Original Research

Opioids are effective analgesics that are widely accepted as therapy for cancer pain and pain related to other terminal illnesses.^1-2 However, the use of opioids to treat chronic nonmalignant pain (CNMP) is controversial.^3-8 Few clinical studies of opioids in the alleviation of CNMP have been conducted, and most have been small, retrospective, uncontrolled, or focused on patients seen in referral settings.^9-21 Together these studies suggest that opioids may benefit certain patients with CNMP, though the results have not been conclusive.

In clinical practice the absence of definitive data on the risks and benefits of opioids for CNMP presents a dilemma. Decisions about potency, frequency, and duration of treatment must be made without the benefit of evidence-based guidelines and with the knowledge that state medical boards or other legal authorities may scrutinize opioid prescriptions. We conducted this study to learn more about attitudes, prescribing practices, and factors associated with the willingness of primary care physicians to prescribe opioids for their patients with CNMP.

Methods

Sample

The University of California, San Francisco/Stanford Collaborative Research Network (CRN) is a practice-based research network composed predominantly of family physicians practicing in Northern and Central California. In 1997 the CRN conducted this survey of all 230 primary care physician members who were not involved in designing our study. Up to 2 mailed reminders and 3 telephone calls were made to initial nonresponders to improve the response rate.

Instrument

The survey instrument was developed through a collaborative process involving 7 volunteer physicians from the CRN. It was pilot-tested and refined using focus groups of practicing non-CRN primary care physicians.

On the first page of the survey, CNMP was defined as pain lasting longer than 6 months that was not related to cancer or another condition expected to end a patient’s life within 6 months. The survey included 3 clinical vignettes Table 1 designed to evoke responses to a variety of patient characteristics, such as medical history, age, sex, and socioeconomic status. Each vignette was followed by a set of specific questions. The survey also contained questions unrelated to the vignettes, regarding general attitudes toward opioids and opioid prescribing practices. We asked about documentation practices, referral resources, and familiarity with state guidelines. The respondents were also queried about personal, patient, and practice characteristics.

Statistical Analysis

We conducted analyses using SAS software.²² Means and standard deviations (SDs) for continuous variables, and frequency distributions for categorical variables, were calculated to summarize physician respondent characteristics, estimates of the characteristics of their caseloads, and summaries of their responses to questions about the clinical vignettes. We used correlation coefficients to examine the strength of relationships between attitudes and practice. The results of the correlation coefficients were used to choose a set of independent variables that were most predictive of willingness to prescribe opioids for CNMP.

We examined with stepwise linear regression the association between willingness to prescribe opioid medications and specific physician characteristics, including year of medical school graduation, size of patient caseload, and concerns about physical dependence, tolerance, addiction, side effects, regulatory scrutiny, and diversion for illegal use. In selecting the final set of variables for the stepwise linear regression predicting willingness to prescribe opioids for CNMP, we found that concern about physical dependence, tolerance, and addiction were highly intercorrelated. Among these variables, concern about physical dependence was the most consistently predictive of willingness to prescribe opioids for CNMP. When concern about physical dependence was entered into stepwise models the variables measuring concern about tolerance and addiction dropped out. Therefore, we chose to use concern about physical dependence as a proxy for measuring generalized concerns about all 3 concerns taken together.

We constructed 3 models to predict willingness to prescribe opioids. In Model 1 the dependent variable to designate willingness to prescribe was constructed from the sum of responses to the 5-point Likert-scaled question asked after each of the 3 vignettes: “If the pain persisted unchanged, would you prescribe opioids for this patient on a long-term basis?” In Model 2 the dependent variable was defined according to the range of agreement on a 5-point Likert scale with the following statement: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids not requiring triplicates (such as Tylenol with codeine) on an as-needed basis.” In Model 3, the dependent variable was defined according to the range of agreement with the following statement on a 5-point Likert scale: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids requiring triplicates (eg, fentanyl patch, methadone, or sustained-release morphine) on a fixed, around-the-clock basis.”

Results

Physician and Practice Characteristics

A total of 161 of 230 physicians (70%) completed the survey. The demographic characteristics of the respondents are presented in Table 2. Table 3 shows physician estimates of patient demographics in their practices. As a group the CRN physicians were mostly white men, but they care for an ethnically, financially, and age-diverse population. The large SDs in Table 3 reflect the wide variety of practice types included in the CRN membership.

Physicians reported seeing an average of 280 patients (SD=157), including 18 CNMP patients (SD=26), per month. An average of 7 patients (SD=8) with CNMP were prescribed opioid analgesics per month, and 90% of the physicians reported prescribing opioids for CNMP at least once a month. The wide SDs again reflect broad variation in the number of patients seen, the number of patients encountered with CNMP, and the number of patients treated with opioid analgesics by different physicians.

Attitudes and Practices of Physicians

Only 15% of respondents agreed with the statement: “I enjoy working with patients who have CNMP.” However, only 15% also felt that daily opioids have no place in the treatment of CNMP. Only 7% agreed with the statement: “I never prescribe opioids for CNMP.”

Many physicians wait for their patients to bring up the subject of opioid treatment, as indicated by the fact that 41% of the respondents agreed that “most of my patients who get opioid prescriptions from me for CNMP requested an opioid before I suggested their use.” In addition, 37% responded that they rarely or never are the first physician to prescribe opioids to their patients with CNMP, possibly waiting for other specialists to take the initiative.

The responses to questions about the 3 clinical vignettes are presented in Table 4. Nearly all physicians felt that the vignettes were realistic, and most believed they were knowledgeable about evaluation and treatment for these patients. However, each case generated substantial variation of opinion with regard to the level of optimism about being able to help the patient, the need for specialty referral, and the willingness to treat with opioids. For each vignette respondents were generally more concerned about physical dependence, tolerance, and addiction than they were about diversion for illegal use, regulatory scrutiny, or side effects. However, physicians’ level of concern about each of these outcomes varied substantially for each vignette.

The physicians were asked general questions about situations in which they would never prescribe opioids. Although none of the respondents said that they had a policy of refusing opioids to patients aged older than 65 years, 19% said they would never prescribe opioids to a child younger than 18 years. In addition, 16% said they would never prescribe opioids to a previous substance abuser, and 42% said they would never prescribe opioids to a current substance abuser, even if recommended by an appropriate specialist.

Also, respondents expressed an increased reluctance to prescribe opioids to CNMP patients as the frequency and potency of the medication was increased. Although only 2% of physicians said they would never prescribe low-potency (schedule III) opioids on an as-needed basis, 35% said they would never prescribe high-potency (schedule II) opioids around the clock, even after exhaustive evaluation and attempts at treatment.

In addition, the willingness of respondents to prescribe opioids varied according to the medical condition being treated. Forty-two percent of respondents said they would never prescribe long-acting schedule II opioids to a patient with post-herpetic neuralgia; 57% would never prescribe them for chronic low back pain; and 75% would never prescribe them for chronic daily headache.

We asked about the use of specialists to assist in the evaluation and treatment of patients who may benefit from opioid treatment for CNMP. Fifty-two percent of the physicians reported always or usually requiring their patients to undergo evaluation by a specialist before prescribing opioids on an ongoing basis for CNMP. Yet only 55% felt they had adequate consultation and referral resources to assist with patients who have CNMP. In addition, only 29% felt they had adequate consultation and referral resources in their communities to assist them with patients who might be abusing or selling opioid prescriptions.

Familiarity with State Prescribing and Documentation Guidelines

In 1994, the Medical Board of California issued guidelines for prescribing opioids for CNMP that were designed to standardize referral and documentation practices and to reduce fear of regulatory scrutiny among physicians who prescribe opioids for CNMP. The guidelines were mailed to all licensed physicians in the state on 3 occasions between 1994 and 1996.²³ We found that 39% of respondents remembered reading the guidelines 1 year after the third mailing. We also found that physicians varied in their self-reported compliance with recommended documentation practices. Ninety percent said they always or usually document a history and physical examination before prescribing opioids, and 86% document periodic reassessment of chronic pain. However, only 60% said that they always or usually document rules of use and misuse of opioid medications; 45% document treatment objectives; and 24% document informed consent. When asked about regulatory scrutiny, 40% of physicians agreed that fear of legal investigation tempers their use of opioids for patients with CNMP.

Predictors of Willingness to Prescribe Opioids

Three models were postulated to clarify the determinants of willingness to prescribe opioids for CNMP. The results of these analyses are presented in Table 5. The stepwise linear regression for each model generated a value for each variable (R2) that represents the proportion of the variance that can be explained by the given variable.

In all 3 models lower levels of concern about physical dependence in response to the vignettes were associated with greater willingness to prescribe opioids. Other variables that were significant predictors of willingness to prescribe opioids in 1 or more models were more recent graduation from medical school, enjoyment in working with chronic pain patients, less fear of regulatory scrutiny, and fewer total patients seen per month.

Discussion

Nearly all the physicians in our sample were willing to treat certain CNMP patients with schedule III opioids on an as-needed basis. However, a third of these physicians said they never use the more potent long-acting schedule II opioids for CNMP. There was also substantial disagreement about which patients would benefit from opioids and which might be likely to suffer adverse effects.

Concern about physical dependence appears to be among the most important barriers to the use of opioids for patients with CNMP. Whether this is always an appropriate concern is debatable. For example, in the case of using schedule III opioids on an as-needed basis, the lack of continuous exposure should limit the risk of physical dependence.

Our finding that physician concerns about physical dependence, tolerance, and addiction were highly intercorrelated raises the possibility that many physicians believe, correctly or incorrectly, that these 3 conditions are closely related effects of opioids. It is also possible that physicians are unclear about what distinguishes one of these outcomes from another. More research is needed to determine the root of physician concerns about physical dependence, tolerance, and addiction. Although all 3 of these outcomes can result when opioids are used around the clock, they nonetheless do not always occur together or necessarily all have equally serious implications when they occur.²⁴ Only a slight majority of respondents felt that they had adequate consultation and referral resources in their community to assist with patients who have CNMP. Primary care physicians may benefit from more information about pain management resources in their communities. In addition, communities without these resources may benefit from the development of pain management centers that can assist primary care physicians with patients who suffer from CNMP.

More recent graduation from medical school was a predictor of increased willingness to prescribe opioids. Recently trained physicians may be more likely to have been exposed to an environment of more liberal use of opioids for CNMP. Conversely, the decreased willingness of more experienced physicians to prescribe opioids may be influenced by their clinical experiences with the complications of opioid use.

Fear of regulatory scrutiny also appeared to limit willingness to prescribe as-needed low-potency schedule III opioids. Recent laws and guidelines have attempted to reduce both the risk and fear of regulatory scrutiny when opioids are prescribed for chronic pain.^23,25,26 However, it is not clear whether awareness of these guidelines would increase or decrease physician concern with regard to regulatory scrutiny, since many physicians reported that their documentation standards are not up to those recommended. In addition, we found no differences in willingness to prescribe opioids based on awareness of the guidelines in California.

We found that physicians who saw more patients were less likely to use more potent opioids. In California, schedule II opioids must be prescribed using triplicate forms purchased from the state. Physicians with high-volume practices may be less inclined to prescribe high-potency opioids because of the time required to complete triplicate forms. Other possible explanations are that these physicians have had more adverse experiences with the use of opioids for CNMP or that they feel more vulnerable to regulatory scrutiny because of their increased volume of patients who might receive opioid prescriptions.

We found that most physicians did not enjoy working with chronic pain patients, and this lack of enjoyment with treating CNMP was a significant barrier to willingness to prescribe opioids in 2 of our 3 models. More investigation of why most physicians do not enjoy working with these patients could further illuminate barriers to the use of opioids for CNMP.

Limitations

There are several limitations to our study. First, the physicians surveyed may not be representative of all practicing primary care physicians. However, CRN physicians are quite similar in many characteristics to family physicians practicing in California.²⁷ Barriers to prescribing opioids in California may also be different from barriers faced by physicians in other parts of the country, so our results may not be easily generalized to other geographic regions. In addition, the data were generated by self-report, and actual practices may differ.

However, our findings are consistent with a 1991 survey of 90 Wisconsin physicians that concluded that concerns about addiction outweigh concerns about regulatory scrutiny for most physicians.²⁸ A national survey of 1912 physicians from multiple specialties found, as we did, a high level of intercorrelation among physician concerns about physical dependence, tolerance, and addiction.²⁹ Unfortunately, that study was not designed to elucidate the relative importance of factors that determine a physician’s willingness to prescribe opioids.

Another limitation of our study is that none of the models we postulated could explain more than a small proportion (24%) of the total variance in the willingness to prescribe opioids for CNMP. Clearly other factors, unmeasured in the current study, also influence physicians’ willingness to prescribe opioids for CNMP. For example, in a study of the prescribing habits at a referral center in Seattle, pain specialists were significantly influenced in their willingness to prescribe opioids for CNMP by a set of pain behaviors exhibited by the patient.³⁰ These behaviors included distorted ambulation or posture, negative affect, facial and audible expressions of distress, and avoidance of activity. The nature of our study did not allow for such factors in our models of willingness to prescribe opioids, but these factors may be worthy of further investigation in direct observation studies of primary care.

Conclusions

Our results suggest that primary care physicians disagree about the relative risks and benefits of opioids in the treatment of individuals who suffer from CNMP. Also, these physicians function with limited reliable information or specialty resources to guide them in choosing which of these patients to treat with opioids. Concerns about addiction, tolerance, and physical dependence appear to be important barriers to the use of opioids by many physicians. More research is needed in primary care settings to determine appropriate uses for opioids in the treatment of CNMP and to further elucidate the concerns of physicians and barriers to more effective use.

Acknowledgments

This research was partially supported by grant #5D32PE19036-09 from the Department of Health and Human Services Health Resources Services Administration to support the establishment of a Department of Family Practice and by a grant from the California Academy of Family Physicians. We would like to acknowledge Dr Eric Sanford, Dr Lawrence Bruguera, Dr Charles Kano, Dr Joyce Hightower, and Dr Yeva Johnson for their assistance with data interpretation and preparation of this manuscript. In addition, we would like to acknowledge the work of Ms Catherine Brosnan and Ms Elizabeth Dito in assisting us with coordination and data collection for our study.

Related resources

• American Pain Society http://www.ampainsoc.org/
• American Academy of Pain Management http://www.aapainmanage.org
• American Pain Foundation http://www.painfoundation.org

Opioids are effective analgesics that are widely accepted as therapy for cancer pain and pain related to other terminal illnesses.^1-2 However, the use of opioids to treat chronic nonmalignant pain (CNMP) is controversial.^3-8 Few clinical studies of opioids in the alleviation of CNMP have been conducted, and most have been small, retrospective, uncontrolled, or focused on patients seen in referral settings.^9-21 Together these studies suggest that opioids may benefit certain patients with CNMP, though the results have not been conclusive.

In clinical practice the absence of definitive data on the risks and benefits of opioids for CNMP presents a dilemma. Decisions about potency, frequency, and duration of treatment must be made without the benefit of evidence-based guidelines and with the knowledge that state medical boards or other legal authorities may scrutinize opioid prescriptions. We conducted this study to learn more about attitudes, prescribing practices, and factors associated with the willingness of primary care physicians to prescribe opioids for their patients with CNMP.

Methods

Sample

The University of California, San Francisco/Stanford Collaborative Research Network (CRN) is a practice-based research network composed predominantly of family physicians practicing in Northern and Central California. In 1997 the CRN conducted this survey of all 230 primary care physician members who were not involved in designing our study. Up to 2 mailed reminders and 3 telephone calls were made to initial nonresponders to improve the response rate.

Instrument

The survey instrument was developed through a collaborative process involving 7 volunteer physicians from the CRN. It was pilot-tested and refined using focus groups of practicing non-CRN primary care physicians.

On the first page of the survey, CNMP was defined as pain lasting longer than 6 months that was not related to cancer or another condition expected to end a patient’s life within 6 months. The survey included 3 clinical vignettes Table 1 designed to evoke responses to a variety of patient characteristics, such as medical history, age, sex, and socioeconomic status. Each vignette was followed by a set of specific questions. The survey also contained questions unrelated to the vignettes, regarding general attitudes toward opioids and opioid prescribing practices. We asked about documentation practices, referral resources, and familiarity with state guidelines. The respondents were also queried about personal, patient, and practice characteristics.

Statistical Analysis

We conducted analyses using SAS software.²² Means and standard deviations (SDs) for continuous variables, and frequency distributions for categorical variables, were calculated to summarize physician respondent characteristics, estimates of the characteristics of their caseloads, and summaries of their responses to questions about the clinical vignettes. We used correlation coefficients to examine the strength of relationships between attitudes and practice. The results of the correlation coefficients were used to choose a set of independent variables that were most predictive of willingness to prescribe opioids for CNMP.

We examined with stepwise linear regression the association between willingness to prescribe opioid medications and specific physician characteristics, including year of medical school graduation, size of patient caseload, and concerns about physical dependence, tolerance, addiction, side effects, regulatory scrutiny, and diversion for illegal use. In selecting the final set of variables for the stepwise linear regression predicting willingness to prescribe opioids for CNMP, we found that concern about physical dependence, tolerance, and addiction were highly intercorrelated. Among these variables, concern about physical dependence was the most consistently predictive of willingness to prescribe opioids for CNMP. When concern about physical dependence was entered into stepwise models the variables measuring concern about tolerance and addiction dropped out. Therefore, we chose to use concern about physical dependence as a proxy for measuring generalized concerns about all 3 concerns taken together.

We constructed 3 models to predict willingness to prescribe opioids. In Model 1 the dependent variable to designate willingness to prescribe was constructed from the sum of responses to the 5-point Likert-scaled question asked after each of the 3 vignettes: “If the pain persisted unchanged, would you prescribe opioids for this patient on a long-term basis?” In Model 2 the dependent variable was defined according to the range of agreement on a 5-point Likert scale with the following statement: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids not requiring triplicates (such as Tylenol with codeine) on an as-needed basis.” In Model 3, the dependent variable was defined according to the range of agreement with the following statement on a 5-point Likert scale: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids requiring triplicates (eg, fentanyl patch, methadone, or sustained-release morphine) on a fixed, around-the-clock basis.”

Results

Physician and Practice Characteristics

A total of 161 of 230 physicians (70%) completed the survey. The demographic characteristics of the respondents are presented in Table 2. Table 3 shows physician estimates of patient demographics in their practices. As a group the CRN physicians were mostly white men, but they care for an ethnically, financially, and age-diverse population. The large SDs in Table 3 reflect the wide variety of practice types included in the CRN membership.

Physicians reported seeing an average of 280 patients (SD=157), including 18 CNMP patients (SD=26), per month. An average of 7 patients (SD=8) with CNMP were prescribed opioid analgesics per month, and 90% of the physicians reported prescribing opioids for CNMP at least once a month. The wide SDs again reflect broad variation in the number of patients seen, the number of patients encountered with CNMP, and the number of patients treated with opioid analgesics by different physicians.

Attitudes and Practices of Physicians

Only 15% of respondents agreed with the statement: “I enjoy working with patients who have CNMP.” However, only 15% also felt that daily opioids have no place in the treatment of CNMP. Only 7% agreed with the statement: “I never prescribe opioids for CNMP.”

Many physicians wait for their patients to bring up the subject of opioid treatment, as indicated by the fact that 41% of the respondents agreed that “most of my patients who get opioid prescriptions from me for CNMP requested an opioid before I suggested their use.” In addition, 37% responded that they rarely or never are the first physician to prescribe opioids to their patients with CNMP, possibly waiting for other specialists to take the initiative.

The responses to questions about the 3 clinical vignettes are presented in Table 4. Nearly all physicians felt that the vignettes were realistic, and most believed they were knowledgeable about evaluation and treatment for these patients. However, each case generated substantial variation of opinion with regard to the level of optimism about being able to help the patient, the need for specialty referral, and the willingness to treat with opioids. For each vignette respondents were generally more concerned about physical dependence, tolerance, and addiction than they were about diversion for illegal use, regulatory scrutiny, or side effects. However, physicians’ level of concern about each of these outcomes varied substantially for each vignette.

The physicians were asked general questions about situations in which they would never prescribe opioids. Although none of the respondents said that they had a policy of refusing opioids to patients aged older than 65 years, 19% said they would never prescribe opioids to a child younger than 18 years. In addition, 16% said they would never prescribe opioids to a previous substance abuser, and 42% said they would never prescribe opioids to a current substance abuser, even if recommended by an appropriate specialist.

Also, respondents expressed an increased reluctance to prescribe opioids to CNMP patients as the frequency and potency of the medication was increased. Although only 2% of physicians said they would never prescribe low-potency (schedule III) opioids on an as-needed basis, 35% said they would never prescribe high-potency (schedule II) opioids around the clock, even after exhaustive evaluation and attempts at treatment.

In addition, the willingness of respondents to prescribe opioids varied according to the medical condition being treated. Forty-two percent of respondents said they would never prescribe long-acting schedule II opioids to a patient with post-herpetic neuralgia; 57% would never prescribe them for chronic low back pain; and 75% would never prescribe them for chronic daily headache.

We asked about the use of specialists to assist in the evaluation and treatment of patients who may benefit from opioid treatment for CNMP. Fifty-two percent of the physicians reported always or usually requiring their patients to undergo evaluation by a specialist before prescribing opioids on an ongoing basis for CNMP. Yet only 55% felt they had adequate consultation and referral resources to assist with patients who have CNMP. In addition, only 29% felt they had adequate consultation and referral resources in their communities to assist them with patients who might be abusing or selling opioid prescriptions.

Familiarity with State Prescribing and Documentation Guidelines

In 1994, the Medical Board of California issued guidelines for prescribing opioids for CNMP that were designed to standardize referral and documentation practices and to reduce fear of regulatory scrutiny among physicians who prescribe opioids for CNMP. The guidelines were mailed to all licensed physicians in the state on 3 occasions between 1994 and 1996.²³ We found that 39% of respondents remembered reading the guidelines 1 year after the third mailing. We also found that physicians varied in their self-reported compliance with recommended documentation practices. Ninety percent said they always or usually document a history and physical examination before prescribing opioids, and 86% document periodic reassessment of chronic pain. However, only 60% said that they always or usually document rules of use and misuse of opioid medications; 45% document treatment objectives; and 24% document informed consent. When asked about regulatory scrutiny, 40% of physicians agreed that fear of legal investigation tempers their use of opioids for patients with CNMP.

Predictors of Willingness to Prescribe Opioids

Three models were postulated to clarify the determinants of willingness to prescribe opioids for CNMP. The results of these analyses are presented in Table 5. The stepwise linear regression for each model generated a value for each variable (R2) that represents the proportion of the variance that can be explained by the given variable.

In all 3 models lower levels of concern about physical dependence in response to the vignettes were associated with greater willingness to prescribe opioids. Other variables that were significant predictors of willingness to prescribe opioids in 1 or more models were more recent graduation from medical school, enjoyment in working with chronic pain patients, less fear of regulatory scrutiny, and fewer total patients seen per month.

Discussion

Nearly all the physicians in our sample were willing to treat certain CNMP patients with schedule III opioids on an as-needed basis. However, a third of these physicians said they never use the more potent long-acting schedule II opioids for CNMP. There was also substantial disagreement about which patients would benefit from opioids and which might be likely to suffer adverse effects.

Concern about physical dependence appears to be among the most important barriers to the use of opioids for patients with CNMP. Whether this is always an appropriate concern is debatable. For example, in the case of using schedule III opioids on an as-needed basis, the lack of continuous exposure should limit the risk of physical dependence.

Our finding that physician concerns about physical dependence, tolerance, and addiction were highly intercorrelated raises the possibility that many physicians believe, correctly or incorrectly, that these 3 conditions are closely related effects of opioids. It is also possible that physicians are unclear about what distinguishes one of these outcomes from another. More research is needed to determine the root of physician concerns about physical dependence, tolerance, and addiction. Although all 3 of these outcomes can result when opioids are used around the clock, they nonetheless do not always occur together or necessarily all have equally serious implications when they occur.²⁴ Only a slight majority of respondents felt that they had adequate consultation and referral resources in their community to assist with patients who have CNMP. Primary care physicians may benefit from more information about pain management resources in their communities. In addition, communities without these resources may benefit from the development of pain management centers that can assist primary care physicians with patients who suffer from CNMP.

More recent graduation from medical school was a predictor of increased willingness to prescribe opioids. Recently trained physicians may be more likely to have been exposed to an environment of more liberal use of opioids for CNMP. Conversely, the decreased willingness of more experienced physicians to prescribe opioids may be influenced by their clinical experiences with the complications of opioid use.

Fear of regulatory scrutiny also appeared to limit willingness to prescribe as-needed low-potency schedule III opioids. Recent laws and guidelines have attempted to reduce both the risk and fear of regulatory scrutiny when opioids are prescribed for chronic pain.^23,25,26 However, it is not clear whether awareness of these guidelines would increase or decrease physician concern with regard to regulatory scrutiny, since many physicians reported that their documentation standards are not up to those recommended. In addition, we found no differences in willingness to prescribe opioids based on awareness of the guidelines in California.

We found that physicians who saw more patients were less likely to use more potent opioids. In California, schedule II opioids must be prescribed using triplicate forms purchased from the state. Physicians with high-volume practices may be less inclined to prescribe high-potency opioids because of the time required to complete triplicate forms. Other possible explanations are that these physicians have had more adverse experiences with the use of opioids for CNMP or that they feel more vulnerable to regulatory scrutiny because of their increased volume of patients who might receive opioid prescriptions.

We found that most physicians did not enjoy working with chronic pain patients, and this lack of enjoyment with treating CNMP was a significant barrier to willingness to prescribe opioids in 2 of our 3 models. More investigation of why most physicians do not enjoy working with these patients could further illuminate barriers to the use of opioids for CNMP.

Limitations

There are several limitations to our study. First, the physicians surveyed may not be representative of all practicing primary care physicians. However, CRN physicians are quite similar in many characteristics to family physicians practicing in California.²⁷ Barriers to prescribing opioids in California may also be different from barriers faced by physicians in other parts of the country, so our results may not be easily generalized to other geographic regions. In addition, the data were generated by self-report, and actual practices may differ.

However, our findings are consistent with a 1991 survey of 90 Wisconsin physicians that concluded that concerns about addiction outweigh concerns about regulatory scrutiny for most physicians.²⁸ A national survey of 1912 physicians from multiple specialties found, as we did, a high level of intercorrelation among physician concerns about physical dependence, tolerance, and addiction.²⁹ Unfortunately, that study was not designed to elucidate the relative importance of factors that determine a physician’s willingness to prescribe opioids.

Another limitation of our study is that none of the models we postulated could explain more than a small proportion (24%) of the total variance in the willingness to prescribe opioids for CNMP. Clearly other factors, unmeasured in the current study, also influence physicians’ willingness to prescribe opioids for CNMP. For example, in a study of the prescribing habits at a referral center in Seattle, pain specialists were significantly influenced in their willingness to prescribe opioids for CNMP by a set of pain behaviors exhibited by the patient.³⁰ These behaviors included distorted ambulation or posture, negative affect, facial and audible expressions of distress, and avoidance of activity. The nature of our study did not allow for such factors in our models of willingness to prescribe opioids, but these factors may be worthy of further investigation in direct observation studies of primary care.

Conclusions

Our results suggest that primary care physicians disagree about the relative risks and benefits of opioids in the treatment of individuals who suffer from CNMP. Also, these physicians function with limited reliable information or specialty resources to guide them in choosing which of these patients to treat with opioids. Concerns about addiction, tolerance, and physical dependence appear to be important barriers to the use of opioids by many physicians. More research is needed in primary care settings to determine appropriate uses for opioids in the treatment of CNMP and to further elucidate the concerns of physicians and barriers to more effective use.

Acknowledgments

This research was partially supported by grant #5D32PE19036-09 from the Department of Health and Human Services Health Resources Services Administration to support the establishment of a Department of Family Practice and by a grant from the California Academy of Family Physicians. We would like to acknowledge Dr Eric Sanford, Dr Lawrence Bruguera, Dr Charles Kano, Dr Joyce Hightower, and Dr Yeva Johnson for their assistance with data interpretation and preparation of this manuscript. In addition, we would like to acknowledge the work of Ms Catherine Brosnan and Ms Elizabeth Dito in assisting us with coordination and data collection for our study.

Related resources

• American Pain Society http://www.ampainsoc.org/
• American Academy of Pain Management http://www.aapainmanage.org
• American Pain Foundation http://www.painfoundation.org

Opioids are effective analgesics that are widely accepted as therapy for cancer pain and pain related to other terminal illnesses.^1-2 However, the use of opioids to treat chronic nonmalignant pain (CNMP) is controversial.^3-8 Few clinical studies of opioids in the alleviation of CNMP have been conducted, and most have been small, retrospective, uncontrolled, or focused on patients seen in referral settings.^9-21 Together these studies suggest that opioids may benefit certain patients with CNMP, though the results have not been conclusive.

In clinical practice the absence of definitive data on the risks and benefits of opioids for CNMP presents a dilemma. Decisions about potency, frequency, and duration of treatment must be made without the benefit of evidence-based guidelines and with the knowledge that state medical boards or other legal authorities may scrutinize opioid prescriptions. We conducted this study to learn more about attitudes, prescribing practices, and factors associated with the willingness of primary care physicians to prescribe opioids for their patients with CNMP.

Methods

Sample

The University of California, San Francisco/Stanford Collaborative Research Network (CRN) is a practice-based research network composed predominantly of family physicians practicing in Northern and Central California. In 1997 the CRN conducted this survey of all 230 primary care physician members who were not involved in designing our study. Up to 2 mailed reminders and 3 telephone calls were made to initial nonresponders to improve the response rate.

Instrument

The survey instrument was developed through a collaborative process involving 7 volunteer physicians from the CRN. It was pilot-tested and refined using focus groups of practicing non-CRN primary care physicians.

On the first page of the survey, CNMP was defined as pain lasting longer than 6 months that was not related to cancer or another condition expected to end a patient’s life within 6 months. The survey included 3 clinical vignettes Table 1 designed to evoke responses to a variety of patient characteristics, such as medical history, age, sex, and socioeconomic status. Each vignette was followed by a set of specific questions. The survey also contained questions unrelated to the vignettes, regarding general attitudes toward opioids and opioid prescribing practices. We asked about documentation practices, referral resources, and familiarity with state guidelines. The respondents were also queried about personal, patient, and practice characteristics.

Statistical Analysis

We conducted analyses using SAS software.²² Means and standard deviations (SDs) for continuous variables, and frequency distributions for categorical variables, were calculated to summarize physician respondent characteristics, estimates of the characteristics of their caseloads, and summaries of their responses to questions about the clinical vignettes. We used correlation coefficients to examine the strength of relationships between attitudes and practice. The results of the correlation coefficients were used to choose a set of independent variables that were most predictive of willingness to prescribe opioids for CNMP.

We examined with stepwise linear regression the association between willingness to prescribe opioid medications and specific physician characteristics, including year of medical school graduation, size of patient caseload, and concerns about physical dependence, tolerance, addiction, side effects, regulatory scrutiny, and diversion for illegal use. In selecting the final set of variables for the stepwise linear regression predicting willingness to prescribe opioids for CNMP, we found that concern about physical dependence, tolerance, and addiction were highly intercorrelated. Among these variables, concern about physical dependence was the most consistently predictive of willingness to prescribe opioids for CNMP. When concern about physical dependence was entered into stepwise models the variables measuring concern about tolerance and addiction dropped out. Therefore, we chose to use concern about physical dependence as a proxy for measuring generalized concerns about all 3 concerns taken together.

We constructed 3 models to predict willingness to prescribe opioids. In Model 1 the dependent variable to designate willingness to prescribe was constructed from the sum of responses to the 5-point Likert-scaled question asked after each of the 3 vignettes: “If the pain persisted unchanged, would you prescribe opioids for this patient on a long-term basis?” In Model 2 the dependent variable was defined according to the range of agreement on a 5-point Likert scale with the following statement: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids not requiring triplicates (such as Tylenol with codeine) on an as-needed basis.” In Model 3, the dependent variable was defined according to the range of agreement with the following statement on a 5-point Likert scale: “For patients with CNMP that persists unchanged after exhaustive evaluation and attempts at treatment, I am willing to prescribe opioids requiring triplicates (eg, fentanyl patch, methadone, or sustained-release morphine) on a fixed, around-the-clock basis.”

Results

Physician and Practice Characteristics

A total of 161 of 230 physicians (70%) completed the survey. The demographic characteristics of the respondents are presented in Table 2. Table 3 shows physician estimates of patient demographics in their practices. As a group the CRN physicians were mostly white men, but they care for an ethnically, financially, and age-diverse population. The large SDs in Table 3 reflect the wide variety of practice types included in the CRN membership.

Physicians reported seeing an average of 280 patients (SD=157), including 18 CNMP patients (SD=26), per month. An average of 7 patients (SD=8) with CNMP were prescribed opioid analgesics per month, and 90% of the physicians reported prescribing opioids for CNMP at least once a month. The wide SDs again reflect broad variation in the number of patients seen, the number of patients encountered with CNMP, and the number of patients treated with opioid analgesics by different physicians.

Attitudes and Practices of Physicians

Only 15% of respondents agreed with the statement: “I enjoy working with patients who have CNMP.” However, only 15% also felt that daily opioids have no place in the treatment of CNMP. Only 7% agreed with the statement: “I never prescribe opioids for CNMP.”

Many physicians wait for their patients to bring up the subject of opioid treatment, as indicated by the fact that 41% of the respondents agreed that “most of my patients who get opioid prescriptions from me for CNMP requested an opioid before I suggested their use.” In addition, 37% responded that they rarely or never are the first physician to prescribe opioids to their patients with CNMP, possibly waiting for other specialists to take the initiative.

The responses to questions about the 3 clinical vignettes are presented in Table 4. Nearly all physicians felt that the vignettes were realistic, and most believed they were knowledgeable about evaluation and treatment for these patients. However, each case generated substantial variation of opinion with regard to the level of optimism about being able to help the patient, the need for specialty referral, and the willingness to treat with opioids. For each vignette respondents were generally more concerned about physical dependence, tolerance, and addiction than they were about diversion for illegal use, regulatory scrutiny, or side effects. However, physicians’ level of concern about each of these outcomes varied substantially for each vignette.

The physicians were asked general questions about situations in which they would never prescribe opioids. Although none of the respondents said that they had a policy of refusing opioids to patients aged older than 65 years, 19% said they would never prescribe opioids to a child younger than 18 years. In addition, 16% said they would never prescribe opioids to a previous substance abuser, and 42% said they would never prescribe opioids to a current substance abuser, even if recommended by an appropriate specialist.

Also, respondents expressed an increased reluctance to prescribe opioids to CNMP patients as the frequency and potency of the medication was increased. Although only 2% of physicians said they would never prescribe low-potency (schedule III) opioids on an as-needed basis, 35% said they would never prescribe high-potency (schedule II) opioids around the clock, even after exhaustive evaluation and attempts at treatment.

In addition, the willingness of respondents to prescribe opioids varied according to the medical condition being treated. Forty-two percent of respondents said they would never prescribe long-acting schedule II opioids to a patient with post-herpetic neuralgia; 57% would never prescribe them for chronic low back pain; and 75% would never prescribe them for chronic daily headache.

We asked about the use of specialists to assist in the evaluation and treatment of patients who may benefit from opioid treatment for CNMP. Fifty-two percent of the physicians reported always or usually requiring their patients to undergo evaluation by a specialist before prescribing opioids on an ongoing basis for CNMP. Yet only 55% felt they had adequate consultation and referral resources to assist with patients who have CNMP. In addition, only 29% felt they had adequate consultation and referral resources in their communities to assist them with patients who might be abusing or selling opioid prescriptions.

Familiarity with State Prescribing and Documentation Guidelines

In 1994, the Medical Board of California issued guidelines for prescribing opioids for CNMP that were designed to standardize referral and documentation practices and to reduce fear of regulatory scrutiny among physicians who prescribe opioids for CNMP. The guidelines were mailed to all licensed physicians in the state on 3 occasions between 1994 and 1996.²³ We found that 39% of respondents remembered reading the guidelines 1 year after the third mailing. We also found that physicians varied in their self-reported compliance with recommended documentation practices. Ninety percent said they always or usually document a history and physical examination before prescribing opioids, and 86% document periodic reassessment of chronic pain. However, only 60% said that they always or usually document rules of use and misuse of opioid medications; 45% document treatment objectives; and 24% document informed consent. When asked about regulatory scrutiny, 40% of physicians agreed that fear of legal investigation tempers their use of opioids for patients with CNMP.

Predictors of Willingness to Prescribe Opioids

Three models were postulated to clarify the determinants of willingness to prescribe opioids for CNMP. The results of these analyses are presented in Table 5. The stepwise linear regression for each model generated a value for each variable (R2) that represents the proportion of the variance that can be explained by the given variable.

In all 3 models lower levels of concern about physical dependence in response to the vignettes were associated with greater willingness to prescribe opioids. Other variables that were significant predictors of willingness to prescribe opioids in 1 or more models were more recent graduation from medical school, enjoyment in working with chronic pain patients, less fear of regulatory scrutiny, and fewer total patients seen per month.

Discussion

Nearly all the physicians in our sample were willing to treat certain CNMP patients with schedule III opioids on an as-needed basis. However, a third of these physicians said they never use the more potent long-acting schedule II opioids for CNMP. There was also substantial disagreement about which patients would benefit from opioids and which might be likely to suffer adverse effects.

Concern about physical dependence appears to be among the most important barriers to the use of opioids for patients with CNMP. Whether this is always an appropriate concern is debatable. For example, in the case of using schedule III opioids on an as-needed basis, the lack of continuous exposure should limit the risk of physical dependence.

Our finding that physician concerns about physical dependence, tolerance, and addiction were highly intercorrelated raises the possibility that many physicians believe, correctly or incorrectly, that these 3 conditions are closely related effects of opioids. It is also possible that physicians are unclear about what distinguishes one of these outcomes from another. More research is needed to determine the root of physician concerns about physical dependence, tolerance, and addiction. Although all 3 of these outcomes can result when opioids are used around the clock, they nonetheless do not always occur together or necessarily all have equally serious implications when they occur.²⁴ Only a slight majority of respondents felt that they had adequate consultation and referral resources in their community to assist with patients who have CNMP. Primary care physicians may benefit from more information about pain management resources in their communities. In addition, communities without these resources may benefit from the development of pain management centers that can assist primary care physicians with patients who suffer from CNMP.

More recent graduation from medical school was a predictor of increased willingness to prescribe opioids. Recently trained physicians may be more likely to have been exposed to an environment of more liberal use of opioids for CNMP. Conversely, the decreased willingness of more experienced physicians to prescribe opioids may be influenced by their clinical experiences with the complications of opioid use.

Fear of regulatory scrutiny also appeared to limit willingness to prescribe as-needed low-potency schedule III opioids. Recent laws and guidelines have attempted to reduce both the risk and fear of regulatory scrutiny when opioids are prescribed for chronic pain.^23,25,26 However, it is not clear whether awareness of these guidelines would increase or decrease physician concern with regard to regulatory scrutiny, since many physicians reported that their documentation standards are not up to those recommended. In addition, we found no differences in willingness to prescribe opioids based on awareness of the guidelines in California.

We found that physicians who saw more patients were less likely to use more potent opioids. In California, schedule II opioids must be prescribed using triplicate forms purchased from the state. Physicians with high-volume practices may be less inclined to prescribe high-potency opioids because of the time required to complete triplicate forms. Other possible explanations are that these physicians have had more adverse experiences with the use of opioids for CNMP or that they feel more vulnerable to regulatory scrutiny because of their increased volume of patients who might receive opioid prescriptions.

We found that most physicians did not enjoy working with chronic pain patients, and this lack of enjoyment with treating CNMP was a significant barrier to willingness to prescribe opioids in 2 of our 3 models. More investigation of why most physicians do not enjoy working with these patients could further illuminate barriers to the use of opioids for CNMP.

Limitations

There are several limitations to our study. First, the physicians surveyed may not be representative of all practicing primary care physicians. However, CRN physicians are quite similar in many characteristics to family physicians practicing in California.²⁷ Barriers to prescribing opioids in California may also be different from barriers faced by physicians in other parts of the country, so our results may not be easily generalized to other geographic regions. In addition, the data were generated by self-report, and actual practices may differ.

However, our findings are consistent with a 1991 survey of 90 Wisconsin physicians that concluded that concerns about addiction outweigh concerns about regulatory scrutiny for most physicians.²⁸ A national survey of 1912 physicians from multiple specialties found, as we did, a high level of intercorrelation among physician concerns about physical dependence, tolerance, and addiction.²⁹ Unfortunately, that study was not designed to elucidate the relative importance of factors that determine a physician’s willingness to prescribe opioids.

Another limitation of our study is that none of the models we postulated could explain more than a small proportion (24%) of the total variance in the willingness to prescribe opioids for CNMP. Clearly other factors, unmeasured in the current study, also influence physicians’ willingness to prescribe opioids for CNMP. For example, in a study of the prescribing habits at a referral center in Seattle, pain specialists were significantly influenced in their willingness to prescribe opioids for CNMP by a set of pain behaviors exhibited by the patient.³⁰ These behaviors included distorted ambulation or posture, negative affect, facial and audible expressions of distress, and avoidance of activity. The nature of our study did not allow for such factors in our models of willingness to prescribe opioids, but these factors may be worthy of further investigation in direct observation studies of primary care.

Conclusions

Our results suggest that primary care physicians disagree about the relative risks and benefits of opioids in the treatment of individuals who suffer from CNMP. Also, these physicians function with limited reliable information or specialty resources to guide them in choosing which of these patients to treat with opioids. Concerns about addiction, tolerance, and physical dependence appear to be important barriers to the use of opioids by many physicians. More research is needed in primary care settings to determine appropriate uses for opioids in the treatment of CNMP and to further elucidate the concerns of physicians and barriers to more effective use.

Acknowledgments

This research was partially supported by grant #5D32PE19036-09 from the Department of Health and Human Services Health Resources Services Administration to support the establishment of a Department of Family Practice and by a grant from the California Academy of Family Physicians. We would like to acknowledge Dr Eric Sanford, Dr Lawrence Bruguera, Dr Charles Kano, Dr Joyce Hightower, and Dr Yeva Johnson for their assistance with data interpretation and preparation of this manuscript. In addition, we would like to acknowledge the work of Ms Catherine Brosnan and Ms Elizabeth Dito in assisting us with coordination and data collection for our study.

Related resources

• American Pain Society http://www.ampainsoc.org/
• American Academy of Pain Management http://www.aapainmanage.org
• American Pain Foundation http://www.painfoundation.org

Clinical practice guidelines on smoking cessation1 advocate that clinicians identify all smokers, advise them to quit, and arrange follow-up care. Arranging systematic follow-up care is often the most difficult of those steps in a primary medical practice because counseling for smoking cessation is often not reimbursed.² Telephone support counseling services offering proactive follow-up with scheduled sessions have achieved long-term success rates from 25% to 30%.^3-8 We were able to achieve a long-term quit rate of 36% in a community-based trial of computer-assisted telephone support counseling by nurses and telephone counselors trained in computer skills and relapse prevention.⁹ In this study more than 57% of the practice-based participants were covered by Medicaid insurance. There were no statistically significant differences in quit rates for Medicaid (33%) and non-Medicaid (36%) smokers at 6 months using a community denominator analysis approach.⁹

Managed care provides an advantageous system for the delivery of preventive services.¹⁰ Most indemnity insurance plans cover few preventive services, mostly limited to screening and immunizations, despite “findings…that the counseling and education services are among the most effective interventions available to clinicians to achieve the goals of health promotion and disease prevention.”¹⁰ Group Health Cooperative (GHC) of Puget Sound has demonstrated with a comprehensive systematic population-based health care approach that the prevalence of smoking can be reduced from 25% to 15.5% over 10 years among more than 550,000 adult enrollees.¹¹ This tremendous change within a population was achieved by multiple approaches, including identification, tracking, community outreach, comprehensive clinician and staff education, free coverage of services to participants, accessible telephone counseling, and self-help materials.¹¹ This is an outstanding example of an effective comprehensive program on smoking cessation within a private managed care system resulting in the overall reduction of smoking for a large population. It serves as a model for other preventive services concerning such common topics as alcohol consumption abuse, cancer screening, or coronary artery disease.

The prevalence of smoking among Medicaid health maintenance organizations (HMOs) versus commercial HMO participants is reported to be much higher. In recent Michigan surveys of health plans, 19.4% of the participants in commercial HMOs reported being current smokers compared with 44.1% of Medicaid HMO participants.¹² Medicaid participants are clearly a high-risk population for tobacco use and the medical consequences of smoking. Many states are moving from a fee-for-service approach to Medicaid coverage as a prospective capitated payment approach within managed care. Before the course of our study all Medicaid participants were moved into managed care plans by the State of Michigan. This context provided the ideal setting for examination of the impact of a systematic approach to smoking cessation by office-based and telephone counseling follow-up care for smokers covered by Medicaid managed care.

Brief advice on smoking cessation from a physician alone results in long-term quit rates of less than 10%.¹³ With the supplementation of brief physician advice with higher-dose nicotine gum or transdermal nicotine in randomized-controlled settings, long-term quit rates are increased to 15% to 25%.^14-16 In the context of community practice relying on general volunteers, long-term quit rates are lower than strictly controlled trials.¹⁷ It seems that pharmacotherapy clearly enhances brief advice by physicians for smoking cessation.¹⁶ In a primary care medical practice-based study, Daughton and colleagues¹⁸ state that “data clearly indicate that counseling seems to maximize smoking cessation rates with the nicotine patch.” Most studies (including that by Daughton and coworkers) have examined the relative effectiveness of pharmacotherapy against placebo. Our study proposes to answer the following questions: What is the comparative efficacy in quit rates by adding nurse and telephone counseling support for follow-up care to physician advice alone when all smokers receive the same pharmacotherapy? Does added behavioral support actually improve quit rates when all smokers use pharmacotherapy, or is there no difference? Can significant quit rates be achieved in low-income populations? Are there special measures required to maintain follow-up and protocol compliance in Medicaid smokers? What are the barriers to decreasing the high prevalence of smoking among participants in Medicaid managed care plans?

Methods

Recruitment of Participants

Participants were enrolled from January 19, 1998, to June 20, 1998, from 3 community heath center sites (Hackley, Baldwin, and Muskegon) in Michigan. Each practice site had the designation of a federally underserved site with the majority of care provided to very low-income patients. Each practice site had 5 to 7 providers with approximately 10,000 to 15,000 active patients on record. All participants were smokers older than 21 years with Medicaid managed care insurance. Participants were covered by 4 different managed care plans that agreed to allow the participation of their patients in the study. All participants had no medical contraindications to the use of transdermal nicotine, including pregnancy, and were willing to commit to quitting smoking within the next 30 days. Smokers were invited to participate in the study during their usual office visits and were offered 21-mg transdermal nicotine for 8 weeks as covered by Medicaid and determined appropriate by their providers.

Recruitment Rates

During the 6-month recruitment period, 501 smokers on Medicaid managed care were identified as eligible by office nurses and referred to participate in our study at the 3 practice sites. Of the referred group, 259 (52%) enrolled in our study and were randomized to either usual or relapse prevention care. A total of 233 (48% of the referred group) participated in our study. Participation was defined as receiving brief physician advice for the usual care group or brief physician advice and 1 telephonic-counseling session for the telephonic-counseling care group. These rates of enrollment were consistent across the 3 study sites. Participants were excluded from the final analysis if it was discovered after randomization that they did not have telephones. The informed consent process was approved by the institutional review board of Michigan State University. Figure 1 shows the recruitment and randomization flow.

Training of Providers and Staff

A total of 20 primary care physicians were trained to provide brief advice for smoking cessation consistent with the national guidelines.¹ Physician training consisted of a 2-hour update session on the guidelines, an overview of the study protocol, and role playing. Physicians received continuing medical education credit for participating. Ten nurses^3-4 per site and 10 telephone counselors were trained in computer-assisted relapse prevention. Nurse and telephone-counselor training consisted of 3 2-hour sessions on relapse prevention, computer skills, and individual case management. Nurses and counselors were encouraged to practice on case examples between training sessions. Their intervention skills were evaluated before they began counseling study participants. Quality assurance of counseling performance was performed through weekly audiotape review by research assistants. The computer program (“I’d Rather Cope than Smoke”⁹) provided a continuing record of counseling time and accuracy of data collection per nurse and counselor.

Study Design and Counseling Interventions

The participants who were assigned to the usual-care group participated in an intake session, received brief advice on smoking by their provider according to the guidelines,¹ were given a prescription for transdermal nicotine if medically appropriate, and had a follow-up scheduled for at least 1 visit (usually 7 to 30 days after the quit date) consistent with their medical condition. All participants also received “Clearing the Air” (National Cancer Institute publication no. 95-1647). The intake session lasted approximately 45 minutes. The intake was conducted by study staff (not the nurses providing telephone counseling) to prevent selection bias. Randomized assignment to either usual care or telephonic-counseling care groups occurred immediately after the intake session.

The participants assigned to the relapse prevention telephonic-counseling group received an intake session, usual care, a copy of “Clearing the Air,” a diary of coping responses the size of a cigarette pack, and 6 telephonic-counseling sessions. The sequence of follow-up sessions was determined according to the quit date: Session 1 was scheduled for 1 day after the quit date; session 2, 3 days; session 3, 7 days; session 4, 14 days; session 5, 30 days; and session 6, 60 days. Follow-up sessions lasted approximately 15 to 20 minutes. This sequence is consistent with previously reported studies by the authors⁹ and other investigators.^4,8 As previously reported, the computer software program, “I’d Rather Cope than Smoke,” was developed to assist in compliance with the relapse prevention protocol.⁹ All counseling sessions were done telephonically. Trained office nurses who used the software on laptop computers performed the first 3 treatment sessions. The intake and follow-up data of the first 3 sessions were electronically transferred to a computer network at Michigan State University where trained telephone counselors provided sessions 4 to 6.

Barriers to Maintaining Telephone Treatment

Before the study onset, focus group analysis of low-income smokers reported that the majority preferred counseling sessions on relapse prevention to be done by telephone rather than in person by the office nurse at the practice sites. We anticipated frequent disruptions in telephone service for the study population, so several innovative methods to maintain telephone treatment were developed, such as: (1) immediately contacting directory assistance for disruptions in service; (2) verifying site records for phone numbers changes; (3) contacting participants during subsequent clinic visits to update phone numbers; and (4) mailing a self-addressed stamped postcard requesting immediate feedback.

Independent Variables

Participants were evaluated for standard demographic characteristics of sex, age, socioeconomic status, education level, and working status. Baseline smoking activity was evaluated on the basis of the number of cigarettes smoked per day, the number of years of smoking, the mini-Fagerstrom Tolerance Questionnaire (FTQ),19 household activity, confidence in quitting, and personal reasons for quitting. Medicaid insurance status was verified. Personal patterns of relapse triggers and coping response were recorded.

Outcomes Measured

The main outcome measure was carbon monoxide verified smoke-free status at a telephone follow-up 90 days after the quit date in both usual and telephonic-counseling groups. Multiple attempts were made to contact participants, regardless of the level of participation at 3 months. Participants reporting 7-day smoke-free status at 3 months were invited to have carbon monoxide verification at the office and were paid $50 for their time.

The secondary outcome measures included physician, nurse, counselor, and participant compliance with protocols; provider and staff satisfaction with the program; and nicotine replacement use.

Statistics

Comparisons of study group characteristics were made using standard statistical measures. Categorical variables were tested using the chi-square test for contingency tables and the Student t test for continuous variables. Several continuous variables were categorized and analyzed by both methods.

The study denominator was based on intention-to-treat assignment as in randomized controlled trials^20,21 for evaluation of pharmacotherapy for nicotine addiction. Participants who refused follow-up, failed to call back, gave incorrect contact numbers, or dropped out were counted as smokers.

Smoking quit rates at 90-day follow-ups were compared using the z score for equality of proportions. Adjustments were made in self-reported outcomes based on carbon monoxide verification rates.

Results

Demographic Comparison of Study Groups

A total of 238 smokers participated in the study (N=123 usual care group, and N=110 in the telephonic-counseling group) and patient demographics are reported in Table 2. The smoking characteristics of the study groups are provided in Table 3. Adjustments for participants without telephones did not induce any significant differences.

As shown in Table 4, the most common reasons for quitting by far were personal health reasons and health problems related to smoking. Very few participants reported advice from their physician as the reason to quit smoking. The groups were comparable and did not differ significantly in their reasons for quitting.

Smoke-Free Status

Of the 233 patients with telephones enrolled in the study, 80 (65%) in the usual care group and 74 (67%) in the telephonic-counseling group were successfully contacted. Of those contacted, 19 in the usual care group and 24 in the telephonic-counseling group reported that they were smoke free. However, smoke-free status was successfully confirmed using carbon monoxide (CO) monitoring in only 56% of patients claiming to be smoke free in the usual care group, while 95% of patients in the telephonic-counseling group had their smoke-free status confirmed. Thus, in the per-protocol analysis, smoke-free status was confirmed in 10 of 80 (12.5%) in the usual care group and 23 of 74 (31%) in the telephonic-counseling group (P=.004).

In the intention-to-treat analysis we assumed that all missing cases were smoke free and used denominators of 123 and 110 for the usual care and telephonic-counseling groups, respectively. In this intention-to-treat analysis, the rates of self-reported smoke-free status were 15% and 19% (P=ns). In the intention-to-treat analysis of CO-verified smoke-free status, patients in the telephonic-counseling group were more likely to be smoke free (8.1% vs 21%, P <.01).

Nicotine Replacement Use

Prescriptions for nicotine replacement were received by 91% of the usual care and 99% of the telephonic-counseling care participants. At follow-up evaluation, 73% of the usual care and 67% of telephonic-counseling care participants reported using at least an initial course of nicotine replacement. These proportions of use did not differ significantly between the study groups

Discussion

Smoking has been shown to be one of the most modifiable health risks significantly related to higher health care charges, even after controlling for age, sex, race, diabetes, and heart disease.²² Although indemnity plans have been largely unsupportive of services for smoking cessation counseling, managed care plans have shown considerable success at decreasing the prevalence of smoking by offering comprehensive smoking cessation services.^23,24 In fact, offering full coverage of both behavioral and pharmacotherapy services results in a greater reduction in smoking prevalence than partial coverage.²⁴ The studies mentioned on smoking cessation were conducted with participants who were employed and had commercial insurance coverage.

Our study examined the effectiveness of a comprehensive program for smoking cessation provided by nurse and telephone counselors who were assisted by a computer-guided program focusing on relapse prevention in very low-income smokers covered by Medicaid managed care. The intention-to-treat results of a 21% quit rate at 3 months were consistent with our previously reported study,⁹ which included a sizable subpopulation of Medicaid patients. If adjustments are made in the denominator based on community trials¹⁷ as our previous study⁹ for reasonable loss-to-follow, then the CO-verified quit rates at 3 months would be 13% (usual care) and 31% (telephonic care) (P=.011). Our report is unique because we directly compared the effectiveness of telephone counseling support with usual care (brief physician advice and follow-up) in a true experimental trial in community practice. Though most participants received prescriptions for transdermal nicotine, the variation in usage was similar in both study groups because randomization allows a true comparison of the behavioral intervention effects. The recruitment data showing that approximately 50% of referred smokers in primary care are willing to enroll in a program is consistent with our previous study⁹ and other reports.²⁴ This demonstrates that Medicaid smokers are generally as willing to participate in smoking cessation services as other smokers.

Although all providers received formal training on the smoking cessation guidelines,¹ were aware of the study, and had “green card” reminders on study charts, they offered appropriate follow-up care only 26% of the time at return visits (based on post-study chart audit documentation). These findings are consistent with national surveys of physicians in primary care practices² that show follow-up care as the greatest shortcoming. It seems that physicians need to have comprehensive office systems in place to ensure even brief follow-up care²⁶ for smoking cessation. Telephone counseling support with a guided computer system definitely enhances follow-up care. By closely tracking participants for changes in addresses and telephone services, reasonable follow-up can be maintained even in low-income smokers. In our study, 60% of the participants in the telephonic-counseling group received at least 4 treatment sessions. Opinions of providers and staff during post-study focus groups were very positive. All 3 practices decided to continue a nurse-based approach for relapse prevention counseling after the study and expressed a need for the telephone support services to continue.

Limitations

One of the possible weaknesses of this study is the lack of long-term follow-up at 6 to 12 months for quit rates to ensure continued differences in effectiveness. Because of lack of funding, we were only able to obtain follow-up at 3 months. However, our findings are similar to the data in our previously reported community demonstration trial,⁹ which did not have a usual care comparison. Though the 2 reports refer to different populations, in our previous report⁹ using an intention-to-treat denominator the CO-verified quit rates were approximately 20% at 6 months in the Medicaid population. When using a community-based denominator that accounted for loss to follow-up, the 6-month quit rate was 33%. These results are consistent with strictly controlled trials where the majority of participants used nicotine replacement therapies.¹⁶

It is of interest to note that in this very low-income population, providing $50 to verify self-reported smoking cessation by CO monitor not only yielded considerable follow-up at 3 months but may have biased self-reporting in the usual care group where only 56% of the reports were verified. This finding shows the importance of using biochemical verification of smoking cessation even in community-based clinical trials.

Continued Research

Our study poses several questions for further research. Are the quit rates obtained by the described telephonic-counseling program sustainable over time at 1 to 2 years post-treatment in low-income populations? Can these approaches for relapse prevention be adapted to meet the needs of special groups, such as pregnant smokers, difficult to reach smokers at home, and high-risk smokers with diseases such as diabetes, heart disease, asthma, and severe disabilities when offered in conjunction with disease management services within managed care plans? This is of particular importance when the majority of low-income smokers report personal, smoking-related, and family health problems as reasons for quitting smoking. Though such behavioral support services are reported to be cost-effective in commercial managed care populations,²⁵ what is the cost-effectiveness of these services when adapted to meet the needs of special populations?

Conclusions

Telephonic-counseling for smoking cessation supported by a computer-guided program on relapse prevention is both practical and effective even for low-income smokers covered by Medicaid managed care. Special tracking approaches are required to maintain low-income smokers in treatment and to ensure provider follow-up. State Medicaid programs and insurance plans should consider investing in both office-based and centralized telephonic smoking cessation services to enhance smoking cessation for low-income smokers.

Acknowledgments

Our research was supported by a grant from the Michigan Department of Community Health to the Institute for Managed Care of Michigan State University (MSU) as a subproject on “Cancer Prevention, Outreach and Screening/Detection for Cancer Patients.”

Joseph Farrell, MA, director of the Institute for Managed Care, acted as the overall project director. Barbara Given, PhD, RN, and professor in the College of Nursing at MSU, was overall project manager. Wei Pan, MS, provided data management and statistical support. Kathy Ives, research assistant, contributed project coordination, data entry, and analysis. Dorothy Pathak, PhD, biostatistics consultant, verified the statistical analysis. We thank the Medicaid managed care plans of Wellness, Care Choices, Physicians Health Plan, and Community Choice for allowing patient participation and covering pharmacotherapy during the study. We thank the community health center physicians and staff in the Michigan communities of Hackley, Baldwin, and Muskegon for participating in the study.

Clinical practice guidelines on smoking cessation1 advocate that clinicians identify all smokers, advise them to quit, and arrange follow-up care. Arranging systematic follow-up care is often the most difficult of those steps in a primary medical practice because counseling for smoking cessation is often not reimbursed.² Telephone support counseling services offering proactive follow-up with scheduled sessions have achieved long-term success rates from 25% to 30%.^3-8 We were able to achieve a long-term quit rate of 36% in a community-based trial of computer-assisted telephone support counseling by nurses and telephone counselors trained in computer skills and relapse prevention.⁹ In this study more than 57% of the practice-based participants were covered by Medicaid insurance. There were no statistically significant differences in quit rates for Medicaid (33%) and non-Medicaid (36%) smokers at 6 months using a community denominator analysis approach.⁹

Managed care provides an advantageous system for the delivery of preventive services.¹⁰ Most indemnity insurance plans cover few preventive services, mostly limited to screening and immunizations, despite “findings…that the counseling and education services are among the most effective interventions available to clinicians to achieve the goals of health promotion and disease prevention.”¹⁰ Group Health Cooperative (GHC) of Puget Sound has demonstrated with a comprehensive systematic population-based health care approach that the prevalence of smoking can be reduced from 25% to 15.5% over 10 years among more than 550,000 adult enrollees.¹¹ This tremendous change within a population was achieved by multiple approaches, including identification, tracking, community outreach, comprehensive clinician and staff education, free coverage of services to participants, accessible telephone counseling, and self-help materials.¹¹ This is an outstanding example of an effective comprehensive program on smoking cessation within a private managed care system resulting in the overall reduction of smoking for a large population. It serves as a model for other preventive services concerning such common topics as alcohol consumption abuse, cancer screening, or coronary artery disease.

The prevalence of smoking among Medicaid health maintenance organizations (HMOs) versus commercial HMO participants is reported to be much higher. In recent Michigan surveys of health plans, 19.4% of the participants in commercial HMOs reported being current smokers compared with 44.1% of Medicaid HMO participants.¹² Medicaid participants are clearly a high-risk population for tobacco use and the medical consequences of smoking. Many states are moving from a fee-for-service approach to Medicaid coverage as a prospective capitated payment approach within managed care. Before the course of our study all Medicaid participants were moved into managed care plans by the State of Michigan. This context provided the ideal setting for examination of the impact of a systematic approach to smoking cessation by office-based and telephone counseling follow-up care for smokers covered by Medicaid managed care.

Brief advice on smoking cessation from a physician alone results in long-term quit rates of less than 10%.¹³ With the supplementation of brief physician advice with higher-dose nicotine gum or transdermal nicotine in randomized-controlled settings, long-term quit rates are increased to 15% to 25%.^14-16 In the context of community practice relying on general volunteers, long-term quit rates are lower than strictly controlled trials.¹⁷ It seems that pharmacotherapy clearly enhances brief advice by physicians for smoking cessation.¹⁶ In a primary care medical practice-based study, Daughton and colleagues¹⁸ state that “data clearly indicate that counseling seems to maximize smoking cessation rates with the nicotine patch.” Most studies (including that by Daughton and coworkers) have examined the relative effectiveness of pharmacotherapy against placebo. Our study proposes to answer the following questions: What is the comparative efficacy in quit rates by adding nurse and telephone counseling support for follow-up care to physician advice alone when all smokers receive the same pharmacotherapy? Does added behavioral support actually improve quit rates when all smokers use pharmacotherapy, or is there no difference? Can significant quit rates be achieved in low-income populations? Are there special measures required to maintain follow-up and protocol compliance in Medicaid smokers? What are the barriers to decreasing the high prevalence of smoking among participants in Medicaid managed care plans?

Methods

Recruitment of Participants

Participants were enrolled from January 19, 1998, to June 20, 1998, from 3 community heath center sites (Hackley, Baldwin, and Muskegon) in Michigan. Each practice site had the designation of a federally underserved site with the majority of care provided to very low-income patients. Each practice site had 5 to 7 providers with approximately 10,000 to 15,000 active patients on record. All participants were smokers older than 21 years with Medicaid managed care insurance. Participants were covered by 4 different managed care plans that agreed to allow the participation of their patients in the study. All participants had no medical contraindications to the use of transdermal nicotine, including pregnancy, and were willing to commit to quitting smoking within the next 30 days. Smokers were invited to participate in the study during their usual office visits and were offered 21-mg transdermal nicotine for 8 weeks as covered by Medicaid and determined appropriate by their providers.

Recruitment Rates

During the 6-month recruitment period, 501 smokers on Medicaid managed care were identified as eligible by office nurses and referred to participate in our study at the 3 practice sites. Of the referred group, 259 (52%) enrolled in our study and were randomized to either usual or relapse prevention care. A total of 233 (48% of the referred group) participated in our study. Participation was defined as receiving brief physician advice for the usual care group or brief physician advice and 1 telephonic-counseling session for the telephonic-counseling care group. These rates of enrollment were consistent across the 3 study sites. Participants were excluded from the final analysis if it was discovered after randomization that they did not have telephones. The informed consent process was approved by the institutional review board of Michigan State University. Figure 1 shows the recruitment and randomization flow.

Training of Providers and Staff

A total of 20 primary care physicians were trained to provide brief advice for smoking cessation consistent with the national guidelines.¹ Physician training consisted of a 2-hour update session on the guidelines, an overview of the study protocol, and role playing. Physicians received continuing medical education credit for participating. Ten nurses^3-4 per site and 10 telephone counselors were trained in computer-assisted relapse prevention. Nurse and telephone-counselor training consisted of 3 2-hour sessions on relapse prevention, computer skills, and individual case management. Nurses and counselors were encouraged to practice on case examples between training sessions. Their intervention skills were evaluated before they began counseling study participants. Quality assurance of counseling performance was performed through weekly audiotape review by research assistants. The computer program (“I’d Rather Cope than Smoke”⁹) provided a continuing record of counseling time and accuracy of data collection per nurse and counselor.

Study Design and Counseling Interventions

The participants who were assigned to the usual-care group participated in an intake session, received brief advice on smoking by their provider according to the guidelines,¹ were given a prescription for transdermal nicotine if medically appropriate, and had a follow-up scheduled for at least 1 visit (usually 7 to 30 days after the quit date) consistent with their medical condition. All participants also received “Clearing the Air” (National Cancer Institute publication no. 95-1647). The intake session lasted approximately 45 minutes. The intake was conducted by study staff (not the nurses providing telephone counseling) to prevent selection bias. Randomized assignment to either usual care or telephonic-counseling care groups occurred immediately after the intake session.

The participants assigned to the relapse prevention telephonic-counseling group received an intake session, usual care, a copy of “Clearing the Air,” a diary of coping responses the size of a cigarette pack, and 6 telephonic-counseling sessions. The sequence of follow-up sessions was determined according to the quit date: Session 1 was scheduled for 1 day after the quit date; session 2, 3 days; session 3, 7 days; session 4, 14 days; session 5, 30 days; and session 6, 60 days. Follow-up sessions lasted approximately 15 to 20 minutes. This sequence is consistent with previously reported studies by the authors⁹ and other investigators.^4,8 As previously reported, the computer software program, “I’d Rather Cope than Smoke,” was developed to assist in compliance with the relapse prevention protocol.⁹ All counseling sessions were done telephonically. Trained office nurses who used the software on laptop computers performed the first 3 treatment sessions. The intake and follow-up data of the first 3 sessions were electronically transferred to a computer network at Michigan State University where trained telephone counselors provided sessions 4 to 6.

Barriers to Maintaining Telephone Treatment

Before the study onset, focus group analysis of low-income smokers reported that the majority preferred counseling sessions on relapse prevention to be done by telephone rather than in person by the office nurse at the practice sites. We anticipated frequent disruptions in telephone service for the study population, so several innovative methods to maintain telephone treatment were developed, such as: (1) immediately contacting directory assistance for disruptions in service; (2) verifying site records for phone numbers changes; (3) contacting participants during subsequent clinic visits to update phone numbers; and (4) mailing a self-addressed stamped postcard requesting immediate feedback.

Independent Variables

Participants were evaluated for standard demographic characteristics of sex, age, socioeconomic status, education level, and working status. Baseline smoking activity was evaluated on the basis of the number of cigarettes smoked per day, the number of years of smoking, the mini-Fagerstrom Tolerance Questionnaire (FTQ),19 household activity, confidence in quitting, and personal reasons for quitting. Medicaid insurance status was verified. Personal patterns of relapse triggers and coping response were recorded.

Outcomes Measured

The main outcome measure was carbon monoxide verified smoke-free status at a telephone follow-up 90 days after the quit date in both usual and telephonic-counseling groups. Multiple attempts were made to contact participants, regardless of the level of participation at 3 months. Participants reporting 7-day smoke-free status at 3 months were invited to have carbon monoxide verification at the office and were paid $50 for their time.

The secondary outcome measures included physician, nurse, counselor, and participant compliance with protocols; provider and staff satisfaction with the program; and nicotine replacement use.

Statistics

Comparisons of study group characteristics were made using standard statistical measures. Categorical variables were tested using the chi-square test for contingency tables and the Student t test for continuous variables. Several continuous variables were categorized and analyzed by both methods.

The study denominator was based on intention-to-treat assignment as in randomized controlled trials^20,21 for evaluation of pharmacotherapy for nicotine addiction. Participants who refused follow-up, failed to call back, gave incorrect contact numbers, or dropped out were counted as smokers.

Smoking quit rates at 90-day follow-ups were compared using the z score for equality of proportions. Adjustments were made in self-reported outcomes based on carbon monoxide verification rates.

Results

Demographic Comparison of Study Groups

A total of 238 smokers participated in the study (N=123 usual care group, and N=110 in the telephonic-counseling group) and patient demographics are reported in Table 2. The smoking characteristics of the study groups are provided in Table 3. Adjustments for participants without telephones did not induce any significant differences.

As shown in Table 4, the most common reasons for quitting by far were personal health reasons and health problems related to smoking. Very few participants reported advice from their physician as the reason to quit smoking. The groups were comparable and did not differ significantly in their reasons for quitting.

Smoke-Free Status

Of the 233 patients with telephones enrolled in the study, 80 (65%) in the usual care group and 74 (67%) in the telephonic-counseling group were successfully contacted. Of those contacted, 19 in the usual care group and 24 in the telephonic-counseling group reported that they were smoke free. However, smoke-free status was successfully confirmed using carbon monoxide (CO) monitoring in only 56% of patients claiming to be smoke free in the usual care group, while 95% of patients in the telephonic-counseling group had their smoke-free status confirmed. Thus, in the per-protocol analysis, smoke-free status was confirmed in 10 of 80 (12.5%) in the usual care group and 23 of 74 (31%) in the telephonic-counseling group (P=.004).

In the intention-to-treat analysis we assumed that all missing cases were smoke free and used denominators of 123 and 110 for the usual care and telephonic-counseling groups, respectively. In this intention-to-treat analysis, the rates of self-reported smoke-free status were 15% and 19% (P=ns). In the intention-to-treat analysis of CO-verified smoke-free status, patients in the telephonic-counseling group were more likely to be smoke free (8.1% vs 21%, P <.01).

Nicotine Replacement Use

Prescriptions for nicotine replacement were received by 91% of the usual care and 99% of the telephonic-counseling care participants. At follow-up evaluation, 73% of the usual care and 67% of telephonic-counseling care participants reported using at least an initial course of nicotine replacement. These proportions of use did not differ significantly between the study groups

Discussion

Smoking has been shown to be one of the most modifiable health risks significantly related to higher health care charges, even after controlling for age, sex, race, diabetes, and heart disease.²² Although indemnity plans have been largely unsupportive of services for smoking cessation counseling, managed care plans have shown considerable success at decreasing the prevalence of smoking by offering comprehensive smoking cessation services.^23,24 In fact, offering full coverage of both behavioral and pharmacotherapy services results in a greater reduction in smoking prevalence than partial coverage.²⁴ The studies mentioned on smoking cessation were conducted with participants who were employed and had commercial insurance coverage.

Our study examined the effectiveness of a comprehensive program for smoking cessation provided by nurse and telephone counselors who were assisted by a computer-guided program focusing on relapse prevention in very low-income smokers covered by Medicaid managed care. The intention-to-treat results of a 21% quit rate at 3 months were consistent with our previously reported study,⁹ which included a sizable subpopulation of Medicaid patients. If adjustments are made in the denominator based on community trials¹⁷ as our previous study⁹ for reasonable loss-to-follow, then the CO-verified quit rates at 3 months would be 13% (usual care) and 31% (telephonic care) (P=.011). Our report is unique because we directly compared the effectiveness of telephone counseling support with usual care (brief physician advice and follow-up) in a true experimental trial in community practice. Though most participants received prescriptions for transdermal nicotine, the variation in usage was similar in both study groups because randomization allows a true comparison of the behavioral intervention effects. The recruitment data showing that approximately 50% of referred smokers in primary care are willing to enroll in a program is consistent with our previous study⁹ and other reports.²⁴ This demonstrates that Medicaid smokers are generally as willing to participate in smoking cessation services as other smokers.

Although all providers received formal training on the smoking cessation guidelines,¹ were aware of the study, and had “green card” reminders on study charts, they offered appropriate follow-up care only 26% of the time at return visits (based on post-study chart audit documentation). These findings are consistent with national surveys of physicians in primary care practices² that show follow-up care as the greatest shortcoming. It seems that physicians need to have comprehensive office systems in place to ensure even brief follow-up care²⁶ for smoking cessation. Telephone counseling support with a guided computer system definitely enhances follow-up care. By closely tracking participants for changes in addresses and telephone services, reasonable follow-up can be maintained even in low-income smokers. In our study, 60% of the participants in the telephonic-counseling group received at least 4 treatment sessions. Opinions of providers and staff during post-study focus groups were very positive. All 3 practices decided to continue a nurse-based approach for relapse prevention counseling after the study and expressed a need for the telephone support services to continue.

Limitations

One of the possible weaknesses of this study is the lack of long-term follow-up at 6 to 12 months for quit rates to ensure continued differences in effectiveness. Because of lack of funding, we were only able to obtain follow-up at 3 months. However, our findings are similar to the data in our previously reported community demonstration trial,⁹ which did not have a usual care comparison. Though the 2 reports refer to different populations, in our previous report⁹ using an intention-to-treat denominator the CO-verified quit rates were approximately 20% at 6 months in the Medicaid population. When using a community-based denominator that accounted for loss to follow-up, the 6-month quit rate was 33%. These results are consistent with strictly controlled trials where the majority of participants used nicotine replacement therapies.¹⁶

It is of interest to note that in this very low-income population, providing $50 to verify self-reported smoking cessation by CO monitor not only yielded considerable follow-up at 3 months but may have biased self-reporting in the usual care group where only 56% of the reports were verified. This finding shows the importance of using biochemical verification of smoking cessation even in community-based clinical trials.

Continued Research

Our study poses several questions for further research. Are the quit rates obtained by the described telephonic-counseling program sustainable over time at 1 to 2 years post-treatment in low-income populations? Can these approaches for relapse prevention be adapted to meet the needs of special groups, such as pregnant smokers, difficult to reach smokers at home, and high-risk smokers with diseases such as diabetes, heart disease, asthma, and severe disabilities when offered in conjunction with disease management services within managed care plans? This is of particular importance when the majority of low-income smokers report personal, smoking-related, and family health problems as reasons for quitting smoking. Though such behavioral support services are reported to be cost-effective in commercial managed care populations,²⁵ what is the cost-effectiveness of these services when adapted to meet the needs of special populations?

Conclusions

Telephonic-counseling for smoking cessation supported by a computer-guided program on relapse prevention is both practical and effective even for low-income smokers covered by Medicaid managed care. Special tracking approaches are required to maintain low-income smokers in treatment and to ensure provider follow-up. State Medicaid programs and insurance plans should consider investing in both office-based and centralized telephonic smoking cessation services to enhance smoking cessation for low-income smokers.

Acknowledgments

Our research was supported by a grant from the Michigan Department of Community Health to the Institute for Managed Care of Michigan State University (MSU) as a subproject on “Cancer Prevention, Outreach and Screening/Detection for Cancer Patients.”

Joseph Farrell, MA, director of the Institute for Managed Care, acted as the overall project director. Barbara Given, PhD, RN, and professor in the College of Nursing at MSU, was overall project manager. Wei Pan, MS, provided data management and statistical support. Kathy Ives, research assistant, contributed project coordination, data entry, and analysis. Dorothy Pathak, PhD, biostatistics consultant, verified the statistical analysis. We thank the Medicaid managed care plans of Wellness, Care Choices, Physicians Health Plan, and Community Choice for allowing patient participation and covering pharmacotherapy during the study. We thank the community health center physicians and staff in the Michigan communities of Hackley, Baldwin, and Muskegon for participating in the study.

Clinical practice guidelines on smoking cessation1 advocate that clinicians identify all smokers, advise them to quit, and arrange follow-up care. Arranging systematic follow-up care is often the most difficult of those steps in a primary medical practice because counseling for smoking cessation is often not reimbursed.² Telephone support counseling services offering proactive follow-up with scheduled sessions have achieved long-term success rates from 25% to 30%.^3-8 We were able to achieve a long-term quit rate of 36% in a community-based trial of computer-assisted telephone support counseling by nurses and telephone counselors trained in computer skills and relapse prevention.⁹ In this study more than 57% of the practice-based participants were covered by Medicaid insurance. There were no statistically significant differences in quit rates for Medicaid (33%) and non-Medicaid (36%) smokers at 6 months using a community denominator analysis approach.⁹

Managed care provides an advantageous system for the delivery of preventive services.¹⁰ Most indemnity insurance plans cover few preventive services, mostly limited to screening and immunizations, despite “findings…that the counseling and education services are among the most effective interventions available to clinicians to achieve the goals of health promotion and disease prevention.”¹⁰ Group Health Cooperative (GHC) of Puget Sound has demonstrated with a comprehensive systematic population-based health care approach that the prevalence of smoking can be reduced from 25% to 15.5% over 10 years among more than 550,000 adult enrollees.¹¹ This tremendous change within a population was achieved by multiple approaches, including identification, tracking, community outreach, comprehensive clinician and staff education, free coverage of services to participants, accessible telephone counseling, and self-help materials.¹¹ This is an outstanding example of an effective comprehensive program on smoking cessation within a private managed care system resulting in the overall reduction of smoking for a large population. It serves as a model for other preventive services concerning such common topics as alcohol consumption abuse, cancer screening, or coronary artery disease.

The prevalence of smoking among Medicaid health maintenance organizations (HMOs) versus commercial HMO participants is reported to be much higher. In recent Michigan surveys of health plans, 19.4% of the participants in commercial HMOs reported being current smokers compared with 44.1% of Medicaid HMO participants.¹² Medicaid participants are clearly a high-risk population for tobacco use and the medical consequences of smoking. Many states are moving from a fee-for-service approach to Medicaid coverage as a prospective capitated payment approach within managed care. Before the course of our study all Medicaid participants were moved into managed care plans by the State of Michigan. This context provided the ideal setting for examination of the impact of a systematic approach to smoking cessation by office-based and telephone counseling follow-up care for smokers covered by Medicaid managed care.

Brief advice on smoking cessation from a physician alone results in long-term quit rates of less than 10%.¹³ With the supplementation of brief physician advice with higher-dose nicotine gum or transdermal nicotine in randomized-controlled settings, long-term quit rates are increased to 15% to 25%.^14-16 In the context of community practice relying on general volunteers, long-term quit rates are lower than strictly controlled trials.¹⁷ It seems that pharmacotherapy clearly enhances brief advice by physicians for smoking cessation.¹⁶ In a primary care medical practice-based study, Daughton and colleagues¹⁸ state that “data clearly indicate that counseling seems to maximize smoking cessation rates with the nicotine patch.” Most studies (including that by Daughton and coworkers) have examined the relative effectiveness of pharmacotherapy against placebo. Our study proposes to answer the following questions: What is the comparative efficacy in quit rates by adding nurse and telephone counseling support for follow-up care to physician advice alone when all smokers receive the same pharmacotherapy? Does added behavioral support actually improve quit rates when all smokers use pharmacotherapy, or is there no difference? Can significant quit rates be achieved in low-income populations? Are there special measures required to maintain follow-up and protocol compliance in Medicaid smokers? What are the barriers to decreasing the high prevalence of smoking among participants in Medicaid managed care plans?

Methods

Recruitment of Participants

Participants were enrolled from January 19, 1998, to June 20, 1998, from 3 community heath center sites (Hackley, Baldwin, and Muskegon) in Michigan. Each practice site had the designation of a federally underserved site with the majority of care provided to very low-income patients. Each practice site had 5 to 7 providers with approximately 10,000 to 15,000 active patients on record. All participants were smokers older than 21 years with Medicaid managed care insurance. Participants were covered by 4 different managed care plans that agreed to allow the participation of their patients in the study. All participants had no medical contraindications to the use of transdermal nicotine, including pregnancy, and were willing to commit to quitting smoking within the next 30 days. Smokers were invited to participate in the study during their usual office visits and were offered 21-mg transdermal nicotine for 8 weeks as covered by Medicaid and determined appropriate by their providers.

Recruitment Rates

During the 6-month recruitment period, 501 smokers on Medicaid managed care were identified as eligible by office nurses and referred to participate in our study at the 3 practice sites. Of the referred group, 259 (52%) enrolled in our study and were randomized to either usual or relapse prevention care. A total of 233 (48% of the referred group) participated in our study. Participation was defined as receiving brief physician advice for the usual care group or brief physician advice and 1 telephonic-counseling session for the telephonic-counseling care group. These rates of enrollment were consistent across the 3 study sites. Participants were excluded from the final analysis if it was discovered after randomization that they did not have telephones. The informed consent process was approved by the institutional review board of Michigan State University. Figure 1 shows the recruitment and randomization flow.

Training of Providers and Staff

A total of 20 primary care physicians were trained to provide brief advice for smoking cessation consistent with the national guidelines.¹ Physician training consisted of a 2-hour update session on the guidelines, an overview of the study protocol, and role playing. Physicians received continuing medical education credit for participating. Ten nurses^3-4 per site and 10 telephone counselors were trained in computer-assisted relapse prevention. Nurse and telephone-counselor training consisted of 3 2-hour sessions on relapse prevention, computer skills, and individual case management. Nurses and counselors were encouraged to practice on case examples between training sessions. Their intervention skills were evaluated before they began counseling study participants. Quality assurance of counseling performance was performed through weekly audiotape review by research assistants. The computer program (“I’d Rather Cope than Smoke”⁹) provided a continuing record of counseling time and accuracy of data collection per nurse and counselor.

Study Design and Counseling Interventions

The participants who were assigned to the usual-care group participated in an intake session, received brief advice on smoking by their provider according to the guidelines,¹ were given a prescription for transdermal nicotine if medically appropriate, and had a follow-up scheduled for at least 1 visit (usually 7 to 30 days after the quit date) consistent with their medical condition. All participants also received “Clearing the Air” (National Cancer Institute publication no. 95-1647). The intake session lasted approximately 45 minutes. The intake was conducted by study staff (not the nurses providing telephone counseling) to prevent selection bias. Randomized assignment to either usual care or telephonic-counseling care groups occurred immediately after the intake session.

The participants assigned to the relapse prevention telephonic-counseling group received an intake session, usual care, a copy of “Clearing the Air,” a diary of coping responses the size of a cigarette pack, and 6 telephonic-counseling sessions. The sequence of follow-up sessions was determined according to the quit date: Session 1 was scheduled for 1 day after the quit date; session 2, 3 days; session 3, 7 days; session 4, 14 days; session 5, 30 days; and session 6, 60 days. Follow-up sessions lasted approximately 15 to 20 minutes. This sequence is consistent with previously reported studies by the authors⁹ and other investigators.^4,8 As previously reported, the computer software program, “I’d Rather Cope than Smoke,” was developed to assist in compliance with the relapse prevention protocol.⁹ All counseling sessions were done telephonically. Trained office nurses who used the software on laptop computers performed the first 3 treatment sessions. The intake and follow-up data of the first 3 sessions were electronically transferred to a computer network at Michigan State University where trained telephone counselors provided sessions 4 to 6.

Barriers to Maintaining Telephone Treatment

Before the study onset, focus group analysis of low-income smokers reported that the majority preferred counseling sessions on relapse prevention to be done by telephone rather than in person by the office nurse at the practice sites. We anticipated frequent disruptions in telephone service for the study population, so several innovative methods to maintain telephone treatment were developed, such as: (1) immediately contacting directory assistance for disruptions in service; (2) verifying site records for phone numbers changes; (3) contacting participants during subsequent clinic visits to update phone numbers; and (4) mailing a self-addressed stamped postcard requesting immediate feedback.

Independent Variables

Participants were evaluated for standard demographic characteristics of sex, age, socioeconomic status, education level, and working status. Baseline smoking activity was evaluated on the basis of the number of cigarettes smoked per day, the number of years of smoking, the mini-Fagerstrom Tolerance Questionnaire (FTQ),19 household activity, confidence in quitting, and personal reasons for quitting. Medicaid insurance status was verified. Personal patterns of relapse triggers and coping response were recorded.

Outcomes Measured

The main outcome measure was carbon monoxide verified smoke-free status at a telephone follow-up 90 days after the quit date in both usual and telephonic-counseling groups. Multiple attempts were made to contact participants, regardless of the level of participation at 3 months. Participants reporting 7-day smoke-free status at 3 months were invited to have carbon monoxide verification at the office and were paid $50 for their time.

The secondary outcome measures included physician, nurse, counselor, and participant compliance with protocols; provider and staff satisfaction with the program; and nicotine replacement use.

Statistics

Comparisons of study group characteristics were made using standard statistical measures. Categorical variables were tested using the chi-square test for contingency tables and the Student t test for continuous variables. Several continuous variables were categorized and analyzed by both methods.

The study denominator was based on intention-to-treat assignment as in randomized controlled trials^20,21 for evaluation of pharmacotherapy for nicotine addiction. Participants who refused follow-up, failed to call back, gave incorrect contact numbers, or dropped out were counted as smokers.

Smoking quit rates at 90-day follow-ups were compared using the z score for equality of proportions. Adjustments were made in self-reported outcomes based on carbon monoxide verification rates.

Results

Demographic Comparison of Study Groups

A total of 238 smokers participated in the study (N=123 usual care group, and N=110 in the telephonic-counseling group) and patient demographics are reported in Table 2. The smoking characteristics of the study groups are provided in Table 3. Adjustments for participants without telephones did not induce any significant differences.

As shown in Table 4, the most common reasons for quitting by far were personal health reasons and health problems related to smoking. Very few participants reported advice from their physician as the reason to quit smoking. The groups were comparable and did not differ significantly in their reasons for quitting.

Smoke-Free Status

Of the 233 patients with telephones enrolled in the study, 80 (65%) in the usual care group and 74 (67%) in the telephonic-counseling group were successfully contacted. Of those contacted, 19 in the usual care group and 24 in the telephonic-counseling group reported that they were smoke free. However, smoke-free status was successfully confirmed using carbon monoxide (CO) monitoring in only 56% of patients claiming to be smoke free in the usual care group, while 95% of patients in the telephonic-counseling group had their smoke-free status confirmed. Thus, in the per-protocol analysis, smoke-free status was confirmed in 10 of 80 (12.5%) in the usual care group and 23 of 74 (31%) in the telephonic-counseling group (P=.004).

In the intention-to-treat analysis we assumed that all missing cases were smoke free and used denominators of 123 and 110 for the usual care and telephonic-counseling groups, respectively. In this intention-to-treat analysis, the rates of self-reported smoke-free status were 15% and 19% (P=ns). In the intention-to-treat analysis of CO-verified smoke-free status, patients in the telephonic-counseling group were more likely to be smoke free (8.1% vs 21%, P <.01).

Nicotine Replacement Use

Prescriptions for nicotine replacement were received by 91% of the usual care and 99% of the telephonic-counseling care participants. At follow-up evaluation, 73% of the usual care and 67% of telephonic-counseling care participants reported using at least an initial course of nicotine replacement. These proportions of use did not differ significantly between the study groups

Discussion

Smoking has been shown to be one of the most modifiable health risks significantly related to higher health care charges, even after controlling for age, sex, race, diabetes, and heart disease.²² Although indemnity plans have been largely unsupportive of services for smoking cessation counseling, managed care plans have shown considerable success at decreasing the prevalence of smoking by offering comprehensive smoking cessation services.^23,24 In fact, offering full coverage of both behavioral and pharmacotherapy services results in a greater reduction in smoking prevalence than partial coverage.²⁴ The studies mentioned on smoking cessation were conducted with participants who were employed and had commercial insurance coverage.

Our study examined the effectiveness of a comprehensive program for smoking cessation provided by nurse and telephone counselors who were assisted by a computer-guided program focusing on relapse prevention in very low-income smokers covered by Medicaid managed care. The intention-to-treat results of a 21% quit rate at 3 months were consistent with our previously reported study,⁹ which included a sizable subpopulation of Medicaid patients. If adjustments are made in the denominator based on community trials¹⁷ as our previous study⁹ for reasonable loss-to-follow, then the CO-verified quit rates at 3 months would be 13% (usual care) and 31% (telephonic care) (P=.011). Our report is unique because we directly compared the effectiveness of telephone counseling support with usual care (brief physician advice and follow-up) in a true experimental trial in community practice. Though most participants received prescriptions for transdermal nicotine, the variation in usage was similar in both study groups because randomization allows a true comparison of the behavioral intervention effects. The recruitment data showing that approximately 50% of referred smokers in primary care are willing to enroll in a program is consistent with our previous study⁹ and other reports.²⁴ This demonstrates that Medicaid smokers are generally as willing to participate in smoking cessation services as other smokers.

Although all providers received formal training on the smoking cessation guidelines,¹ were aware of the study, and had “green card” reminders on study charts, they offered appropriate follow-up care only 26% of the time at return visits (based on post-study chart audit documentation). These findings are consistent with national surveys of physicians in primary care practices² that show follow-up care as the greatest shortcoming. It seems that physicians need to have comprehensive office systems in place to ensure even brief follow-up care²⁶ for smoking cessation. Telephone counseling support with a guided computer system definitely enhances follow-up care. By closely tracking participants for changes in addresses and telephone services, reasonable follow-up can be maintained even in low-income smokers. In our study, 60% of the participants in the telephonic-counseling group received at least 4 treatment sessions. Opinions of providers and staff during post-study focus groups were very positive. All 3 practices decided to continue a nurse-based approach for relapse prevention counseling after the study and expressed a need for the telephone support services to continue.

Limitations

One of the possible weaknesses of this study is the lack of long-term follow-up at 6 to 12 months for quit rates to ensure continued differences in effectiveness. Because of lack of funding, we were only able to obtain follow-up at 3 months. However, our findings are similar to the data in our previously reported community demonstration trial,⁹ which did not have a usual care comparison. Though the 2 reports refer to different populations, in our previous report⁹ using an intention-to-treat denominator the CO-verified quit rates were approximately 20% at 6 months in the Medicaid population. When using a community-based denominator that accounted for loss to follow-up, the 6-month quit rate was 33%. These results are consistent with strictly controlled trials where the majority of participants used nicotine replacement therapies.¹⁶

It is of interest to note that in this very low-income population, providing $50 to verify self-reported smoking cessation by CO monitor not only yielded considerable follow-up at 3 months but may have biased self-reporting in the usual care group where only 56% of the reports were verified. This finding shows the importance of using biochemical verification of smoking cessation even in community-based clinical trials.

Continued Research

Our study poses several questions for further research. Are the quit rates obtained by the described telephonic-counseling program sustainable over time at 1 to 2 years post-treatment in low-income populations? Can these approaches for relapse prevention be adapted to meet the needs of special groups, such as pregnant smokers, difficult to reach smokers at home, and high-risk smokers with diseases such as diabetes, heart disease, asthma, and severe disabilities when offered in conjunction with disease management services within managed care plans? This is of particular importance when the majority of low-income smokers report personal, smoking-related, and family health problems as reasons for quitting smoking. Though such behavioral support services are reported to be cost-effective in commercial managed care populations,²⁵ what is the cost-effectiveness of these services when adapted to meet the needs of special populations?

Conclusions

Telephonic-counseling for smoking cessation supported by a computer-guided program on relapse prevention is both practical and effective even for low-income smokers covered by Medicaid managed care. Special tracking approaches are required to maintain low-income smokers in treatment and to ensure provider follow-up. State Medicaid programs and insurance plans should consider investing in both office-based and centralized telephonic smoking cessation services to enhance smoking cessation for low-income smokers.

Acknowledgments

Our research was supported by a grant from the Michigan Department of Community Health to the Institute for Managed Care of Michigan State University (MSU) as a subproject on “Cancer Prevention, Outreach and Screening/Detection for Cancer Patients.”

Joseph Farrell, MA, director of the Institute for Managed Care, acted as the overall project director. Barbara Given, PhD, RN, and professor in the College of Nursing at MSU, was overall project manager. Wei Pan, MS, provided data management and statistical support. Kathy Ives, research assistant, contributed project coordination, data entry, and analysis. Dorothy Pathak, PhD, biostatistics consultant, verified the statistical analysis. We thank the Medicaid managed care plans of Wellness, Care Choices, Physicians Health Plan, and Community Choice for allowing patient participation and covering pharmacotherapy during the study. We thank the community health center physicians and staff in the Michigan communities of Hackley, Baldwin, and Muskegon for participating in the study.

The presence of sustained relationships between physicians and patients is a defining characteristic of primary care.¹ Family physicians use these relationships to acquire the depth of medical and personal knowledge about a patient that is essential to primary care practice.² It is also the reason some physicians choose this area of medicine.

A substantial body of empirical research points to the value of continuity in the physician-patient relationship, particularly in primary care. The benefits of continuity have been shown to accrue in the form of cost savings, improved health outcomes, and greater satisfaction for patients and physicians.^3-15 Yet little empirical research exists to indicate the amount of physician switching that occurs in primary care or the reasons for it.

In 1976 Kastler and colleagues¹⁶ examined the association between patients’ assessments of their care and their “doctor shopping” behavior. They found that patients’ evaluations of both interpersonal and structural features of care were significantly associated with the likelihood of voluntarily changing physicians. Those authors did not attempt to determine the relative importance of the 2 domains with respect to physician switching. The cross-sectional design precluded the study from determining which factor (if either) prospectively predicted switching.

Marquis and coworkers¹⁷ studied the sequencing of the satisfaction-disenrollment relationship using longitudinal data from the RAND Health Insurance Experiment (HIE). The HIE data showed that patients’ general satisfaction with their medical care significantly predicted physician switching over the following year. However, the HIE data did not afford the ability to differentiate among the many components of patient satisfaction and to discern which aspects specifically drive disenrollment.

Thus, little is known about the relative importance of the many factors that shape patients’ overall satisfaction with their physician and the extent to which performance on any of these ultimately drives a patient’s decision to leave a physician’s practice. Moreover, these earlier studies pre-date the recent surge in managed care enrollment and in consumerism among patients, both of which are presumed to be having a substantial impact on the rates of physician switching and the reasons for it. The generalizability of earlier findings to the present circumstances is unclear.

Methods

Our longitudinal observational study includes a population of insured adults who were employed by the Commonwealth of Massachusetts at baseline (1996), completed a self-administered questionnaire at baseline and follow-up (1999), and reported having a regular personal physician at baseline. Between January 1996 and April 1996 the baseline questionnaire was administered to a random sample of commonwealth employees who subscribed to any of 12 health plans available to employees, their dependents, and retirees. A 68.5% response rate was achieved (n=7204) using a standard 3-stage mail survey protocol with limited telephone follow-up of nonrespondents (mail responses=6810; telephone responses=394). Further details of the baseline sampling and data collection methods are documented elsewhere.

Follow-up data collection occurred precisely 3 years after baseline (January 1999-April 1999). Respondents who identified a primary care physician at baseline and participated by mail were eligible for follow-up (n=6075). Data were obtained using a standard 3-stage mail survey protocol with a final targeted mailing to racial and ethnic minorities (n=311) and to those without a college diploma (n=521). The targeted mailings were done because these subgroups were found to be underrepresented among follow-up respondents near the conclusion of data collection, and their representation in the longitudinal sample was important to our objectives. A 69.4% response rate was achieved in follow-up (n=4108) after accounting for respondents who died (n=21), were too ill to participate (n=2), or could not be located by mail in 1999 (n=136). At baseline and follow-up, respondents were somewhat older than nonrespondents, more likely to be women and white, and less likely to be poor (Appendix, Table 1A.

The questionnaire administered to patients at both baseline and follow-up included 4 scales measuring features of the physician-patient relationship (quality of communication, interpersonal treatment, physician’s knowledge of the patient, and patient trust) and 4 scales measuring structural aspects of care (access to care, visit-based continuity, duration of primary care relationship, and integration of care). The 8 scales are part of the Primary Care Assessment Survey (PCAS), a validated questionnaire with measures corresponding to the defining features of primary care posited by the Institute of Medicine (IOM). All concepts are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit-specific). All PCAS scales are scored on a 0 to 100 scale, with higher scores indicating more of the referent attribute. Details of the development and psychometric performance of the PCAS scales are available elsewhere. The item content and reliability coefficient (Cronbach a)for each scale are summarized in the Appendix Table 2A. In addition to completing the PCAS items referencing their experiences with and assessments of their primary physicians, the respondents also provided their physicians’ names.

Using the physician-identifying information provided by the patients, we linked data from the Massachusetts Board of Registration in Medicine (BRM) to the study database. The BRM data provided the physician’s practice address and several characteristics of the physician’s training and practice. We linked with the BRM data by using a matching algorithm based on the spelling of the physician’s name as provided by the patient, the distance between the patient’s home ZIP code and the physician’s practice site (BRM database), and the physician’s medical specialty. At both baseline and follow-up, matches from the BRM data were identified for 94.0% of the patients who named a physician.

Identifying Voluntary Versus Involuntary Disenrollment

Patients were classified as having changed physicians during the study period if their follow-up questionnaire reported having been in their primary physician’s practice for less than 3 years and if the physician named at follow-up was different from the one named in 1996. Those who changed physicians were then classified as having switched either voluntarily or involuntarily. A switch was considered involuntary if: (1) the patient’s baseline physician was no longer listed as active in the Massachusetts BRM database (n=77), (2) the baseline physician had moved more than 10 miles (n=91), or (3) the patient had moved more than 15 miles from the baseline residence (n=62).

In addition we considered the possible involuntary nature of physician switches that occurred along with a change in health plan enrollment. Because the employer in our study did not force or even incentivize health plan changes during the study period (ie, there was a consistent offering of health plans and no notable changes in the employee contributions for coverage), respondents who were insured by the commonwealth throughout the study period did not incur any involuntary physician switching owing to employer-imposed health plan changes. Among respondents not insured by the commonwealth throughout the study (ie, respondents who left state employment [n=40] or deferred coverage [n=7]), there were 6 who changed physicians. Five of these did so while remaining in the same health plan and were thus coded as having voluntarily changed physicians. The remaining individual who both changed health plans and physicians was dropped from our analysis of voluntary disenrollment, since we were unable to ascertain whether the plan change forced a change in physician.

Statistical Analyses

We limited the analytic sample to patients who completed both the baseline and follow-up questionnaire, who identified a primary physician at baseline for whom a BRM database match was found, and who had either remained with their baseline physician throughout the study period or had voluntarily left the physician’s practice (n=3052). Patients who had involuntarily disenrolled from their baseline physician’s practice (n=230) were excluded. Their exclusion was necessary, since there was no way to determine whether those who involuntarily switched physicians would have otherwise voluntarily left their physician. The sociodemographic and health profile of the analytic sample (n=3052) did not differ from that of the complete 1999 sample (n=4108).

Multiple logistic regression methods were used to evaluate interpersonal and structural features of care, as measured by the baseline PCAS scales, as predictors of voluntary disenrollment from a physician’s practice. All scales were standardized ([X1-mean]/standard deviation) to permit direct comparison of results across scales. First, the 8 PCAS scales were tested individually as predictors of voluntary disenrollment. Testing scales independently in this way is useful in cases such as this where moderate to high correlations exist among some scales. Although the majority of PCAS scale correlations are small, higher correlations exist among some scales (r=0.40-0.86). We applied the Bonferroni correction for multiple comparisons to this set of analyses.

Next, we modeled voluntary disenrollment as a function of the 4 relationship-quality scales together and tested for the equality of their effects (odd ratio [OR]) using a chi-square test. We repeated this using the 4 measures of structural features of care. Finally, using factor analysis methods (principal factor), we explored the potential for defining a single factor denoting relationship quality and a single factor denoting structural features of care. The 4 scales denoting structural features of care failed to generate an acceptable factor (range of factor loadings=0.20 [relationship duration] to 0.67 [access to care]), so this factor was dropped. The relationship-quality factor was retained (range of factor loadings=0.84 [knowledge of patient] to 0.92 [communication]) and tested in multiple logistic regression along with each of the 4 structure-of-care measures. A chi-square test was used to test the equivalence of the effects (OR) associated with the relationship quality factor and each of the 4 structure-of-care scales.

All regression models controlled for patients’ baseline sociodemographic profile (age, sex, race, years of education, household income), baseline health status (physical functioning, mental functioning, number of primary care sensitive conditions [PCSC], and number of primary care insensitive conditions [PCIC]), and baseline utilization (number of ambulatory visits in the previous 6 months). Physical and mental functioning were measured with data from the Medical Outcomes Study Short Form-12 (SF-12) Health Survey, which was included in the patient questionnaire. The numbers of primary care sensitive and insensitive conditions were classified using patients’ baseline reports about 20 chronic medical conditions with high prevalence among adults in the United States. The classification of PCSC and PCIC was defined by 9 generalist physicians, blind to the study objectives, who were asked to identify those conditions for which good primary care management could substantially affect outcomes (PCSC) and those for which it could not (PCIC). PCSC included hypertension, recent myocardial infarction, congestive heart failure, diabetes, angina, migraines, seasonal allergies, asthma, ulcers, arthritis, cancer, back pain, weight problem, and depression. PCIC included blindness, deafness, liver disease, insomnia, nonseasonal allergies (eg, dust, food, pets), and limb paralysis or amputation.

We assessed the goodness-of-fit of the final models using the Hosmer and Lemeshow method. For each scale, the P on the chi-square test statistic was greater than .05, indicating that the model fit the data well.

Results

Slightly more than one fourth of the patients in the longitudinal study panel changed physicians during the 3-year follow-up period (n=899), while approximately three fourths remained with their baseline physician throughout the study (n=2383). Of those who changed physicians, most changed voluntarily (n=669), but some changed involuntarily (n=230) because the physician had moved, retired, died, or the patient had moved a substantial distance. Table 1 shows the unadjusted sociodemographic, health, and utilization characteristics of the analytic sample, comparing those who voluntarily changed physicians with those who remained with their baseline physician throughout our study. Voluntary disenrollees were younger and more likely to be women and nonwhite than those who stayed with their baseline physician (P <.01). There were no differences in the baseline health status or outpatient utilization of the 2 groups.

Table 2 presents the results of the regression analyses examining the 8 PCAS scales as individual predictors of voluntary disenrollment (column 1) and the results of a multivariable model, including the composite relationship-quality factor (RQ) and the 4 structure-of-care scales as predictors of voluntary disenrollment (column 2, columns 3-7). When all scales were modeled independently (column 1), each was a significant predictor of voluntary disenrollment (P <.001), with somewhat larger effects associated with the relationship quality scales (OR=1.49-1.56) than the structure-of-care scales (OR=1.29-1.44). Pairwise tests of the ORs associated with each of the 4 relationship quality scales indicated that they were statistically equivalent in their ability to predict voluntary disenrollment. When the 4 indicators of relationship quality were included together in a multiple regression model, a chi-square test of their effects (OR) revealed the 4 to be statistically equivalent predictors of voluntary disenrollment. Similarly, in a model including the 4 structure-of-care scales, chi-square testing showed these 4 variables to have statistically equivalent effects. With the exception of sex, patient characteristics (sociodemographics, health, utilization) did not significantly predict voluntary disenrollment in any of these models. The gender effect had marginal significance in most cases (.05

Table 2 (column 2) shows the results of modeling voluntary disenrollment as a function of both relationship quality and structure-of-care together. In that multivariable model, the composite relationship quality factor (RQ) emerged as the leading predictor of voluntary disenrollment (OR=1.59; P <.001). This OR signifies that a standard deviation (SD) decline in relationship quality was associated with a 59% increase in the odds of voluntary disenrollment. The results indicate that after accounting for patients’ baseline characteristics (sociodemographic, health, and utilization) and the 4 structural features of care, patients with relationship quality scores in the 5th percentile in 1996 were 3 times more likely to voluntarily disenroll from their physician’s practice than those with 95th percentile relationship quality scores (37.8% vs 12.2%). The 2 measures of continuity also significantly predicted disenrollment in the multivariable model (visit-based continuity: OR=1.14, P=.03; relationship duration: OR=1.16, P=.01). Access to care predicted disenrollment with marginal significance (OR=1.14; P=.08), and integration did not significantly predict disenrollment (P=.59) in this model. None of the patient characteristics (sociodemograhics, health, utilization) significantly predicted disenrollment in the presence of these 5 quality-of-care measures.

Discussion

In our observational study of insured employed adults, 20% of the patients voluntarily left their primary care physician’s practice over a 3-year period. Another 5% left involuntarily, owing to factors that forced a change (eg, the physician moved, retired, or died). For the average full-time primary physician, this translates into approximately 400 patients voluntarily leaving the practice over a given 3-year period and another 100 leaving involuntarily. Rates of involuntary switching are almost certainly higher among physicians whose patients face more employer-imposed disruptions than occurred in our study population.

Our data indicate that the quality of the physician-patient relationship significantly predicts patients’ loyalty. With patient characteristics and structural features of care taken into account, those with the poorest-quality physician-patient relationships in 1996 were 3 times more likely to leave the physician’s practice over the ensuing 3 years than those with the highest-quality relationships.

Structural features of care also emerged as important determinants of patients’ disenrollment decisions. When considered independently of relationship quality, each of the 4 structural elements of care significantly predicted voluntary disenrollment. With relationship quality taken into account, continuity of care (both relationship duration and visit-based continuity) remained significant predictors of disenrollment, while access to care and integration of care did not. The results suggest that although these patients put a high priority on being given timely and convenient access to their physician’s office, the issue of who they are given access to and the quality of their connection with that clinician mattered more.

Our findings are consistent with those reported more than 2 decades ago by Kasteler and colleagues,¹⁶ who found both interpersonal quality of care and structural features of care to be significantly associated with voluntary physician switching in a cross-sectional study. Our study has the advantage of longitudinal data through which the sequencing of effects is clear. In addition, our study advances beyond earlier studies that evaluated a single patient-based measure of care in predicting disenrollment.^17,25 Marquis and coworkers¹⁷ showed that patients’ general satisfaction with their physician predicted disenrollment from the physician’s practice over the following year. Thom and colleagues²⁵ found that patients’ trust in their physicians significantly predicted disenrollment over the next 6 months. Our study includes measures of 8 characteristics that encompass the defining features of primary care as posited by the IOM1 and others,^26-29 with several features for which the relationship to disenrollment have not been previously studied. Our study contributes evidence concerning both the absolute and relative importance of interpersonal and structural features of care as predictors of patients’ loyalty to their primary care physician’s practice.

Limitations

Our study is limited to a population of adults in Massachusetts who were employed and insured at baseline. Rates of involuntary physician switching in this population were likely lower than would be observed in other employed populations (particularly in competitive health care markets) for the reasons mentioned (ie, benefit policies that minimized employer-imposed disruption of employees’ health care arrangements). However, the observed rates of voluntary physician switching and the predictors of voluntary disenrollment should not be affected. Those findings may be presumed to generalize more broadly.

A second limitation is the absence of information about salient health events that occurred between the baseline and follow-up phases of our study. For patients who incurred a serious episode of illness, information about the intervening health events and their experiences with their physician during that time might have enhanced our understanding of the factors that influenced their decisions about whether to remain in that physician’s practice.

Similarly, the study lacked detailed indicators of the technical quality of care provided and therefore could not assess the role that technical quality—and patients’ perceptions of it—play in shaping patients’ loyalty to their physician.

Finally, our study could not fully account for one potential source of involuntary disenrollment: patients leaving practices because the physicians no longer accepted their health plan. However, rates of physician turnover during the study period were no more than 5% in any of the health plans studied and were substantially lower in most.²¹ Thus, our findings are unlikely to have been substantively altered by a detailed accounting of this form of involuntary physician switching.

Conclusions

Previous empirical research has underscored the importance of physician-patient relationship quality by demonstrating its association with important outcomes, including adherence to medical advice,^19,30-32 satisfaction with care,^19,33-35 and litigation against physicians.^36-38 However, few studies have had the benefit of longitudinal data with which to verify the sequencing of effects between relationship quality and outcomes.

In our study the strength of physician-patient relationships in primary care—as indicated by patients’ trust in their physician, their assessments of how well the physician knows them, and the quality of communication and interpersonal treatment—was the leading predictor of patients’ loyalty to their primary physician’s practice. Continuity of care also significantly predicted voluntary disenrollment. The findings are noteworthy against a backdrop of health care delivery changes nationwide that many describe as threatening the therapeutic alliance between the physician and the patient.^29,39-42

The recent IOM report on the future of primary care called attention to the importance of the physician-patient relationship in primary care, asserting that primary care is predicated on sustained clinician-patient partnerships and on a whole-person orientation.1 In our study, longitudinal data demonstrate that the strength of connection between a patient and his or her primary care physician significantly predicts the likelihood of that patient remaining in that physician’s practice (vs voluntarily leaving) over the next several years. In an era marked by increasing pressure on clinicians and health care organizations to attend to such factors as market share, productivity, and efficiency, these findings point to a set of attributes that might otherwise be overlooked. They suggest that medical practices and health plans cannot afford to ignore that the essence of medical care delivery involves the interaction of one human being with another.

Acknowledgments

This research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, executive director of the Massachusetts Group Insurance Commission, whose commitment and participation have made this study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study. Finally, we acknowledge each of the health plans involved in the study, and particularly the health plan executives who served on our Advisory Committee and those who provided interviews, for generously giving their time, insights, and information that was critical to our research.

The presence of sustained relationships between physicians and patients is a defining characteristic of primary care.¹ Family physicians use these relationships to acquire the depth of medical and personal knowledge about a patient that is essential to primary care practice.² It is also the reason some physicians choose this area of medicine.

A substantial body of empirical research points to the value of continuity in the physician-patient relationship, particularly in primary care. The benefits of continuity have been shown to accrue in the form of cost savings, improved health outcomes, and greater satisfaction for patients and physicians.^3-15 Yet little empirical research exists to indicate the amount of physician switching that occurs in primary care or the reasons for it.

In 1976 Kastler and colleagues¹⁶ examined the association between patients’ assessments of their care and their “doctor shopping” behavior. They found that patients’ evaluations of both interpersonal and structural features of care were significantly associated with the likelihood of voluntarily changing physicians. Those authors did not attempt to determine the relative importance of the 2 domains with respect to physician switching. The cross-sectional design precluded the study from determining which factor (if either) prospectively predicted switching.

Marquis and coworkers¹⁷ studied the sequencing of the satisfaction-disenrollment relationship using longitudinal data from the RAND Health Insurance Experiment (HIE). The HIE data showed that patients’ general satisfaction with their medical care significantly predicted physician switching over the following year. However, the HIE data did not afford the ability to differentiate among the many components of patient satisfaction and to discern which aspects specifically drive disenrollment.

Thus, little is known about the relative importance of the many factors that shape patients’ overall satisfaction with their physician and the extent to which performance on any of these ultimately drives a patient’s decision to leave a physician’s practice. Moreover, these earlier studies pre-date the recent surge in managed care enrollment and in consumerism among patients, both of which are presumed to be having a substantial impact on the rates of physician switching and the reasons for it. The generalizability of earlier findings to the present circumstances is unclear.

Methods

Our longitudinal observational study includes a population of insured adults who were employed by the Commonwealth of Massachusetts at baseline (1996), completed a self-administered questionnaire at baseline and follow-up (1999), and reported having a regular personal physician at baseline. Between January 1996 and April 1996 the baseline questionnaire was administered to a random sample of commonwealth employees who subscribed to any of 12 health plans available to employees, their dependents, and retirees. A 68.5% response rate was achieved (n=7204) using a standard 3-stage mail survey protocol with limited telephone follow-up of nonrespondents (mail responses=6810; telephone responses=394). Further details of the baseline sampling and data collection methods are documented elsewhere.

Follow-up data collection occurred precisely 3 years after baseline (January 1999-April 1999). Respondents who identified a primary care physician at baseline and participated by mail were eligible for follow-up (n=6075). Data were obtained using a standard 3-stage mail survey protocol with a final targeted mailing to racial and ethnic minorities (n=311) and to those without a college diploma (n=521). The targeted mailings were done because these subgroups were found to be underrepresented among follow-up respondents near the conclusion of data collection, and their representation in the longitudinal sample was important to our objectives. A 69.4% response rate was achieved in follow-up (n=4108) after accounting for respondents who died (n=21), were too ill to participate (n=2), or could not be located by mail in 1999 (n=136). At baseline and follow-up, respondents were somewhat older than nonrespondents, more likely to be women and white, and less likely to be poor (Appendix, Table 1A.

The questionnaire administered to patients at both baseline and follow-up included 4 scales measuring features of the physician-patient relationship (quality of communication, interpersonal treatment, physician’s knowledge of the patient, and patient trust) and 4 scales measuring structural aspects of care (access to care, visit-based continuity, duration of primary care relationship, and integration of care). The 8 scales are part of the Primary Care Assessment Survey (PCAS), a validated questionnaire with measures corresponding to the defining features of primary care posited by the Institute of Medicine (IOM). All concepts are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit-specific). All PCAS scales are scored on a 0 to 100 scale, with higher scores indicating more of the referent attribute. Details of the development and psychometric performance of the PCAS scales are available elsewhere. The item content and reliability coefficient (Cronbach a)for each scale are summarized in the Appendix Table 2A. In addition to completing the PCAS items referencing their experiences with and assessments of their primary physicians, the respondents also provided their physicians’ names.

Using the physician-identifying information provided by the patients, we linked data from the Massachusetts Board of Registration in Medicine (BRM) to the study database. The BRM data provided the physician’s practice address and several characteristics of the physician’s training and practice. We linked with the BRM data by using a matching algorithm based on the spelling of the physician’s name as provided by the patient, the distance between the patient’s home ZIP code and the physician’s practice site (BRM database), and the physician’s medical specialty. At both baseline and follow-up, matches from the BRM data were identified for 94.0% of the patients who named a physician.

Identifying Voluntary Versus Involuntary Disenrollment

Patients were classified as having changed physicians during the study period if their follow-up questionnaire reported having been in their primary physician’s practice for less than 3 years and if the physician named at follow-up was different from the one named in 1996. Those who changed physicians were then classified as having switched either voluntarily or involuntarily. A switch was considered involuntary if: (1) the patient’s baseline physician was no longer listed as active in the Massachusetts BRM database (n=77), (2) the baseline physician had moved more than 10 miles (n=91), or (3) the patient had moved more than 15 miles from the baseline residence (n=62).

In addition we considered the possible involuntary nature of physician switches that occurred along with a change in health plan enrollment. Because the employer in our study did not force or even incentivize health plan changes during the study period (ie, there was a consistent offering of health plans and no notable changes in the employee contributions for coverage), respondents who were insured by the commonwealth throughout the study period did not incur any involuntary physician switching owing to employer-imposed health plan changes. Among respondents not insured by the commonwealth throughout the study (ie, respondents who left state employment [n=40] or deferred coverage [n=7]), there were 6 who changed physicians. Five of these did so while remaining in the same health plan and were thus coded as having voluntarily changed physicians. The remaining individual who both changed health plans and physicians was dropped from our analysis of voluntary disenrollment, since we were unable to ascertain whether the plan change forced a change in physician.

Statistical Analyses

We limited the analytic sample to patients who completed both the baseline and follow-up questionnaire, who identified a primary physician at baseline for whom a BRM database match was found, and who had either remained with their baseline physician throughout the study period or had voluntarily left the physician’s practice (n=3052). Patients who had involuntarily disenrolled from their baseline physician’s practice (n=230) were excluded. Their exclusion was necessary, since there was no way to determine whether those who involuntarily switched physicians would have otherwise voluntarily left their physician. The sociodemographic and health profile of the analytic sample (n=3052) did not differ from that of the complete 1999 sample (n=4108).

Multiple logistic regression methods were used to evaluate interpersonal and structural features of care, as measured by the baseline PCAS scales, as predictors of voluntary disenrollment from a physician’s practice. All scales were standardized ([X1-mean]/standard deviation) to permit direct comparison of results across scales. First, the 8 PCAS scales were tested individually as predictors of voluntary disenrollment. Testing scales independently in this way is useful in cases such as this where moderate to high correlations exist among some scales. Although the majority of PCAS scale correlations are small, higher correlations exist among some scales (r=0.40-0.86). We applied the Bonferroni correction for multiple comparisons to this set of analyses.

Next, we modeled voluntary disenrollment as a function of the 4 relationship-quality scales together and tested for the equality of their effects (odd ratio [OR]) using a chi-square test. We repeated this using the 4 measures of structural features of care. Finally, using factor analysis methods (principal factor), we explored the potential for defining a single factor denoting relationship quality and a single factor denoting structural features of care. The 4 scales denoting structural features of care failed to generate an acceptable factor (range of factor loadings=0.20 [relationship duration] to 0.67 [access to care]), so this factor was dropped. The relationship-quality factor was retained (range of factor loadings=0.84 [knowledge of patient] to 0.92 [communication]) and tested in multiple logistic regression along with each of the 4 structure-of-care measures. A chi-square test was used to test the equivalence of the effects (OR) associated with the relationship quality factor and each of the 4 structure-of-care scales.

All regression models controlled for patients’ baseline sociodemographic profile (age, sex, race, years of education, household income), baseline health status (physical functioning, mental functioning, number of primary care sensitive conditions [PCSC], and number of primary care insensitive conditions [PCIC]), and baseline utilization (number of ambulatory visits in the previous 6 months). Physical and mental functioning were measured with data from the Medical Outcomes Study Short Form-12 (SF-12) Health Survey, which was included in the patient questionnaire. The numbers of primary care sensitive and insensitive conditions were classified using patients’ baseline reports about 20 chronic medical conditions with high prevalence among adults in the United States. The classification of PCSC and PCIC was defined by 9 generalist physicians, blind to the study objectives, who were asked to identify those conditions for which good primary care management could substantially affect outcomes (PCSC) and those for which it could not (PCIC). PCSC included hypertension, recent myocardial infarction, congestive heart failure, diabetes, angina, migraines, seasonal allergies, asthma, ulcers, arthritis, cancer, back pain, weight problem, and depression. PCIC included blindness, deafness, liver disease, insomnia, nonseasonal allergies (eg, dust, food, pets), and limb paralysis or amputation.

We assessed the goodness-of-fit of the final models using the Hosmer and Lemeshow method. For each scale, the P on the chi-square test statistic was greater than .05, indicating that the model fit the data well.

Results

Slightly more than one fourth of the patients in the longitudinal study panel changed physicians during the 3-year follow-up period (n=899), while approximately three fourths remained with their baseline physician throughout the study (n=2383). Of those who changed physicians, most changed voluntarily (n=669), but some changed involuntarily (n=230) because the physician had moved, retired, died, or the patient had moved a substantial distance. Table 1 shows the unadjusted sociodemographic, health, and utilization characteristics of the analytic sample, comparing those who voluntarily changed physicians with those who remained with their baseline physician throughout our study. Voluntary disenrollees were younger and more likely to be women and nonwhite than those who stayed with their baseline physician (P <.01). There were no differences in the baseline health status or outpatient utilization of the 2 groups.

Table 2 presents the results of the regression analyses examining the 8 PCAS scales as individual predictors of voluntary disenrollment (column 1) and the results of a multivariable model, including the composite relationship-quality factor (RQ) and the 4 structure-of-care scales as predictors of voluntary disenrollment (column 2, columns 3-7). When all scales were modeled independently (column 1), each was a significant predictor of voluntary disenrollment (P <.001), with somewhat larger effects associated with the relationship quality scales (OR=1.49-1.56) than the structure-of-care scales (OR=1.29-1.44). Pairwise tests of the ORs associated with each of the 4 relationship quality scales indicated that they were statistically equivalent in their ability to predict voluntary disenrollment. When the 4 indicators of relationship quality were included together in a multiple regression model, a chi-square test of their effects (OR) revealed the 4 to be statistically equivalent predictors of voluntary disenrollment. Similarly, in a model including the 4 structure-of-care scales, chi-square testing showed these 4 variables to have statistically equivalent effects. With the exception of sex, patient characteristics (sociodemographics, health, utilization) did not significantly predict voluntary disenrollment in any of these models. The gender effect had marginal significance in most cases (.05

Table 2 (column 2) shows the results of modeling voluntary disenrollment as a function of both relationship quality and structure-of-care together. In that multivariable model, the composite relationship quality factor (RQ) emerged as the leading predictor of voluntary disenrollment (OR=1.59; P <.001). This OR signifies that a standard deviation (SD) decline in relationship quality was associated with a 59% increase in the odds of voluntary disenrollment. The results indicate that after accounting for patients’ baseline characteristics (sociodemographic, health, and utilization) and the 4 structural features of care, patients with relationship quality scores in the 5th percentile in 1996 were 3 times more likely to voluntarily disenroll from their physician’s practice than those with 95th percentile relationship quality scores (37.8% vs 12.2%). The 2 measures of continuity also significantly predicted disenrollment in the multivariable model (visit-based continuity: OR=1.14, P=.03; relationship duration: OR=1.16, P=.01). Access to care predicted disenrollment with marginal significance (OR=1.14; P=.08), and integration did not significantly predict disenrollment (P=.59) in this model. None of the patient characteristics (sociodemograhics, health, utilization) significantly predicted disenrollment in the presence of these 5 quality-of-care measures.

Discussion

In our observational study of insured employed adults, 20% of the patients voluntarily left their primary care physician’s practice over a 3-year period. Another 5% left involuntarily, owing to factors that forced a change (eg, the physician moved, retired, or died). For the average full-time primary physician, this translates into approximately 400 patients voluntarily leaving the practice over a given 3-year period and another 100 leaving involuntarily. Rates of involuntary switching are almost certainly higher among physicians whose patients face more employer-imposed disruptions than occurred in our study population.

Our data indicate that the quality of the physician-patient relationship significantly predicts patients’ loyalty. With patient characteristics and structural features of care taken into account, those with the poorest-quality physician-patient relationships in 1996 were 3 times more likely to leave the physician’s practice over the ensuing 3 years than those with the highest-quality relationships.

Structural features of care also emerged as important determinants of patients’ disenrollment decisions. When considered independently of relationship quality, each of the 4 structural elements of care significantly predicted voluntary disenrollment. With relationship quality taken into account, continuity of care (both relationship duration and visit-based continuity) remained significant predictors of disenrollment, while access to care and integration of care did not. The results suggest that although these patients put a high priority on being given timely and convenient access to their physician’s office, the issue of who they are given access to and the quality of their connection with that clinician mattered more.

Our findings are consistent with those reported more than 2 decades ago by Kasteler and colleagues,¹⁶ who found both interpersonal quality of care and structural features of care to be significantly associated with voluntary physician switching in a cross-sectional study. Our study has the advantage of longitudinal data through which the sequencing of effects is clear. In addition, our study advances beyond earlier studies that evaluated a single patient-based measure of care in predicting disenrollment.^17,25 Marquis and coworkers¹⁷ showed that patients’ general satisfaction with their physician predicted disenrollment from the physician’s practice over the following year. Thom and colleagues²⁵ found that patients’ trust in their physicians significantly predicted disenrollment over the next 6 months. Our study includes measures of 8 characteristics that encompass the defining features of primary care as posited by the IOM1 and others,^26-29 with several features for which the relationship to disenrollment have not been previously studied. Our study contributes evidence concerning both the absolute and relative importance of interpersonal and structural features of care as predictors of patients’ loyalty to their primary care physician’s practice.

Limitations

Our study is limited to a population of adults in Massachusetts who were employed and insured at baseline. Rates of involuntary physician switching in this population were likely lower than would be observed in other employed populations (particularly in competitive health care markets) for the reasons mentioned (ie, benefit policies that minimized employer-imposed disruption of employees’ health care arrangements). However, the observed rates of voluntary physician switching and the predictors of voluntary disenrollment should not be affected. Those findings may be presumed to generalize more broadly.

A second limitation is the absence of information about salient health events that occurred between the baseline and follow-up phases of our study. For patients who incurred a serious episode of illness, information about the intervening health events and their experiences with their physician during that time might have enhanced our understanding of the factors that influenced their decisions about whether to remain in that physician’s practice.

Similarly, the study lacked detailed indicators of the technical quality of care provided and therefore could not assess the role that technical quality—and patients’ perceptions of it—play in shaping patients’ loyalty to their physician.

Finally, our study could not fully account for one potential source of involuntary disenrollment: patients leaving practices because the physicians no longer accepted their health plan. However, rates of physician turnover during the study period were no more than 5% in any of the health plans studied and were substantially lower in most.²¹ Thus, our findings are unlikely to have been substantively altered by a detailed accounting of this form of involuntary physician switching.

Conclusions

Previous empirical research has underscored the importance of physician-patient relationship quality by demonstrating its association with important outcomes, including adherence to medical advice,^19,30-32 satisfaction with care,^19,33-35 and litigation against physicians.^36-38 However, few studies have had the benefit of longitudinal data with which to verify the sequencing of effects between relationship quality and outcomes.

In our study the strength of physician-patient relationships in primary care—as indicated by patients’ trust in their physician, their assessments of how well the physician knows them, and the quality of communication and interpersonal treatment—was the leading predictor of patients’ loyalty to their primary physician’s practice. Continuity of care also significantly predicted voluntary disenrollment. The findings are noteworthy against a backdrop of health care delivery changes nationwide that many describe as threatening the therapeutic alliance between the physician and the patient.^29,39-42

The recent IOM report on the future of primary care called attention to the importance of the physician-patient relationship in primary care, asserting that primary care is predicated on sustained clinician-patient partnerships and on a whole-person orientation.1 In our study, longitudinal data demonstrate that the strength of connection between a patient and his or her primary care physician significantly predicts the likelihood of that patient remaining in that physician’s practice (vs voluntarily leaving) over the next several years. In an era marked by increasing pressure on clinicians and health care organizations to attend to such factors as market share, productivity, and efficiency, these findings point to a set of attributes that might otherwise be overlooked. They suggest that medical practices and health plans cannot afford to ignore that the essence of medical care delivery involves the interaction of one human being with another.

Acknowledgments

This research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, executive director of the Massachusetts Group Insurance Commission, whose commitment and participation have made this study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study. Finally, we acknowledge each of the health plans involved in the study, and particularly the health plan executives who served on our Advisory Committee and those who provided interviews, for generously giving their time, insights, and information that was critical to our research.

The presence of sustained relationships between physicians and patients is a defining characteristic of primary care.¹ Family physicians use these relationships to acquire the depth of medical and personal knowledge about a patient that is essential to primary care practice.² It is also the reason some physicians choose this area of medicine.

A substantial body of empirical research points to the value of continuity in the physician-patient relationship, particularly in primary care. The benefits of continuity have been shown to accrue in the form of cost savings, improved health outcomes, and greater satisfaction for patients and physicians.^3-15 Yet little empirical research exists to indicate the amount of physician switching that occurs in primary care or the reasons for it.

In 1976 Kastler and colleagues¹⁶ examined the association between patients’ assessments of their care and their “doctor shopping” behavior. They found that patients’ evaluations of both interpersonal and structural features of care were significantly associated with the likelihood of voluntarily changing physicians. Those authors did not attempt to determine the relative importance of the 2 domains with respect to physician switching. The cross-sectional design precluded the study from determining which factor (if either) prospectively predicted switching.

Marquis and coworkers¹⁷ studied the sequencing of the satisfaction-disenrollment relationship using longitudinal data from the RAND Health Insurance Experiment (HIE). The HIE data showed that patients’ general satisfaction with their medical care significantly predicted physician switching over the following year. However, the HIE data did not afford the ability to differentiate among the many components of patient satisfaction and to discern which aspects specifically drive disenrollment.

Thus, little is known about the relative importance of the many factors that shape patients’ overall satisfaction with their physician and the extent to which performance on any of these ultimately drives a patient’s decision to leave a physician’s practice. Moreover, these earlier studies pre-date the recent surge in managed care enrollment and in consumerism among patients, both of which are presumed to be having a substantial impact on the rates of physician switching and the reasons for it. The generalizability of earlier findings to the present circumstances is unclear.

Methods

Our longitudinal observational study includes a population of insured adults who were employed by the Commonwealth of Massachusetts at baseline (1996), completed a self-administered questionnaire at baseline and follow-up (1999), and reported having a regular personal physician at baseline. Between January 1996 and April 1996 the baseline questionnaire was administered to a random sample of commonwealth employees who subscribed to any of 12 health plans available to employees, their dependents, and retirees. A 68.5% response rate was achieved (n=7204) using a standard 3-stage mail survey protocol with limited telephone follow-up of nonrespondents (mail responses=6810; telephone responses=394). Further details of the baseline sampling and data collection methods are documented elsewhere.

Follow-up data collection occurred precisely 3 years after baseline (January 1999-April 1999). Respondents who identified a primary care physician at baseline and participated by mail were eligible for follow-up (n=6075). Data were obtained using a standard 3-stage mail survey protocol with a final targeted mailing to racial and ethnic minorities (n=311) and to those without a college diploma (n=521). The targeted mailings were done because these subgroups were found to be underrepresented among follow-up respondents near the conclusion of data collection, and their representation in the longitudinal sample was important to our objectives. A 69.4% response rate was achieved in follow-up (n=4108) after accounting for respondents who died (n=21), were too ill to participate (n=2), or could not be located by mail in 1999 (n=136). At baseline and follow-up, respondents were somewhat older than nonrespondents, more likely to be women and white, and less likely to be poor (Appendix, Table 1A.

The questionnaire administered to patients at both baseline and follow-up included 4 scales measuring features of the physician-patient relationship (quality of communication, interpersonal treatment, physician’s knowledge of the patient, and patient trust) and 4 scales measuring structural aspects of care (access to care, visit-based continuity, duration of primary care relationship, and integration of care). The 8 scales are part of the Primary Care Assessment Survey (PCAS), a validated questionnaire with measures corresponding to the defining features of primary care posited by the Institute of Medicine (IOM). All concepts are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit-specific). All PCAS scales are scored on a 0 to 100 scale, with higher scores indicating more of the referent attribute. Details of the development and psychometric performance of the PCAS scales are available elsewhere. The item content and reliability coefficient (Cronbach a)for each scale are summarized in the Appendix Table 2A. In addition to completing the PCAS items referencing their experiences with and assessments of their primary physicians, the respondents also provided their physicians’ names.

Using the physician-identifying information provided by the patients, we linked data from the Massachusetts Board of Registration in Medicine (BRM) to the study database. The BRM data provided the physician’s practice address and several characteristics of the physician’s training and practice. We linked with the BRM data by using a matching algorithm based on the spelling of the physician’s name as provided by the patient, the distance between the patient’s home ZIP code and the physician’s practice site (BRM database), and the physician’s medical specialty. At both baseline and follow-up, matches from the BRM data were identified for 94.0% of the patients who named a physician.

Identifying Voluntary Versus Involuntary Disenrollment

Patients were classified as having changed physicians during the study period if their follow-up questionnaire reported having been in their primary physician’s practice for less than 3 years and if the physician named at follow-up was different from the one named in 1996. Those who changed physicians were then classified as having switched either voluntarily or involuntarily. A switch was considered involuntary if: (1) the patient’s baseline physician was no longer listed as active in the Massachusetts BRM database (n=77), (2) the baseline physician had moved more than 10 miles (n=91), or (3) the patient had moved more than 15 miles from the baseline residence (n=62).

In addition we considered the possible involuntary nature of physician switches that occurred along with a change in health plan enrollment. Because the employer in our study did not force or even incentivize health plan changes during the study period (ie, there was a consistent offering of health plans and no notable changes in the employee contributions for coverage), respondents who were insured by the commonwealth throughout the study period did not incur any involuntary physician switching owing to employer-imposed health plan changes. Among respondents not insured by the commonwealth throughout the study (ie, respondents who left state employment [n=40] or deferred coverage [n=7]), there were 6 who changed physicians. Five of these did so while remaining in the same health plan and were thus coded as having voluntarily changed physicians. The remaining individual who both changed health plans and physicians was dropped from our analysis of voluntary disenrollment, since we were unable to ascertain whether the plan change forced a change in physician.

Statistical Analyses

We limited the analytic sample to patients who completed both the baseline and follow-up questionnaire, who identified a primary physician at baseline for whom a BRM database match was found, and who had either remained with their baseline physician throughout the study period or had voluntarily left the physician’s practice (n=3052). Patients who had involuntarily disenrolled from their baseline physician’s practice (n=230) were excluded. Their exclusion was necessary, since there was no way to determine whether those who involuntarily switched physicians would have otherwise voluntarily left their physician. The sociodemographic and health profile of the analytic sample (n=3052) did not differ from that of the complete 1999 sample (n=4108).

Multiple logistic regression methods were used to evaluate interpersonal and structural features of care, as measured by the baseline PCAS scales, as predictors of voluntary disenrollment from a physician’s practice. All scales were standardized ([X1-mean]/standard deviation) to permit direct comparison of results across scales. First, the 8 PCAS scales were tested individually as predictors of voluntary disenrollment. Testing scales independently in this way is useful in cases such as this where moderate to high correlations exist among some scales. Although the majority of PCAS scale correlations are small, higher correlations exist among some scales (r=0.40-0.86). We applied the Bonferroni correction for multiple comparisons to this set of analyses.

Next, we modeled voluntary disenrollment as a function of the 4 relationship-quality scales together and tested for the equality of their effects (odd ratio [OR]) using a chi-square test. We repeated this using the 4 measures of structural features of care. Finally, using factor analysis methods (principal factor), we explored the potential for defining a single factor denoting relationship quality and a single factor denoting structural features of care. The 4 scales denoting structural features of care failed to generate an acceptable factor (range of factor loadings=0.20 [relationship duration] to 0.67 [access to care]), so this factor was dropped. The relationship-quality factor was retained (range of factor loadings=0.84 [knowledge of patient] to 0.92 [communication]) and tested in multiple logistic regression along with each of the 4 structure-of-care measures. A chi-square test was used to test the equivalence of the effects (OR) associated with the relationship quality factor and each of the 4 structure-of-care scales.

All regression models controlled for patients’ baseline sociodemographic profile (age, sex, race, years of education, household income), baseline health status (physical functioning, mental functioning, number of primary care sensitive conditions [PCSC], and number of primary care insensitive conditions [PCIC]), and baseline utilization (number of ambulatory visits in the previous 6 months). Physical and mental functioning were measured with data from the Medical Outcomes Study Short Form-12 (SF-12) Health Survey, which was included in the patient questionnaire. The numbers of primary care sensitive and insensitive conditions were classified using patients’ baseline reports about 20 chronic medical conditions with high prevalence among adults in the United States. The classification of PCSC and PCIC was defined by 9 generalist physicians, blind to the study objectives, who were asked to identify those conditions for which good primary care management could substantially affect outcomes (PCSC) and those for which it could not (PCIC). PCSC included hypertension, recent myocardial infarction, congestive heart failure, diabetes, angina, migraines, seasonal allergies, asthma, ulcers, arthritis, cancer, back pain, weight problem, and depression. PCIC included blindness, deafness, liver disease, insomnia, nonseasonal allergies (eg, dust, food, pets), and limb paralysis or amputation.

We assessed the goodness-of-fit of the final models using the Hosmer and Lemeshow method. For each scale, the P on the chi-square test statistic was greater than .05, indicating that the model fit the data well.

Results

Slightly more than one fourth of the patients in the longitudinal study panel changed physicians during the 3-year follow-up period (n=899), while approximately three fourths remained with their baseline physician throughout the study (n=2383). Of those who changed physicians, most changed voluntarily (n=669), but some changed involuntarily (n=230) because the physician had moved, retired, died, or the patient had moved a substantial distance. Table 1 shows the unadjusted sociodemographic, health, and utilization characteristics of the analytic sample, comparing those who voluntarily changed physicians with those who remained with their baseline physician throughout our study. Voluntary disenrollees were younger and more likely to be women and nonwhite than those who stayed with their baseline physician (P <.01). There were no differences in the baseline health status or outpatient utilization of the 2 groups.

Table 2 presents the results of the regression analyses examining the 8 PCAS scales as individual predictors of voluntary disenrollment (column 1) and the results of a multivariable model, including the composite relationship-quality factor (RQ) and the 4 structure-of-care scales as predictors of voluntary disenrollment (column 2, columns 3-7). When all scales were modeled independently (column 1), each was a significant predictor of voluntary disenrollment (P <.001), with somewhat larger effects associated with the relationship quality scales (OR=1.49-1.56) than the structure-of-care scales (OR=1.29-1.44). Pairwise tests of the ORs associated with each of the 4 relationship quality scales indicated that they were statistically equivalent in their ability to predict voluntary disenrollment. When the 4 indicators of relationship quality were included together in a multiple regression model, a chi-square test of their effects (OR) revealed the 4 to be statistically equivalent predictors of voluntary disenrollment. Similarly, in a model including the 4 structure-of-care scales, chi-square testing showed these 4 variables to have statistically equivalent effects. With the exception of sex, patient characteristics (sociodemographics, health, utilization) did not significantly predict voluntary disenrollment in any of these models. The gender effect had marginal significance in most cases (.05

Table 2 (column 2) shows the results of modeling voluntary disenrollment as a function of both relationship quality and structure-of-care together. In that multivariable model, the composite relationship quality factor (RQ) emerged as the leading predictor of voluntary disenrollment (OR=1.59; P <.001). This OR signifies that a standard deviation (SD) decline in relationship quality was associated with a 59% increase in the odds of voluntary disenrollment. The results indicate that after accounting for patients’ baseline characteristics (sociodemographic, health, and utilization) and the 4 structural features of care, patients with relationship quality scores in the 5th percentile in 1996 were 3 times more likely to voluntarily disenroll from their physician’s practice than those with 95th percentile relationship quality scores (37.8% vs 12.2%). The 2 measures of continuity also significantly predicted disenrollment in the multivariable model (visit-based continuity: OR=1.14, P=.03; relationship duration: OR=1.16, P=.01). Access to care predicted disenrollment with marginal significance (OR=1.14; P=.08), and integration did not significantly predict disenrollment (P=.59) in this model. None of the patient characteristics (sociodemograhics, health, utilization) significantly predicted disenrollment in the presence of these 5 quality-of-care measures.

Discussion

In our observational study of insured employed adults, 20% of the patients voluntarily left their primary care physician’s practice over a 3-year period. Another 5% left involuntarily, owing to factors that forced a change (eg, the physician moved, retired, or died). For the average full-time primary physician, this translates into approximately 400 patients voluntarily leaving the practice over a given 3-year period and another 100 leaving involuntarily. Rates of involuntary switching are almost certainly higher among physicians whose patients face more employer-imposed disruptions than occurred in our study population.

Our data indicate that the quality of the physician-patient relationship significantly predicts patients’ loyalty. With patient characteristics and structural features of care taken into account, those with the poorest-quality physician-patient relationships in 1996 were 3 times more likely to leave the physician’s practice over the ensuing 3 years than those with the highest-quality relationships.

Structural features of care also emerged as important determinants of patients’ disenrollment decisions. When considered independently of relationship quality, each of the 4 structural elements of care significantly predicted voluntary disenrollment. With relationship quality taken into account, continuity of care (both relationship duration and visit-based continuity) remained significant predictors of disenrollment, while access to care and integration of care did not. The results suggest that although these patients put a high priority on being given timely and convenient access to their physician’s office, the issue of who they are given access to and the quality of their connection with that clinician mattered more.

Our findings are consistent with those reported more than 2 decades ago by Kasteler and colleagues,¹⁶ who found both interpersonal quality of care and structural features of care to be significantly associated with voluntary physician switching in a cross-sectional study. Our study has the advantage of longitudinal data through which the sequencing of effects is clear. In addition, our study advances beyond earlier studies that evaluated a single patient-based measure of care in predicting disenrollment.^17,25 Marquis and coworkers¹⁷ showed that patients’ general satisfaction with their physician predicted disenrollment from the physician’s practice over the following year. Thom and colleagues²⁵ found that patients’ trust in their physicians significantly predicted disenrollment over the next 6 months. Our study includes measures of 8 characteristics that encompass the defining features of primary care as posited by the IOM1 and others,^26-29 with several features for which the relationship to disenrollment have not been previously studied. Our study contributes evidence concerning both the absolute and relative importance of interpersonal and structural features of care as predictors of patients’ loyalty to their primary care physician’s practice.

Limitations

Our study is limited to a population of adults in Massachusetts who were employed and insured at baseline. Rates of involuntary physician switching in this population were likely lower than would be observed in other employed populations (particularly in competitive health care markets) for the reasons mentioned (ie, benefit policies that minimized employer-imposed disruption of employees’ health care arrangements). However, the observed rates of voluntary physician switching and the predictors of voluntary disenrollment should not be affected. Those findings may be presumed to generalize more broadly.

A second limitation is the absence of information about salient health events that occurred between the baseline and follow-up phases of our study. For patients who incurred a serious episode of illness, information about the intervening health events and their experiences with their physician during that time might have enhanced our understanding of the factors that influenced their decisions about whether to remain in that physician’s practice.

Similarly, the study lacked detailed indicators of the technical quality of care provided and therefore could not assess the role that technical quality—and patients’ perceptions of it—play in shaping patients’ loyalty to their physician.

Finally, our study could not fully account for one potential source of involuntary disenrollment: patients leaving practices because the physicians no longer accepted their health plan. However, rates of physician turnover during the study period were no more than 5% in any of the health plans studied and were substantially lower in most.²¹ Thus, our findings are unlikely to have been substantively altered by a detailed accounting of this form of involuntary physician switching.

Conclusions

Previous empirical research has underscored the importance of physician-patient relationship quality by demonstrating its association with important outcomes, including adherence to medical advice,^19,30-32 satisfaction with care,^19,33-35 and litigation against physicians.^36-38 However, few studies have had the benefit of longitudinal data with which to verify the sequencing of effects between relationship quality and outcomes.

In our study the strength of physician-patient relationships in primary care—as indicated by patients’ trust in their physician, their assessments of how well the physician knows them, and the quality of communication and interpersonal treatment—was the leading predictor of patients’ loyalty to their primary physician’s practice. Continuity of care also significantly predicted voluntary disenrollment. The findings are noteworthy against a backdrop of health care delivery changes nationwide that many describe as threatening the therapeutic alliance between the physician and the patient.^29,39-42

The recent IOM report on the future of primary care called attention to the importance of the physician-patient relationship in primary care, asserting that primary care is predicated on sustained clinician-patient partnerships and on a whole-person orientation.1 In our study, longitudinal data demonstrate that the strength of connection between a patient and his or her primary care physician significantly predicts the likelihood of that patient remaining in that physician’s practice (vs voluntarily leaving) over the next several years. In an era marked by increasing pressure on clinicians and health care organizations to attend to such factors as market share, productivity, and efficiency, these findings point to a set of attributes that might otherwise be overlooked. They suggest that medical practices and health plans cannot afford to ignore that the essence of medical care delivery involves the interaction of one human being with another.

Acknowledgments

This research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, executive director of the Massachusetts Group Insurance Commission, whose commitment and participation have made this study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study. Finally, we acknowledge each of the health plans involved in the study, and particularly the health plan executives who served on our Advisory Committee and those who provided interviews, for generously giving their time, insights, and information that was critical to our research.

The quality of physician-patient relationships alters health outcomes,^1-3 affects patients’ willingness to comply with medical advice or treatment,^4,5 and influences patients’ pursuit of malpractice suits.^6,7 Changes that reflect a decline in patients’ experience of structural and organizational aspects of care are important, because these areas are strong determinants of patient satisfaction.⁸

There is little question that health care delivery systems have undergone tremendous change over the past decade and a half. These changes have affected multiple aspects of medical practice, including financial incentives faced by individual clinicians, the organization of medical practices, and the corporate relationships among provider organizations. Primary care and its position within the health care delivery system has been a focal point for much of the change. Under most forms of insurance primary care physicians now hold a central role in patient care, responsible for coordinating and integrating all aspects of the care provided to their panel of patients and in some cases sharing in the financial risk associated with providing care under a capitated budget arrangement. As changes in the organization and financing of health care have unfolded (most notably over the past several years) they have almost always had direct implications for the primary care physician’s role in interacting with patients. During our study period of 1996 to 1999, these changes in the Commonwealth of Massachusetts have included the restructuring or merging of plans and their member practices, publicly reported financial difficulties, and the departure of plans from the market region. Even within stable plans, primary care physicians have experienced pressures to increase productivity, decrease costs, and attend to patient satisfaction.

We measured changes in patients’ experience of primary care with their primary care physicians over a 3-year study period in the Commonwealth of Massachusetts. We used indices of primary care quality and studied a panel of insured adults who provided detailed information about their care. Both the quality of their relationships with their primary care physicians and their experience with organizational features of care (access, continuity, integration) were monitored during the study period.

Methods

A sample of insured employees who responded to a mailed questionnaire at baseline (1996) and again 3 years later (1999) comprised the population of this longitudinal observational study. The participants belonged to 1 of 12 insurance plans. These were representative of the major health plans in the state. The questionnaire included the Primary Care Assessment Survey (PCAS),⁹ a validated patient-completed questionnaire that measures 7 essential characteristics of primary care, defined by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ All PCAS scales are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit specific).⁹ In these analyses, we examined changes in the 8 PCAS scales over a 3-year study period. The scales that we examined cover 2 broad aspects of the patient’s primary care experience: the quality of the primary care relationship (4 scales: quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust) and organizational features of care (4 scales: financial access, organizational access, visit-based continuity, integration of care). Table 1shows the item content of each scale.

Baseline data were obtained between January and April 1996. Using a 3-stage mail survey that included an initial mailing and 2 additional mailings to nonrespondents and limited telephone follow-up of randomly selected nonrespondents,¹¹ the PCAS was administered to a random sample of 10,733 Commonwealth of Massachusetts employees stratified by age, health plan, and ZIP code. Of the original sample, 221 were excluded as either unable to be located by mail (n=184), deceased (n=11), or no longer a Commonwealth of Massachusetts employee (n=26). In total, 6810 adults completed the baseline questionnaire by mail, and 394 completed it by telephone (response rate=68.5%).

Follow-up data were collected between January and April 1999. The follow-up questionnaire was administered to all baseline study participants who had identified a primary care physician and whose physician was listed in the Massachusetts Board of Registration in Medicine registry of licensed physicians (n=6075). Follow-up data collection employed a 3-step mail survey protocol as at baseline and was supplemented with final targeted mailings to 2 groups of nonrespondents (ethnic minorities, n=31, and those without a college diploma, n=521). The targeted mailings were performed, when nearing the conclusion of data collection, these subgroups were found to be under-represented among follow-up respondents. Completed questionnaires were received from 69.4% of the eligible respondents at follow-up (n=4108). Data collection and entry at baseline and follow-up were managed by the Center for Survey Research, University of Massachusetts (Boston).

In addition to the PCAS measures, the baseline and follow-up questionnaires were used to ascertain the respondents’ sociodemographic profiles (age, sex, race, years of education, household income) and health status. Measures of health status included the Medical Outcomes Study Short Form-12 (SF-12) Health Survey¹² and a checklist of 20 chronic conditions with high prevalence among US adults.¹³

Statistical Analyses

The principal analytic objective was to study the changes in primary care experiences of patients in a sustained primary care relationship during the 3-year study period. Patients who had changed physicians were excluded from the analytic sample. By restricting our analyses to patients who remained with the same physician we were able to isolate changes in their care over the 3-year study period without confounding factors associated with changing physicians. The analytic sample included patients who completed both the baseline and follow-up questionnaires, who identified a primary physician at baseline, and who remained with that physician throughout the 3-year study period (n=2383).

In a comparison of baseline characteristics of 1996 baseline respondents (n=6810) and those who were retained at the 3-year follow-up period (n=4108), the 1999 respondents were on average a year older (49 vs 48 years). They were more likely to be women (57% vs 55%) and more likely to be white (89% vs 86%). Education and income levels and measures of mental and physical health were almost equal in 1996 and 1999.

After exclusions, baseline demographics and health indexes of the analytic sample still remained near equal to those of the overall 1999 respondent group. Members of the analytic sample were more likely to be women (57% vs 55%) and to have a baseline primary care physician relationship duration of more than 5 years (54% vs 49.6%) when compared with the overall 1999 respondents. Compared with respondents, nonrespondents were younger, more likely to be men, poorer, less educated, and of a racial group other than white.

Baseline characteristics of physicians of 1999 respondents were also examined. The specialties of physicians were listed in the Massachusetts Board of Registration as 19% family physicians, 65% internists, 3% obstetrician-gynecologists, 4% cardiologists, and another 9% spread across other specialties. The physicians of patients in our analytic sample had the same specialty group distribution as the overall sample. In comparison with the overall group of 1999 respondents’ physicians, the physicians of the analytic sample were more likely to be men (80% vs 75%) and earlier graduates. Differences are due in part to patients nominating residents in training programs as their primary care physicians (6% of the overall respondent sample, 0% of the analytic sample). These patients subsequently changed their primary care physicians and were not included in the analytic sample. Also, there is an under-representation of women in the older group of physicians who have stable long-term relationships with their patients.

For each patient, we computed the difference between the 1996 and 1999 scores on each of the 8 PCAS scales (quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust, financial access, organizational access, visit-based continuity, integration of care). We determined the unadjusted mean change in each scale and the 95% confidence interval around this change. To permit comparison across scales, we computed a standardized difference score (the effect size), which was the mean change of scales divided by the standard deviation of the referent scale scores at baseline.

Because previous empiric medical literature,^14,15 other research,^16-22 and our own cross-sectional data from baseline suggested that physician-patient relationship quality improves with increased relationship duration, we also examined the changes in indexes of relationship quality, controlling for the increased relationship length that occurred during our study for the group of patients who did not switch physicians. The adjusted changes in the relationship quality scales of communication, interpersonal treatment, physician knowledge of the patient, and trust were calculated using regression models. We used the longitudinal data in a stacked data set in which each patient’s pair of observation sets (from 1996 and 1999) was entered and included a binary variable indicating the survey round (round 1=0 and round 2=1). The stacked data make possible regression of the scale scores for each of the 4 indicators of relationship quality against a measure of relationship length, which (inherent in the selection criteria for the analytic sample) increased from the first to the second round of the survey. Inclusion of the survey round indicator in the 4 relevant regressions allowed a regression coefficient to be obtained for each scale, which reflected the magnitude of average difference in scale scores over the study period, adjusting by increases in the length of physician-patient relationship. P values for these regressions were used to evaluate the significance of the findings.

Results

The sociodemographic characteristics and health status of the analytic sample are presented in Table 2. Study participants ranged in age from 20 to 88 years, with a mean of 50.2 years. The majority were women and white, with some college education. On average they began our study with 2.8 chronic conditions. Physical and mental health status (as measured by the SF-12) was consistent with those observed nationally in adults in this age group.²³

Table 3 shows the unadjusted PCAS scores at baseline and follow-up and provides the 3-year score differences for the analytic sample. Two scales showed significant improvements (physician’s knowledge of the patient and visit-based continuity). Four scales showed significant declines (communication quality, interpersonal treatment, patient trust, and organizational access). The standardized measure of change (effect size [ES]) reveals that the largest changes occurred in organization access (ES=0.165), interpersonal treatment (ES=0.115), and communication quality (ES=0.095). The effect sizes for the other scales ranged from 0.016 (integration of care) to 0.060 (visit-based continuity).

For the 4 indicators of relationship quality, the observed (or unadjusted) change and change adjusted for the length of the physician-patient relationship are depicted in the Figure 1. The adjusted change scores from 1996 to 1999 show significant declines in all 4 indicators of relationship quality, ranging from -1.72 (physician’s knowledge of the patient) to -3.28 (interpersonal treatment).

Discussion

This observational study of patients under the continuing care of a primary physician from 1996 to 1999 found significant declines in 3 of the 4 indicators of relationship quality between 1996 and 1999. The largest declines were observed in interpersonal treatment, followed by declines in the quality of communication and trust. The fourth measure of relationship quality—the physician’s whole-person knowledge of the patient—increased, but this increase could not be demonstrated when adjusting for increased relationship duration. The adjusted figures demonstrate the concept that if expected increases in relationship quality due to increased relationship length are controlled for (ie, taken into account), then even larger decreases in relationship quality are demonstrated Figure 1.

Primary care is predicated on sustained physician-patient relationships, as recently noted by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ The importance of relationship quality in health care is underscored by a research literature that links it to important outcomes of care. The quality of the physician-patient relationship in primary care has been associated with outcomes that include patients’ compliance with medical advice,^4,5,24 clinical outcomes of care,^1,3 patients’ willingness to initiate malpractice suits,^6,7 and patients’ decisions to change physicians.^25-27 Interpersonal treatment is a correlate of patient satisfaction,^28,29 which is important to individual patient well-being and as a factor that results in patient disenrollment. Research literature establishes that effective communication builds trust, reduces patients’ emotional stress, facilitates the process of diagnosing medical conditions, affects medical management decisions, and creates positive health outcomes.^1,3,30-34 In this context the observed decline in some of the indicators of quality of primary care relationships across the 3-year study period is concerning. We do not know from our study whether the quality of relationships was already declining before 1996 and, more important, whether declines are continuing at this rate.

Previous analyses employing baseline PCAS indicators of relationship quality as predictors of outcomes of care in this study population²⁷ enable us to estimate with some caution the effects of the observed declines in relationship quality over time. On the basis of this previous evidence, had the observed declines in interpersonal treatment and communication not occurred, an estimated 5% of the rate of patients’ voluntary disenrollment from their physicians’ practices could have been avoided. The observed decline in interpersonal treatment could translate into a measurable decline in patients’ attempts to adhere to their physicians’ counseling about smoking cessation, reducing alcohol consumption, and increasing exercise.

Our study included 4 indicators of organizational/structure features of care, 2 of which were observed to change significantly during the study period. Visit-based continuity between patients and their primary physicians (the ability of patients to see their regular physician for routine care and appointments when sick) increased. Patients’ organizational access to care (which includes patients’ ability to reach their physician’s office by phone and to obtain timely appointments when sick) declined substantially—more so than any other scale in the study. Still, patients in this study sample were better able to see their own physician, and having done so they go on to report that the quality of the encounters is declining. Access to care is a defining feature of primary care^10,35 and an important correlate of patient satisfaction.^8,36 The observed decline in organizational access to care raises concerns about the quality of primary care.

Limitations

Our study population included employed, insured, and generally healthy adults and was not representative of more vulnerable groups. This effect is further accentuated because the nonrespondents are more likely to be from a more vulnerable population. Since the research literature suggests that minority status and low income have an adverse impact on physician-patient interaction,^37-39 it is likely our findings would have been demonstrated more strongly with the inclusion of data from this section of the population.

Our inclusion criteria create selection biases, which reduce generalizability for some scales. Visit-based continuity is most likely to be optimized in this group of patients who have named a regular physician and have stayed with them during the study period.

The observed changes in primary care performance, though statistically significant, are small. But they occur within a reported (ie, observed) range of scores that is approximately one third the size of the range of possible scores. The movement within this range represents a larger shift than the same shift in a more extensive range. In addition to allowing comparisons across scales, the standardized effect size also helps address this issue by representing the data as a proportion of the standard deviation. The usual Cohen classification of effect size is not as pertinent to these results, because it was not developed and described for population studies.⁴⁰ Thus, the changes we observed (1) reflect declines where increases would be expected, (2) reflect the shift of a population, and (3) may reflect an ongoing trend continuing beyond our study period.

We have viewed declines primarily as a result of a change in the patients’ experiences in contrast to a change in the patients. Patient attributes not considered may include: declines in levels of societal trust, raised patient expectations with increasing patient consumerism, and patient education influenced by the rapidly increasing patient access to information on the Internet during the study period. Our scale measures are specific to a domain and should not be as affected by external factors.

Conclusions

• Declines in primary care performance indicators were demonstrated by our study.
• These declines have been reported in an environment of change.
• Further research to examine the factors driving this decline in primary care quality is needed. The distractions of organizational restructuring, mergers, and departures from the market region, and pressures to increase productivity without compromising standards of care may be contributing factors.
• If quality of primary care performance continues to fall, the previously hoped for goals of health care reform through the advancement of primary care are at risk for being undermined.

Acknowledgments

Our research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, the executive director of the Massachusetts Group Insurance Commission, whose commitment to this research and participation in it made the study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study.

Related resources:

• HealthWatch http://healthwatch.medscape.com/medscape/p/gcommunity/ghome.asp
• HealthScout http://www.healthscout.com/cgi-bin/WebObjects/Af.woa
• WebMD http://www.my.webmd.com
• drkoop.com http://www.drkoop.com

The quality of physician-patient relationships alters health outcomes,^1-3 affects patients’ willingness to comply with medical advice or treatment,^4,5 and influences patients’ pursuit of malpractice suits.^6,7 Changes that reflect a decline in patients’ experience of structural and organizational aspects of care are important, because these areas are strong determinants of patient satisfaction.⁸

There is little question that health care delivery systems have undergone tremendous change over the past decade and a half. These changes have affected multiple aspects of medical practice, including financial incentives faced by individual clinicians, the organization of medical practices, and the corporate relationships among provider organizations. Primary care and its position within the health care delivery system has been a focal point for much of the change. Under most forms of insurance primary care physicians now hold a central role in patient care, responsible for coordinating and integrating all aspects of the care provided to their panel of patients and in some cases sharing in the financial risk associated with providing care under a capitated budget arrangement. As changes in the organization and financing of health care have unfolded (most notably over the past several years) they have almost always had direct implications for the primary care physician’s role in interacting with patients. During our study period of 1996 to 1999, these changes in the Commonwealth of Massachusetts have included the restructuring or merging of plans and their member practices, publicly reported financial difficulties, and the departure of plans from the market region. Even within stable plans, primary care physicians have experienced pressures to increase productivity, decrease costs, and attend to patient satisfaction.

We measured changes in patients’ experience of primary care with their primary care physicians over a 3-year study period in the Commonwealth of Massachusetts. We used indices of primary care quality and studied a panel of insured adults who provided detailed information about their care. Both the quality of their relationships with their primary care physicians and their experience with organizational features of care (access, continuity, integration) were monitored during the study period.

Methods

A sample of insured employees who responded to a mailed questionnaire at baseline (1996) and again 3 years later (1999) comprised the population of this longitudinal observational study. The participants belonged to 1 of 12 insurance plans. These were representative of the major health plans in the state. The questionnaire included the Primary Care Assessment Survey (PCAS),⁹ a validated patient-completed questionnaire that measures 7 essential characteristics of primary care, defined by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ All PCAS scales are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit specific).⁹ In these analyses, we examined changes in the 8 PCAS scales over a 3-year study period. The scales that we examined cover 2 broad aspects of the patient’s primary care experience: the quality of the primary care relationship (4 scales: quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust) and organizational features of care (4 scales: financial access, organizational access, visit-based continuity, integration of care). Table 1shows the item content of each scale.

Baseline data were obtained between January and April 1996. Using a 3-stage mail survey that included an initial mailing and 2 additional mailings to nonrespondents and limited telephone follow-up of randomly selected nonrespondents,¹¹ the PCAS was administered to a random sample of 10,733 Commonwealth of Massachusetts employees stratified by age, health plan, and ZIP code. Of the original sample, 221 were excluded as either unable to be located by mail (n=184), deceased (n=11), or no longer a Commonwealth of Massachusetts employee (n=26). In total, 6810 adults completed the baseline questionnaire by mail, and 394 completed it by telephone (response rate=68.5%).

Follow-up data were collected between January and April 1999. The follow-up questionnaire was administered to all baseline study participants who had identified a primary care physician and whose physician was listed in the Massachusetts Board of Registration in Medicine registry of licensed physicians (n=6075). Follow-up data collection employed a 3-step mail survey protocol as at baseline and was supplemented with final targeted mailings to 2 groups of nonrespondents (ethnic minorities, n=31, and those without a college diploma, n=521). The targeted mailings were performed, when nearing the conclusion of data collection, these subgroups were found to be under-represented among follow-up respondents. Completed questionnaires were received from 69.4% of the eligible respondents at follow-up (n=4108). Data collection and entry at baseline and follow-up were managed by the Center for Survey Research, University of Massachusetts (Boston).

In addition to the PCAS measures, the baseline and follow-up questionnaires were used to ascertain the respondents’ sociodemographic profiles (age, sex, race, years of education, household income) and health status. Measures of health status included the Medical Outcomes Study Short Form-12 (SF-12) Health Survey¹² and a checklist of 20 chronic conditions with high prevalence among US adults.¹³

Statistical Analyses

The principal analytic objective was to study the changes in primary care experiences of patients in a sustained primary care relationship during the 3-year study period. Patients who had changed physicians were excluded from the analytic sample. By restricting our analyses to patients who remained with the same physician we were able to isolate changes in their care over the 3-year study period without confounding factors associated with changing physicians. The analytic sample included patients who completed both the baseline and follow-up questionnaires, who identified a primary physician at baseline, and who remained with that physician throughout the 3-year study period (n=2383).

In a comparison of baseline characteristics of 1996 baseline respondents (n=6810) and those who were retained at the 3-year follow-up period (n=4108), the 1999 respondents were on average a year older (49 vs 48 years). They were more likely to be women (57% vs 55%) and more likely to be white (89% vs 86%). Education and income levels and measures of mental and physical health were almost equal in 1996 and 1999.

After exclusions, baseline demographics and health indexes of the analytic sample still remained near equal to those of the overall 1999 respondent group. Members of the analytic sample were more likely to be women (57% vs 55%) and to have a baseline primary care physician relationship duration of more than 5 years (54% vs 49.6%) when compared with the overall 1999 respondents. Compared with respondents, nonrespondents were younger, more likely to be men, poorer, less educated, and of a racial group other than white.

Baseline characteristics of physicians of 1999 respondents were also examined. The specialties of physicians were listed in the Massachusetts Board of Registration as 19% family physicians, 65% internists, 3% obstetrician-gynecologists, 4% cardiologists, and another 9% spread across other specialties. The physicians of patients in our analytic sample had the same specialty group distribution as the overall sample. In comparison with the overall group of 1999 respondents’ physicians, the physicians of the analytic sample were more likely to be men (80% vs 75%) and earlier graduates. Differences are due in part to patients nominating residents in training programs as their primary care physicians (6% of the overall respondent sample, 0% of the analytic sample). These patients subsequently changed their primary care physicians and were not included in the analytic sample. Also, there is an under-representation of women in the older group of physicians who have stable long-term relationships with their patients.

For each patient, we computed the difference between the 1996 and 1999 scores on each of the 8 PCAS scales (quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust, financial access, organizational access, visit-based continuity, integration of care). We determined the unadjusted mean change in each scale and the 95% confidence interval around this change. To permit comparison across scales, we computed a standardized difference score (the effect size), which was the mean change of scales divided by the standard deviation of the referent scale scores at baseline.

Because previous empiric medical literature,^14,15 other research,^16-22 and our own cross-sectional data from baseline suggested that physician-patient relationship quality improves with increased relationship duration, we also examined the changes in indexes of relationship quality, controlling for the increased relationship length that occurred during our study for the group of patients who did not switch physicians. The adjusted changes in the relationship quality scales of communication, interpersonal treatment, physician knowledge of the patient, and trust were calculated using regression models. We used the longitudinal data in a stacked data set in which each patient’s pair of observation sets (from 1996 and 1999) was entered and included a binary variable indicating the survey round (round 1=0 and round 2=1). The stacked data make possible regression of the scale scores for each of the 4 indicators of relationship quality against a measure of relationship length, which (inherent in the selection criteria for the analytic sample) increased from the first to the second round of the survey. Inclusion of the survey round indicator in the 4 relevant regressions allowed a regression coefficient to be obtained for each scale, which reflected the magnitude of average difference in scale scores over the study period, adjusting by increases in the length of physician-patient relationship. P values for these regressions were used to evaluate the significance of the findings.

Results

The sociodemographic characteristics and health status of the analytic sample are presented in Table 2. Study participants ranged in age from 20 to 88 years, with a mean of 50.2 years. The majority were women and white, with some college education. On average they began our study with 2.8 chronic conditions. Physical and mental health status (as measured by the SF-12) was consistent with those observed nationally in adults in this age group.²³

Table 3 shows the unadjusted PCAS scores at baseline and follow-up and provides the 3-year score differences for the analytic sample. Two scales showed significant improvements (physician’s knowledge of the patient and visit-based continuity). Four scales showed significant declines (communication quality, interpersonal treatment, patient trust, and organizational access). The standardized measure of change (effect size [ES]) reveals that the largest changes occurred in organization access (ES=0.165), interpersonal treatment (ES=0.115), and communication quality (ES=0.095). The effect sizes for the other scales ranged from 0.016 (integration of care) to 0.060 (visit-based continuity).

For the 4 indicators of relationship quality, the observed (or unadjusted) change and change adjusted for the length of the physician-patient relationship are depicted in the Figure 1. The adjusted change scores from 1996 to 1999 show significant declines in all 4 indicators of relationship quality, ranging from -1.72 (physician’s knowledge of the patient) to -3.28 (interpersonal treatment).

Discussion

This observational study of patients under the continuing care of a primary physician from 1996 to 1999 found significant declines in 3 of the 4 indicators of relationship quality between 1996 and 1999. The largest declines were observed in interpersonal treatment, followed by declines in the quality of communication and trust. The fourth measure of relationship quality—the physician’s whole-person knowledge of the patient—increased, but this increase could not be demonstrated when adjusting for increased relationship duration. The adjusted figures demonstrate the concept that if expected increases in relationship quality due to increased relationship length are controlled for (ie, taken into account), then even larger decreases in relationship quality are demonstrated Figure 1.

Primary care is predicated on sustained physician-patient relationships, as recently noted by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ The importance of relationship quality in health care is underscored by a research literature that links it to important outcomes of care. The quality of the physician-patient relationship in primary care has been associated with outcomes that include patients’ compliance with medical advice,^4,5,24 clinical outcomes of care,^1,3 patients’ willingness to initiate malpractice suits,^6,7 and patients’ decisions to change physicians.^25-27 Interpersonal treatment is a correlate of patient satisfaction,^28,29 which is important to individual patient well-being and as a factor that results in patient disenrollment. Research literature establishes that effective communication builds trust, reduces patients’ emotional stress, facilitates the process of diagnosing medical conditions, affects medical management decisions, and creates positive health outcomes.^1,3,30-34 In this context the observed decline in some of the indicators of quality of primary care relationships across the 3-year study period is concerning. We do not know from our study whether the quality of relationships was already declining before 1996 and, more important, whether declines are continuing at this rate.

Previous analyses employing baseline PCAS indicators of relationship quality as predictors of outcomes of care in this study population²⁷ enable us to estimate with some caution the effects of the observed declines in relationship quality over time. On the basis of this previous evidence, had the observed declines in interpersonal treatment and communication not occurred, an estimated 5% of the rate of patients’ voluntary disenrollment from their physicians’ practices could have been avoided. The observed decline in interpersonal treatment could translate into a measurable decline in patients’ attempts to adhere to their physicians’ counseling about smoking cessation, reducing alcohol consumption, and increasing exercise.

Our study included 4 indicators of organizational/structure features of care, 2 of which were observed to change significantly during the study period. Visit-based continuity between patients and their primary physicians (the ability of patients to see their regular physician for routine care and appointments when sick) increased. Patients’ organizational access to care (which includes patients’ ability to reach their physician’s office by phone and to obtain timely appointments when sick) declined substantially—more so than any other scale in the study. Still, patients in this study sample were better able to see their own physician, and having done so they go on to report that the quality of the encounters is declining. Access to care is a defining feature of primary care^10,35 and an important correlate of patient satisfaction.^8,36 The observed decline in organizational access to care raises concerns about the quality of primary care.

Limitations

Our study population included employed, insured, and generally healthy adults and was not representative of more vulnerable groups. This effect is further accentuated because the nonrespondents are more likely to be from a more vulnerable population. Since the research literature suggests that minority status and low income have an adverse impact on physician-patient interaction,^37-39 it is likely our findings would have been demonstrated more strongly with the inclusion of data from this section of the population.

Our inclusion criteria create selection biases, which reduce generalizability for some scales. Visit-based continuity is most likely to be optimized in this group of patients who have named a regular physician and have stayed with them during the study period.

The observed changes in primary care performance, though statistically significant, are small. But they occur within a reported (ie, observed) range of scores that is approximately one third the size of the range of possible scores. The movement within this range represents a larger shift than the same shift in a more extensive range. In addition to allowing comparisons across scales, the standardized effect size also helps address this issue by representing the data as a proportion of the standard deviation. The usual Cohen classification of effect size is not as pertinent to these results, because it was not developed and described for population studies.⁴⁰ Thus, the changes we observed (1) reflect declines where increases would be expected, (2) reflect the shift of a population, and (3) may reflect an ongoing trend continuing beyond our study period.

We have viewed declines primarily as a result of a change in the patients’ experiences in contrast to a change in the patients. Patient attributes not considered may include: declines in levels of societal trust, raised patient expectations with increasing patient consumerism, and patient education influenced by the rapidly increasing patient access to information on the Internet during the study period. Our scale measures are specific to a domain and should not be as affected by external factors.

Conclusions

• Declines in primary care performance indicators were demonstrated by our study.
• These declines have been reported in an environment of change.
• Further research to examine the factors driving this decline in primary care quality is needed. The distractions of organizational restructuring, mergers, and departures from the market region, and pressures to increase productivity without compromising standards of care may be contributing factors.
• If quality of primary care performance continues to fall, the previously hoped for goals of health care reform through the advancement of primary care are at risk for being undermined.

Acknowledgments

Our research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, the executive director of the Massachusetts Group Insurance Commission, whose commitment to this research and participation in it made the study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study.

Related resources:

• HealthWatch http://healthwatch.medscape.com/medscape/p/gcommunity/ghome.asp
• HealthScout http://www.healthscout.com/cgi-bin/WebObjects/Af.woa
• WebMD http://www.my.webmd.com
• drkoop.com http://www.drkoop.com

The quality of physician-patient relationships alters health outcomes,^1-3 affects patients’ willingness to comply with medical advice or treatment,^4,5 and influences patients’ pursuit of malpractice suits.^6,7 Changes that reflect a decline in patients’ experience of structural and organizational aspects of care are important, because these areas are strong determinants of patient satisfaction.⁸

There is little question that health care delivery systems have undergone tremendous change over the past decade and a half. These changes have affected multiple aspects of medical practice, including financial incentives faced by individual clinicians, the organization of medical practices, and the corporate relationships among provider organizations. Primary care and its position within the health care delivery system has been a focal point for much of the change. Under most forms of insurance primary care physicians now hold a central role in patient care, responsible for coordinating and integrating all aspects of the care provided to their panel of patients and in some cases sharing in the financial risk associated with providing care under a capitated budget arrangement. As changes in the organization and financing of health care have unfolded (most notably over the past several years) they have almost always had direct implications for the primary care physician’s role in interacting with patients. During our study period of 1996 to 1999, these changes in the Commonwealth of Massachusetts have included the restructuring or merging of plans and their member practices, publicly reported financial difficulties, and the departure of plans from the market region. Even within stable plans, primary care physicians have experienced pressures to increase productivity, decrease costs, and attend to patient satisfaction.

We measured changes in patients’ experience of primary care with their primary care physicians over a 3-year study period in the Commonwealth of Massachusetts. We used indices of primary care quality and studied a panel of insured adults who provided detailed information about their care. Both the quality of their relationships with their primary care physicians and their experience with organizational features of care (access, continuity, integration) were monitored during the study period.

Methods

A sample of insured employees who responded to a mailed questionnaire at baseline (1996) and again 3 years later (1999) comprised the population of this longitudinal observational study. The participants belonged to 1 of 12 insurance plans. These were representative of the major health plans in the state. The questionnaire included the Primary Care Assessment Survey (PCAS),⁹ a validated patient-completed questionnaire that measures 7 essential characteristics of primary care, defined by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ All PCAS scales are measured in the context of a specific physician-patient relationship and reference the entirety of that relationship (ie, they are not visit specific).⁹ In these analyses, we examined changes in the 8 PCAS scales over a 3-year study period. The scales that we examined cover 2 broad aspects of the patient’s primary care experience: the quality of the primary care relationship (4 scales: quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust) and organizational features of care (4 scales: financial access, organizational access, visit-based continuity, integration of care). Table 1shows the item content of each scale.

Baseline data were obtained between January and April 1996. Using a 3-stage mail survey that included an initial mailing and 2 additional mailings to nonrespondents and limited telephone follow-up of randomly selected nonrespondents,¹¹ the PCAS was administered to a random sample of 10,733 Commonwealth of Massachusetts employees stratified by age, health plan, and ZIP code. Of the original sample, 221 were excluded as either unable to be located by mail (n=184), deceased (n=11), or no longer a Commonwealth of Massachusetts employee (n=26). In total, 6810 adults completed the baseline questionnaire by mail, and 394 completed it by telephone (response rate=68.5%).

Follow-up data were collected between January and April 1999. The follow-up questionnaire was administered to all baseline study participants who had identified a primary care physician and whose physician was listed in the Massachusetts Board of Registration in Medicine registry of licensed physicians (n=6075). Follow-up data collection employed a 3-step mail survey protocol as at baseline and was supplemented with final targeted mailings to 2 groups of nonrespondents (ethnic minorities, n=31, and those without a college diploma, n=521). The targeted mailings were performed, when nearing the conclusion of data collection, these subgroups were found to be under-represented among follow-up respondents. Completed questionnaires were received from 69.4% of the eligible respondents at follow-up (n=4108). Data collection and entry at baseline and follow-up were managed by the Center for Survey Research, University of Massachusetts (Boston).

In addition to the PCAS measures, the baseline and follow-up questionnaires were used to ascertain the respondents’ sociodemographic profiles (age, sex, race, years of education, household income) and health status. Measures of health status included the Medical Outcomes Study Short Form-12 (SF-12) Health Survey¹² and a checklist of 20 chronic conditions with high prevalence among US adults.¹³

Statistical Analyses

The principal analytic objective was to study the changes in primary care experiences of patients in a sustained primary care relationship during the 3-year study period. Patients who had changed physicians were excluded from the analytic sample. By restricting our analyses to patients who remained with the same physician we were able to isolate changes in their care over the 3-year study period without confounding factors associated with changing physicians. The analytic sample included patients who completed both the baseline and follow-up questionnaires, who identified a primary physician at baseline, and who remained with that physician throughout the 3-year study period (n=2383).

In a comparison of baseline characteristics of 1996 baseline respondents (n=6810) and those who were retained at the 3-year follow-up period (n=4108), the 1999 respondents were on average a year older (49 vs 48 years). They were more likely to be women (57% vs 55%) and more likely to be white (89% vs 86%). Education and income levels and measures of mental and physical health were almost equal in 1996 and 1999.

After exclusions, baseline demographics and health indexes of the analytic sample still remained near equal to those of the overall 1999 respondent group. Members of the analytic sample were more likely to be women (57% vs 55%) and to have a baseline primary care physician relationship duration of more than 5 years (54% vs 49.6%) when compared with the overall 1999 respondents. Compared with respondents, nonrespondents were younger, more likely to be men, poorer, less educated, and of a racial group other than white.

Baseline characteristics of physicians of 1999 respondents were also examined. The specialties of physicians were listed in the Massachusetts Board of Registration as 19% family physicians, 65% internists, 3% obstetrician-gynecologists, 4% cardiologists, and another 9% spread across other specialties. The physicians of patients in our analytic sample had the same specialty group distribution as the overall sample. In comparison with the overall group of 1999 respondents’ physicians, the physicians of the analytic sample were more likely to be men (80% vs 75%) and earlier graduates. Differences are due in part to patients nominating residents in training programs as their primary care physicians (6% of the overall respondent sample, 0% of the analytic sample). These patients subsequently changed their primary care physicians and were not included in the analytic sample. Also, there is an under-representation of women in the older group of physicians who have stable long-term relationships with their patients.

For each patient, we computed the difference between the 1996 and 1999 scores on each of the 8 PCAS scales (quality of communication, interpersonal treatment, physician’s knowledge of the patient, patient trust, financial access, organizational access, visit-based continuity, integration of care). We determined the unadjusted mean change in each scale and the 95% confidence interval around this change. To permit comparison across scales, we computed a standardized difference score (the effect size), which was the mean change of scales divided by the standard deviation of the referent scale scores at baseline.

Because previous empiric medical literature,^14,15 other research,^16-22 and our own cross-sectional data from baseline suggested that physician-patient relationship quality improves with increased relationship duration, we also examined the changes in indexes of relationship quality, controlling for the increased relationship length that occurred during our study for the group of patients who did not switch physicians. The adjusted changes in the relationship quality scales of communication, interpersonal treatment, physician knowledge of the patient, and trust were calculated using regression models. We used the longitudinal data in a stacked data set in which each patient’s pair of observation sets (from 1996 and 1999) was entered and included a binary variable indicating the survey round (round 1=0 and round 2=1). The stacked data make possible regression of the scale scores for each of the 4 indicators of relationship quality against a measure of relationship length, which (inherent in the selection criteria for the analytic sample) increased from the first to the second round of the survey. Inclusion of the survey round indicator in the 4 relevant regressions allowed a regression coefficient to be obtained for each scale, which reflected the magnitude of average difference in scale scores over the study period, adjusting by increases in the length of physician-patient relationship. P values for these regressions were used to evaluate the significance of the findings.

Results

The sociodemographic characteristics and health status of the analytic sample are presented in Table 2. Study participants ranged in age from 20 to 88 years, with a mean of 50.2 years. The majority were women and white, with some college education. On average they began our study with 2.8 chronic conditions. Physical and mental health status (as measured by the SF-12) was consistent with those observed nationally in adults in this age group.²³

Table 3 shows the unadjusted PCAS scores at baseline and follow-up and provides the 3-year score differences for the analytic sample. Two scales showed significant improvements (physician’s knowledge of the patient and visit-based continuity). Four scales showed significant declines (communication quality, interpersonal treatment, patient trust, and organizational access). The standardized measure of change (effect size [ES]) reveals that the largest changes occurred in organization access (ES=0.165), interpersonal treatment (ES=0.115), and communication quality (ES=0.095). The effect sizes for the other scales ranged from 0.016 (integration of care) to 0.060 (visit-based continuity).

For the 4 indicators of relationship quality, the observed (or unadjusted) change and change adjusted for the length of the physician-patient relationship are depicted in the Figure 1. The adjusted change scores from 1996 to 1999 show significant declines in all 4 indicators of relationship quality, ranging from -1.72 (physician’s knowledge of the patient) to -3.28 (interpersonal treatment).

Discussion

This observational study of patients under the continuing care of a primary physician from 1996 to 1999 found significant declines in 3 of the 4 indicators of relationship quality between 1996 and 1999. The largest declines were observed in interpersonal treatment, followed by declines in the quality of communication and trust. The fourth measure of relationship quality—the physician’s whole-person knowledge of the patient—increased, but this increase could not be demonstrated when adjusting for increased relationship duration. The adjusted figures demonstrate the concept that if expected increases in relationship quality due to increased relationship length are controlled for (ie, taken into account), then even larger decreases in relationship quality are demonstrated Figure 1.

Primary care is predicated on sustained physician-patient relationships, as recently noted by the Institute of Medicine Committee on the Future of Primary Care.¹⁰ The importance of relationship quality in health care is underscored by a research literature that links it to important outcomes of care. The quality of the physician-patient relationship in primary care has been associated with outcomes that include patients’ compliance with medical advice,^4,5,24 clinical outcomes of care,^1,3 patients’ willingness to initiate malpractice suits,^6,7 and patients’ decisions to change physicians.^25-27 Interpersonal treatment is a correlate of patient satisfaction,^28,29 which is important to individual patient well-being and as a factor that results in patient disenrollment. Research literature establishes that effective communication builds trust, reduces patients’ emotional stress, facilitates the process of diagnosing medical conditions, affects medical management decisions, and creates positive health outcomes.^1,3,30-34 In this context the observed decline in some of the indicators of quality of primary care relationships across the 3-year study period is concerning. We do not know from our study whether the quality of relationships was already declining before 1996 and, more important, whether declines are continuing at this rate.

Previous analyses employing baseline PCAS indicators of relationship quality as predictors of outcomes of care in this study population²⁷ enable us to estimate with some caution the effects of the observed declines in relationship quality over time. On the basis of this previous evidence, had the observed declines in interpersonal treatment and communication not occurred, an estimated 5% of the rate of patients’ voluntary disenrollment from their physicians’ practices could have been avoided. The observed decline in interpersonal treatment could translate into a measurable decline in patients’ attempts to adhere to their physicians’ counseling about smoking cessation, reducing alcohol consumption, and increasing exercise.

Our study included 4 indicators of organizational/structure features of care, 2 of which were observed to change significantly during the study period. Visit-based continuity between patients and their primary physicians (the ability of patients to see their regular physician for routine care and appointments when sick) increased. Patients’ organizational access to care (which includes patients’ ability to reach their physician’s office by phone and to obtain timely appointments when sick) declined substantially—more so than any other scale in the study. Still, patients in this study sample were better able to see their own physician, and having done so they go on to report that the quality of the encounters is declining. Access to care is a defining feature of primary care^10,35 and an important correlate of patient satisfaction.^8,36 The observed decline in organizational access to care raises concerns about the quality of primary care.

Limitations

Our study population included employed, insured, and generally healthy adults and was not representative of more vulnerable groups. This effect is further accentuated because the nonrespondents are more likely to be from a more vulnerable population. Since the research literature suggests that minority status and low income have an adverse impact on physician-patient interaction,^37-39 it is likely our findings would have been demonstrated more strongly with the inclusion of data from this section of the population.

Our inclusion criteria create selection biases, which reduce generalizability for some scales. Visit-based continuity is most likely to be optimized in this group of patients who have named a regular physician and have stayed with them during the study period.

The observed changes in primary care performance, though statistically significant, are small. But they occur within a reported (ie, observed) range of scores that is approximately one third the size of the range of possible scores. The movement within this range represents a larger shift than the same shift in a more extensive range. In addition to allowing comparisons across scales, the standardized effect size also helps address this issue by representing the data as a proportion of the standard deviation. The usual Cohen classification of effect size is not as pertinent to these results, because it was not developed and described for population studies.⁴⁰ Thus, the changes we observed (1) reflect declines where increases would be expected, (2) reflect the shift of a population, and (3) may reflect an ongoing trend continuing beyond our study period.

We have viewed declines primarily as a result of a change in the patients’ experiences in contrast to a change in the patients. Patient attributes not considered may include: declines in levels of societal trust, raised patient expectations with increasing patient consumerism, and patient education influenced by the rapidly increasing patient access to information on the Internet during the study period. Our scale measures are specific to a domain and should not be as affected by external factors.

Conclusions

• Declines in primary care performance indicators were demonstrated by our study.
• These declines have been reported in an environment of change.
• Further research to examine the factors driving this decline in primary care quality is needed. The distractions of organizational restructuring, mergers, and departures from the market region, and pressures to increase productivity without compromising standards of care may be contributing factors.
• If quality of primary care performance continues to fall, the previously hoped for goals of health care reform through the advancement of primary care are at risk for being undermined.

Acknowledgments

Our research was supported by grant number R01 HS08841 from the Agency for Healthcare Research and Quality (formerly the Agency for Health Care Policy and Research) and by grant number 035321 from the Robert Wood Johnson Foundation. We are indebted to Dolores Mitchell, the executive director of the Massachusetts Group Insurance Commission, whose commitment to this research and participation in it made the study possible. We also gratefully acknowledge Brian Clarridge, PhD, and his colleagues at The Center for Survey Research, University of Massachusetts, for their technical expertise and commitment to excellence in obtaining the data for our study.

Related resources:

• HealthWatch http://healthwatch.medscape.com/medscape/p/gcommunity/ghome.asp
• HealthScout http://www.healthscout.com/cgi-bin/WebObjects/Af.woa
• WebMD http://www.my.webmd.com
• drkoop.com http://www.drkoop.com

Postpartum depression (PPD) is a serious, common, and treatable condition seen frequently in the primary care setting.^1-3 The effects can be devastating for the entire family. The couple’s relationship often suffers,⁴ and women afflicted with PPD are at high risk for recurrent depression.⁵ Children of depressed mothers have been reported to have impaired cognitive development⁶ and behavioral disturbances.^7,8 Despite the serious consequences and the availability of highly effective pharmacologic and nonpharmacologic therapies,^9-11 PPD often remains unrecognized and untreated.^12,13

Routine office-based screening and the initiation of office systems have been shown to increase recognition and treatment of common conditions with high rates of missed diagnostic and treatment opportunities.¹⁴ Despite the availability of specific validated tools,^{15 17} routine screening for postpartum depression is not common in the United States. Although several population-based studies of PPD screening are available from other countries,^18,19 most studies in the United States have been completed in university settings or among high-risk populations.^20,21 Little published information is available on the effectiveness of routine postpartum screening in a community’s health care practice.²²

In 1997-98, we undertook a 9-month study of routine screening for PPD using the Edinburgh Postnatal Depression Scale (EPDS)¹⁵ at the 6-week postpartum visit in all clinical departments providing postpartum care in the Olmsted Medical Center and the Mayo Clinic, both in Rochester, Minnesota. The EPDS¹⁵ is a self-report scale that has 10 items relating to symptoms of depression and was developed to counter the limitations of other well-established depression scales used to screen postpartum women.^15,17 The scale is brief, easy to use, and avoids interpreting such common postpartum changes as fatigue, poor appetite, and altered sleep patterns as evidence of depression.^15,23

We evaluated changes in the 1-year postdelivery rates of the diagnosis and treatment of PPD before and after the introduction of universal office-based screening with the EPDS. The information obtained should be useful to other communities in determining how to address postpartum depression identification and the potential value of routine screening for PPD.

Methods

The 180 subjects for our study were all women who participated in the routine EPDS screening project, were residents of Olmsted County, and had EPDS scores of 10 or higher (n=172) or scores lower than 10 and an indication of any suicidal ideation (n=8). Nine women with scores of 10 or higher or suicidal ideation refused the general medical records research authorization required by Minnesota statute and could not be included in our study. That left 171 subjects with abnormal EPDS screening results plus an equal number of optimally matched²⁴ women with scores less than 10 and no indication of suicidal ideation for a total of 342 women studied. The matching was based on the age of the mother (±1.5 years) and month of delivery (±2 months).

Olmsted County is a metropolitan statistical area with a population of approximately 106,000 of whom 92% were white non-Hispanic with socioeconomic and educational levels slightly above the average for white citizens in the United States. There are approximately 1750 deliveries annually of Olmsted County women within Olmsted County hospitals. All in-hospital births in Olmsted County (99.5% of all county births) occur at Olmsted Medical Center or Rochester Methodist Hospital. Postpartum care for county residents is delivered at the Olmsted Medical Center, the Mayo Clinic, and their satellite practices, allowing screening of virtually all (98%) postpartum women in Olmsted County using only 2 institutions.²⁵

The screening process as well as the demographic data and scores for the women screened have been described previously.²⁶ Each woman’s screening results were available to her clinician at the time of her 6-week postpartum visit. Women who did not schedule a visit by 6 weeks postpartum were sent the survey by mail, and the results were given to the clinician who supervised her delivery. As required by the institutional review board, we notified the clinician of any EPDS score of 12 or higher or any indication of suicidal ideation on the EPDS, whether completed at the clinic or by mail. All care of the women remained at the discretion of the individual clinician.

Data Collection

All Olmsted Medical Center and Mayo Clinic records of each subject were reviewed for the period of 1 year postpartum. Linking women to all sources of health care is possible because the Rochester Epidemiology Project maintains a database of all health care utilization of all Olmsted County residents.²⁷ The data we collected included any medical record documentation of the EPDS scores, evaluation for depression, referrals to psychiatry or psychology, and any psychiatric diagnoses made during the 1-year period. Documented treatment of depression with reassurance, social services support, counseling/therapy, medications, electroconvulsive therapy, partial or inpatient psychiatric hospitalization, or other modalities was also collected. We recorded remissions and recurrences of depression and suicide attempts. Other basic demographic information was also collected, including gravity, parity, and gestational age at delivery, as well as documented previous affective disorders and previous postpartum depression.

Data Analysis

We calculated simple descriptive statistics. Comparison of depression-related evaluations, treatments, and diagnoses for those with EPDS scores of 12 or higher, scores of 10 or 11, and scores lower than 10 with and without suicidal ideation were completed using Mantel-Haenzel chi-square testing and tests for trends. The number of diagnoses of depression for the entire population of the 909 subjects screened with the EPDS was estimated by applying the rate of diagnosed depression in the 171 women with EPDS scores lower than 10 to the other 558 women with scores of lower than 10. This estimate was based on the assumption that the 558 women with EPDS scores less than 10 whose medical records were not reviewed had similar rates of diagnosed depression as the women with EPDS scores less than 10 whose medical records were reviewed. This assumption appeared justified, since both groups had similar demographic characteristics and similar distributions of EPDS scores from 0 to 9. We compared the post-EPDS screening rates of PPD diagnosis with the prescreening rates obtained from a previous study of the same community²⁸ using the chi-squared statistic.

The institutional review boards of the Olmsted Medical Center and the Mayo Clinic approved our study design.

Results

The mean age at delivery of the 342 women (171 with normal EPDS scores and 171 women with elevated scores) whose medical records were reviewed was 29 years (range=16-46 years). On average this was the second pregnancy for these women, and most (94%) delivered at more than 36 weeks’ gestation. Ninety-two percent (315) of women made a postpartum visit, while 8% (27) did not and received the EPDS by mail. Eighty-two percent of the women saw a physician, and 18% saw a nurse practitioner or nurse midwife for the postpartum visit. The demographic data for the women in this study is similar to that for the entire group of 909 who completed the EPDS during the 9-month study.

Overall, 68 women were diagnosed with postpartum depression Figure 1. The rate of diagnosis of PPD varied by the EPDS score and was highest in women with scores of Ž12 compared with scores of 10 or 11 and <10 (P for trend=.01). When weighted for the whole population of women screened, the community rate of diagnosed PPD was estimated to be 10.7%.

Documentation of mental health evaluations and referrals was not universal and differed between those with normal and elevated EPDS scores Table 1. More than three fourths (77%) of the women with some level of suicidal ideation indicated on the EPDS had no documentation of further immediate evaluation or scheduled follow-up concerning the risk for suicide. This included 5 women whose EPDS scores indicated “sometimes” thinking about suicide and another 28 who “occasionally” thought about suicide.

In the 3 women with documented clinician concern regarding risk of self-harm, immediate action was also documented. All 3 of these women had indicated that they had experienced suicidal ideation during the previous week, according to their EPDS sheets. One of these women was admitted to an inpatient mental health unit for short-term evaluation and initiation of therapy. The others were started on outpatient medical therapy. Two suicide attempts were recorded in the medical records of the study cohort. One woman who expressed sometimes thinking of self-harm but had no documentation of further evaluation made a suicide attempt (by overdose of over-the-counter medications) approximately 1 month after her postpartum visit and EPDS screening. She was hospitalized in the intensive care unit (ICU) for medical stabilization and was later transferred to an inpatient mental health unit. Another suicide attempt in this cohort involved a woman with no thoughts of suicide reproted on the EPDS at 6 weeks postpartum.

Follow-up appointments to monitor confirmed or probable depression were suggested for 57 of the women, including 52 with EPDS scores of 10 or higher. In approximately a third of the cases (21, 37%) the follow-up appointment was with the same clinician. The other two thirds were scheduled to see a psychologist or psychiatrist. Follow-up visits were encouraged for 2.9% (5 of 171) of the women with EPDS scores lower than 10, for 23.5% (16 of 68) of the women with EPDS scores of 10 or 11, and for 45.3% (43 of 95) of the women with EPDS scores of 12 or higher (P for trend <.001).

Postpartum depression was diagnosed in 16 women at follow-up appointments initiated by the postpartum care provider. Altogether, 58 women were diagnosed with postpartum depression at visits clearly related to the 6-week postpartum visit. Most diagnoses of postpartum depression occurred within 90 days of delivery (65%).

An additional 46 subjects had later evaluation for postpartum depression which did not appear to be initiated by their postnatal care clinician. Only 10 of these women were given a diagnosis of depression. Sixteen of these women self-referred directly to a psychiatrist or psychologist, and the others were evaluated for depression during the course of a visit for another reason. The specialty of the other clinicians included family medicine (16), obstetrician/gynecologist or certified nurse-midwife (8), emergency department physician (2), and 1 each by a physiatrist, an endocrinologist, a nurse practitioner, and a physician’s assistant.

Treatment for women with diagnosed postpartum depression was universally documented. Antidepressant medications were prescribed for 49% of these women and counseling was given to 78%; many women received both (39%). In addition, one woman with a history of recurrent depression was started on an antidepressant immediately following delivery. She had no documented recurrence of depression in the postpartum period. None of the subjects in this study underwent electroconvulsive therapy during the first year postpartum. Three women were hospitalized for specific diagnoses of depression and 2 have been described previously. Another woman was hospitalized on a medical service at 4 months postpartum for fatigue, arthralgias, and other nonspecific symptoms that were eventually diagnosed as an unusual presentation of postpartum depression. Her EPDS score was 13 near 6 weeks postpartum, and she had a history of depression, including a pre-pregnancy attempted suicide.

Discussion

Routine screening for postpartum depression with the EPDS was associated with more-than-doubling the rate of physician-diagnosed postpartum depression in this community-based population. Many of the diagnoses of depression (85%) were made at a visit that could be directly linked to the 6-week postpartum visit during which the screening was completed. Depression-related care was offered in all women with the diagnosis of PPD. Consistent with other work,^15,17,18 women with an elevated EPDS score were 7 times more likely to be diagnosed with PPD. Although only an intermediate outcome measure, receiving treatment for PPD is the first step in effecting more patient-oriented outcomes, such as improved ability to carry out usual activities, ability to care for the new infant, and prevention of suicide.¹³

Most of the diagnoses of postpartum depression were made by the physician or midlevel practitioner who cared for the woman at her 6-week postpartum visit, and most were made within 3 months of delivery. These primary care physicians and obstetrical care providers both diagnosed the condition and provided care for many of the women. The importance of primary care physicians in the recognition and treatment of all types of depression has previously been confirmed.^13,14,29,30

The pattern of diagnosis early in the postpartum period is similar to that reported in other studies^2,15,12 with most women receiving the diagnosis within 6 months of delivery. During evaluation for their depression, many women with PPD reported that symptoms began within weeks of delivery and were simply tolerated until the diagnosis was made. Screening for depression at the 6-week postpartum visit is most likely to identify these women with early onset of symptoms.

EPDS screening is done at a single point in time, and not all postpartum depression is evident at or before this time. It is therefore important to continue to consider PPD as a diagnosis for women who have no signs or symptoms at the 6-week postpartum visit but present at a later time with findings that may be consistent with depression.¹⁷ In our study, it is impossible to determine whether the women ultimately diagnosed with PPD but had low EPDS scores near 6 weeks postpartum represent false-negative depression screens or whether these women were not symptomatic at the time of the EPDS screening.

The information documented in the medical records suggests that for some of the women with elevated EPDS scores, at the postpartum visits may have been missed opportunities to diagnose depression. Some women who had a first diagnosis of PPD at 3 to 9 months after delivery mentioned that symptoms had been present since the baby was aged younger than 1 month and had elevated EPDS screening scores. These women may represent the enhanced clarity of hindsight, the failure of the physician to address EPDS scores, the limited ability of the clinician to adequately evaluate depression,^5,31-33 or the failure of the women to disclose the severity of their symptoms.¹² The importance of reducing missed opportunities is exemplified by the woman with no documented response to a high EPDS score followed by a suicide attempt at approximately 3 months postpartum. The ICU record completed at the time of hospitalization for treatment of an attempted suicide by overdose states she had been symptomatic since shortly after the birth of the baby.

The lack of documented response to suicidal ideation indicated on the EPDS of several women is disturbing. It is not clear if the clinicians did not see the response, did not respond, or did not document their response (ie, unreported telephone follow-up). All clinicians received the same information about the program including written material and a presentation at a meeting of each department providing postnatal care. Each clinician was notified of any EPDS indication of thoughts of self-harm.

Other studies of psychiatric screening tools in primary care have found similar results. In their evaluation of the Primary Care Evaluation of Mental Disorders (PRIME-MD), Spitzer and colleagues³⁴ reported that although 80% of clinicians introduced to this diagnostic screening tool supported routine psychiatric screening in primary care settings, only 32% of patients given new diagnoses by screening had new management actions initiated or planned. Among 74 patients in their study with previously unrecognized major depression, 22% were scheduled for follow-up visits, 10% received antidepressant prescriptions, and 5% were referred to a mental health care provider.³⁴ Routine use of the EPDS at 6 weeks postpartum can help to diagnose depression, but it is clearly not a sufficient intervention by itself.

Antidepressant therapy was not universally documented for this group of women. This may reflect the available spectrum of treatment choices and patient and physician preferences noted in the medical literature.⁹ In addition, antidepressant therapy may be discouraged if women are breastfeeding.³⁵ We were unable to make this distinction in most of the women with depression; however, the issue of medication crossing into breast milk was raised in at least 5 medical records and on at least 2 occasions breastfeeding was listed as a reason not to use antidepressant therapy.

Limitations

Because we followed practice as it occurs, it is not possible to benchmark our results against those of clinical intervention trials in which all patients are assessed for the outcome. However, we can provide unique data on the changes in clinical practice following the institution of screening for all women at the 6-week postpartum visit. Women were considered to have PPD on the basis of diagnoses recorded in the medical record. These diagnoses reflect the physicians’ judgment and may not exactly reflect the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, diagnostic criteria for depression. However, it is the diagnoses that physicians and other clinicians make that are the basis for treatment provided to women. Therefore, this type of study offers important information regarding the clinical effectiveness of universal screening with the EPDS. When added to studies of the psychometric properties and the efficacy of the instrument, effectiveness data can help identify barriers that occur in the practice-based implementation of trial programs.

Olmsted County women represent a diversity of socioeconomic status with 22% of pregnancies being covered by Medicaid insurance. Although the screening tool has been validated in multiple racial groups,^17-19 racially diverse groups may respond differently to their physician’s discussion of signs and symptoms of depression. Therefore, our results may not be generalizable to all women in the United States. However, middle-class white women are often considered at low risk for psychosocial problems and may therefore fail to be evaluated for PPD, making this an important group in which to assess this mass screening program.

Conclusions

Universal screening for PPD using the EPDS can be successfully implemented in primary care practices and may be associated with a significant increase in the rate of recognition, diagnosis, and treatment of postpartum depression.

Related Resources

• WebMD
• National Institute of Mental Health (NIMH)
• National Mental Health Association (NMHA)
• Mental Health Online

Postpartum depression (PPD) is a serious, common, and treatable condition seen frequently in the primary care setting.^1-3 The effects can be devastating for the entire family. The couple’s relationship often suffers,⁴ and women afflicted with PPD are at high risk for recurrent depression.⁵ Children of depressed mothers have been reported to have impaired cognitive development⁶ and behavioral disturbances.^7,8 Despite the serious consequences and the availability of highly effective pharmacologic and nonpharmacologic therapies,^9-11 PPD often remains unrecognized and untreated.^12,13

Routine office-based screening and the initiation of office systems have been shown to increase recognition and treatment of common conditions with high rates of missed diagnostic and treatment opportunities.¹⁴ Despite the availability of specific validated tools,^{15 17} routine screening for postpartum depression is not common in the United States. Although several population-based studies of PPD screening are available from other countries,^18,19 most studies in the United States have been completed in university settings or among high-risk populations.^20,21 Little published information is available on the effectiveness of routine postpartum screening in a community’s health care practice.²²

In 1997-98, we undertook a 9-month study of routine screening for PPD using the Edinburgh Postnatal Depression Scale (EPDS)¹⁵ at the 6-week postpartum visit in all clinical departments providing postpartum care in the Olmsted Medical Center and the Mayo Clinic, both in Rochester, Minnesota. The EPDS¹⁵ is a self-report scale that has 10 items relating to symptoms of depression and was developed to counter the limitations of other well-established depression scales used to screen postpartum women.^15,17 The scale is brief, easy to use, and avoids interpreting such common postpartum changes as fatigue, poor appetite, and altered sleep patterns as evidence of depression.^15,23

We evaluated changes in the 1-year postdelivery rates of the diagnosis and treatment of PPD before and after the introduction of universal office-based screening with the EPDS. The information obtained should be useful to other communities in determining how to address postpartum depression identification and the potential value of routine screening for PPD.

Methods

The 180 subjects for our study were all women who participated in the routine EPDS screening project, were residents of Olmsted County, and had EPDS scores of 10 or higher (n=172) or scores lower than 10 and an indication of any suicidal ideation (n=8). Nine women with scores of 10 or higher or suicidal ideation refused the general medical records research authorization required by Minnesota statute and could not be included in our study. That left 171 subjects with abnormal EPDS screening results plus an equal number of optimally matched²⁴ women with scores less than 10 and no indication of suicidal ideation for a total of 342 women studied. The matching was based on the age of the mother (±1.5 years) and month of delivery (±2 months).

Olmsted County is a metropolitan statistical area with a population of approximately 106,000 of whom 92% were white non-Hispanic with socioeconomic and educational levels slightly above the average for white citizens in the United States. There are approximately 1750 deliveries annually of Olmsted County women within Olmsted County hospitals. All in-hospital births in Olmsted County (99.5% of all county births) occur at Olmsted Medical Center or Rochester Methodist Hospital. Postpartum care for county residents is delivered at the Olmsted Medical Center, the Mayo Clinic, and their satellite practices, allowing screening of virtually all (98%) postpartum women in Olmsted County using only 2 institutions.²⁵

The screening process as well as the demographic data and scores for the women screened have been described previously.²⁶ Each woman’s screening results were available to her clinician at the time of her 6-week postpartum visit. Women who did not schedule a visit by 6 weeks postpartum were sent the survey by mail, and the results were given to the clinician who supervised her delivery. As required by the institutional review board, we notified the clinician of any EPDS score of 12 or higher or any indication of suicidal ideation on the EPDS, whether completed at the clinic or by mail. All care of the women remained at the discretion of the individual clinician.

Data Collection

All Olmsted Medical Center and Mayo Clinic records of each subject were reviewed for the period of 1 year postpartum. Linking women to all sources of health care is possible because the Rochester Epidemiology Project maintains a database of all health care utilization of all Olmsted County residents.²⁷ The data we collected included any medical record documentation of the EPDS scores, evaluation for depression, referrals to psychiatry or psychology, and any psychiatric diagnoses made during the 1-year period. Documented treatment of depression with reassurance, social services support, counseling/therapy, medications, electroconvulsive therapy, partial or inpatient psychiatric hospitalization, or other modalities was also collected. We recorded remissions and recurrences of depression and suicide attempts. Other basic demographic information was also collected, including gravity, parity, and gestational age at delivery, as well as documented previous affective disorders and previous postpartum depression.

Data Analysis

We calculated simple descriptive statistics. Comparison of depression-related evaluations, treatments, and diagnoses for those with EPDS scores of 12 or higher, scores of 10 or 11, and scores lower than 10 with and without suicidal ideation were completed using Mantel-Haenzel chi-square testing and tests for trends. The number of diagnoses of depression for the entire population of the 909 subjects screened with the EPDS was estimated by applying the rate of diagnosed depression in the 171 women with EPDS scores lower than 10 to the other 558 women with scores of lower than 10. This estimate was based on the assumption that the 558 women with EPDS scores less than 10 whose medical records were not reviewed had similar rates of diagnosed depression as the women with EPDS scores less than 10 whose medical records were reviewed. This assumption appeared justified, since both groups had similar demographic characteristics and similar distributions of EPDS scores from 0 to 9. We compared the post-EPDS screening rates of PPD diagnosis with the prescreening rates obtained from a previous study of the same community²⁸ using the chi-squared statistic.

The institutional review boards of the Olmsted Medical Center and the Mayo Clinic approved our study design.

Results

The mean age at delivery of the 342 women (171 with normal EPDS scores and 171 women with elevated scores) whose medical records were reviewed was 29 years (range=16-46 years). On average this was the second pregnancy for these women, and most (94%) delivered at more than 36 weeks’ gestation. Ninety-two percent (315) of women made a postpartum visit, while 8% (27) did not and received the EPDS by mail. Eighty-two percent of the women saw a physician, and 18% saw a nurse practitioner or nurse midwife for the postpartum visit. The demographic data for the women in this study is similar to that for the entire group of 909 who completed the EPDS during the 9-month study.

Overall, 68 women were diagnosed with postpartum depression Figure 1. The rate of diagnosis of PPD varied by the EPDS score and was highest in women with scores of Ž12 compared with scores of 10 or 11 and <10 (P for trend=.01). When weighted for the whole population of women screened, the community rate of diagnosed PPD was estimated to be 10.7%.

Documentation of mental health evaluations and referrals was not universal and differed between those with normal and elevated EPDS scores Table 1. More than three fourths (77%) of the women with some level of suicidal ideation indicated on the EPDS had no documentation of further immediate evaluation or scheduled follow-up concerning the risk for suicide. This included 5 women whose EPDS scores indicated “sometimes” thinking about suicide and another 28 who “occasionally” thought about suicide.

In the 3 women with documented clinician concern regarding risk of self-harm, immediate action was also documented. All 3 of these women had indicated that they had experienced suicidal ideation during the previous week, according to their EPDS sheets. One of these women was admitted to an inpatient mental health unit for short-term evaluation and initiation of therapy. The others were started on outpatient medical therapy. Two suicide attempts were recorded in the medical records of the study cohort. One woman who expressed sometimes thinking of self-harm but had no documentation of further evaluation made a suicide attempt (by overdose of over-the-counter medications) approximately 1 month after her postpartum visit and EPDS screening. She was hospitalized in the intensive care unit (ICU) for medical stabilization and was later transferred to an inpatient mental health unit. Another suicide attempt in this cohort involved a woman with no thoughts of suicide reproted on the EPDS at 6 weeks postpartum.

Follow-up appointments to monitor confirmed or probable depression were suggested for 57 of the women, including 52 with EPDS scores of 10 or higher. In approximately a third of the cases (21, 37%) the follow-up appointment was with the same clinician. The other two thirds were scheduled to see a psychologist or psychiatrist. Follow-up visits were encouraged for 2.9% (5 of 171) of the women with EPDS scores lower than 10, for 23.5% (16 of 68) of the women with EPDS scores of 10 or 11, and for 45.3% (43 of 95) of the women with EPDS scores of 12 or higher (P for trend <.001).

Postpartum depression was diagnosed in 16 women at follow-up appointments initiated by the postpartum care provider. Altogether, 58 women were diagnosed with postpartum depression at visits clearly related to the 6-week postpartum visit. Most diagnoses of postpartum depression occurred within 90 days of delivery (65%).

An additional 46 subjects had later evaluation for postpartum depression which did not appear to be initiated by their postnatal care clinician. Only 10 of these women were given a diagnosis of depression. Sixteen of these women self-referred directly to a psychiatrist or psychologist, and the others were evaluated for depression during the course of a visit for another reason. The specialty of the other clinicians included family medicine (16), obstetrician/gynecologist or certified nurse-midwife (8), emergency department physician (2), and 1 each by a physiatrist, an endocrinologist, a nurse practitioner, and a physician’s assistant.

Treatment for women with diagnosed postpartum depression was universally documented. Antidepressant medications were prescribed for 49% of these women and counseling was given to 78%; many women received both (39%). In addition, one woman with a history of recurrent depression was started on an antidepressant immediately following delivery. She had no documented recurrence of depression in the postpartum period. None of the subjects in this study underwent electroconvulsive therapy during the first year postpartum. Three women were hospitalized for specific diagnoses of depression and 2 have been described previously. Another woman was hospitalized on a medical service at 4 months postpartum for fatigue, arthralgias, and other nonspecific symptoms that were eventually diagnosed as an unusual presentation of postpartum depression. Her EPDS score was 13 near 6 weeks postpartum, and she had a history of depression, including a pre-pregnancy attempted suicide.

Discussion

Routine screening for postpartum depression with the EPDS was associated with more-than-doubling the rate of physician-diagnosed postpartum depression in this community-based population. Many of the diagnoses of depression (85%) were made at a visit that could be directly linked to the 6-week postpartum visit during which the screening was completed. Depression-related care was offered in all women with the diagnosis of PPD. Consistent with other work,^15,17,18 women with an elevated EPDS score were 7 times more likely to be diagnosed with PPD. Although only an intermediate outcome measure, receiving treatment for PPD is the first step in effecting more patient-oriented outcomes, such as improved ability to carry out usual activities, ability to care for the new infant, and prevention of suicide.¹³

Most of the diagnoses of postpartum depression were made by the physician or midlevel practitioner who cared for the woman at her 6-week postpartum visit, and most were made within 3 months of delivery. These primary care physicians and obstetrical care providers both diagnosed the condition and provided care for many of the women. The importance of primary care physicians in the recognition and treatment of all types of depression has previously been confirmed.^13,14,29,30

The pattern of diagnosis early in the postpartum period is similar to that reported in other studies^2,15,12 with most women receiving the diagnosis within 6 months of delivery. During evaluation for their depression, many women with PPD reported that symptoms began within weeks of delivery and were simply tolerated until the diagnosis was made. Screening for depression at the 6-week postpartum visit is most likely to identify these women with early onset of symptoms.

EPDS screening is done at a single point in time, and not all postpartum depression is evident at or before this time. It is therefore important to continue to consider PPD as a diagnosis for women who have no signs or symptoms at the 6-week postpartum visit but present at a later time with findings that may be consistent with depression.¹⁷ In our study, it is impossible to determine whether the women ultimately diagnosed with PPD but had low EPDS scores near 6 weeks postpartum represent false-negative depression screens or whether these women were not symptomatic at the time of the EPDS screening.

The information documented in the medical records suggests that for some of the women with elevated EPDS scores, at the postpartum visits may have been missed opportunities to diagnose depression. Some women who had a first diagnosis of PPD at 3 to 9 months after delivery mentioned that symptoms had been present since the baby was aged younger than 1 month and had elevated EPDS screening scores. These women may represent the enhanced clarity of hindsight, the failure of the physician to address EPDS scores, the limited ability of the clinician to adequately evaluate depression,^5,31-33 or the failure of the women to disclose the severity of their symptoms.¹² The importance of reducing missed opportunities is exemplified by the woman with no documented response to a high EPDS score followed by a suicide attempt at approximately 3 months postpartum. The ICU record completed at the time of hospitalization for treatment of an attempted suicide by overdose states she had been symptomatic since shortly after the birth of the baby.

The lack of documented response to suicidal ideation indicated on the EPDS of several women is disturbing. It is not clear if the clinicians did not see the response, did not respond, or did not document their response (ie, unreported telephone follow-up). All clinicians received the same information about the program including written material and a presentation at a meeting of each department providing postnatal care. Each clinician was notified of any EPDS indication of thoughts of self-harm.

Other studies of psychiatric screening tools in primary care have found similar results. In their evaluation of the Primary Care Evaluation of Mental Disorders (PRIME-MD), Spitzer and colleagues³⁴ reported that although 80% of clinicians introduced to this diagnostic screening tool supported routine psychiatric screening in primary care settings, only 32% of patients given new diagnoses by screening had new management actions initiated or planned. Among 74 patients in their study with previously unrecognized major depression, 22% were scheduled for follow-up visits, 10% received antidepressant prescriptions, and 5% were referred to a mental health care provider.³⁴ Routine use of the EPDS at 6 weeks postpartum can help to diagnose depression, but it is clearly not a sufficient intervention by itself.

Antidepressant therapy was not universally documented for this group of women. This may reflect the available spectrum of treatment choices and patient and physician preferences noted in the medical literature.⁹ In addition, antidepressant therapy may be discouraged if women are breastfeeding.³⁵ We were unable to make this distinction in most of the women with depression; however, the issue of medication crossing into breast milk was raised in at least 5 medical records and on at least 2 occasions breastfeeding was listed as a reason not to use antidepressant therapy.

Limitations

Because we followed practice as it occurs, it is not possible to benchmark our results against those of clinical intervention trials in which all patients are assessed for the outcome. However, we can provide unique data on the changes in clinical practice following the institution of screening for all women at the 6-week postpartum visit. Women were considered to have PPD on the basis of diagnoses recorded in the medical record. These diagnoses reflect the physicians’ judgment and may not exactly reflect the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, diagnostic criteria for depression. However, it is the diagnoses that physicians and other clinicians make that are the basis for treatment provided to women. Therefore, this type of study offers important information regarding the clinical effectiveness of universal screening with the EPDS. When added to studies of the psychometric properties and the efficacy of the instrument, effectiveness data can help identify barriers that occur in the practice-based implementation of trial programs.

Olmsted County women represent a diversity of socioeconomic status with 22% of pregnancies being covered by Medicaid insurance. Although the screening tool has been validated in multiple racial groups,^17-19 racially diverse groups may respond differently to their physician’s discussion of signs and symptoms of depression. Therefore, our results may not be generalizable to all women in the United States. However, middle-class white women are often considered at low risk for psychosocial problems and may therefore fail to be evaluated for PPD, making this an important group in which to assess this mass screening program.

Conclusions

Universal screening for PPD using the EPDS can be successfully implemented in primary care practices and may be associated with a significant increase in the rate of recognition, diagnosis, and treatment of postpartum depression.

Related Resources

• WebMD
• National Institute of Mental Health (NIMH)
• National Mental Health Association (NMHA)
• Mental Health Online

Postpartum depression (PPD) is a serious, common, and treatable condition seen frequently in the primary care setting.^1-3 The effects can be devastating for the entire family. The couple’s relationship often suffers,⁴ and women afflicted with PPD are at high risk for recurrent depression.⁵ Children of depressed mothers have been reported to have impaired cognitive development⁶ and behavioral disturbances.^7,8 Despite the serious consequences and the availability of highly effective pharmacologic and nonpharmacologic therapies,^9-11 PPD often remains unrecognized and untreated.^12,13

Routine office-based screening and the initiation of office systems have been shown to increase recognition and treatment of common conditions with high rates of missed diagnostic and treatment opportunities.¹⁴ Despite the availability of specific validated tools,^{15 17} routine screening for postpartum depression is not common in the United States. Although several population-based studies of PPD screening are available from other countries,^18,19 most studies in the United States have been completed in university settings or among high-risk populations.^20,21 Little published information is available on the effectiveness of routine postpartum screening in a community’s health care practice.²²

In 1997-98, we undertook a 9-month study of routine screening for PPD using the Edinburgh Postnatal Depression Scale (EPDS)¹⁵ at the 6-week postpartum visit in all clinical departments providing postpartum care in the Olmsted Medical Center and the Mayo Clinic, both in Rochester, Minnesota. The EPDS¹⁵ is a self-report scale that has 10 items relating to symptoms of depression and was developed to counter the limitations of other well-established depression scales used to screen postpartum women.^15,17 The scale is brief, easy to use, and avoids interpreting such common postpartum changes as fatigue, poor appetite, and altered sleep patterns as evidence of depression.^15,23

We evaluated changes in the 1-year postdelivery rates of the diagnosis and treatment of PPD before and after the introduction of universal office-based screening with the EPDS. The information obtained should be useful to other communities in determining how to address postpartum depression identification and the potential value of routine screening for PPD.

Methods

The 180 subjects for our study were all women who participated in the routine EPDS screening project, were residents of Olmsted County, and had EPDS scores of 10 or higher (n=172) or scores lower than 10 and an indication of any suicidal ideation (n=8). Nine women with scores of 10 or higher or suicidal ideation refused the general medical records research authorization required by Minnesota statute and could not be included in our study. That left 171 subjects with abnormal EPDS screening results plus an equal number of optimally matched²⁴ women with scores less than 10 and no indication of suicidal ideation for a total of 342 women studied. The matching was based on the age of the mother (±1.5 years) and month of delivery (±2 months).

Olmsted County is a metropolitan statistical area with a population of approximately 106,000 of whom 92% were white non-Hispanic with socioeconomic and educational levels slightly above the average for white citizens in the United States. There are approximately 1750 deliveries annually of Olmsted County women within Olmsted County hospitals. All in-hospital births in Olmsted County (99.5% of all county births) occur at Olmsted Medical Center or Rochester Methodist Hospital. Postpartum care for county residents is delivered at the Olmsted Medical Center, the Mayo Clinic, and their satellite practices, allowing screening of virtually all (98%) postpartum women in Olmsted County using only 2 institutions.²⁵

The screening process as well as the demographic data and scores for the women screened have been described previously.²⁶ Each woman’s screening results were available to her clinician at the time of her 6-week postpartum visit. Women who did not schedule a visit by 6 weeks postpartum were sent the survey by mail, and the results were given to the clinician who supervised her delivery. As required by the institutional review board, we notified the clinician of any EPDS score of 12 or higher or any indication of suicidal ideation on the EPDS, whether completed at the clinic or by mail. All care of the women remained at the discretion of the individual clinician.

Data Collection

All Olmsted Medical Center and Mayo Clinic records of each subject were reviewed for the period of 1 year postpartum. Linking women to all sources of health care is possible because the Rochester Epidemiology Project maintains a database of all health care utilization of all Olmsted County residents.²⁷ The data we collected included any medical record documentation of the EPDS scores, evaluation for depression, referrals to psychiatry or psychology, and any psychiatric diagnoses made during the 1-year period. Documented treatment of depression with reassurance, social services support, counseling/therapy, medications, electroconvulsive therapy, partial or inpatient psychiatric hospitalization, or other modalities was also collected. We recorded remissions and recurrences of depression and suicide attempts. Other basic demographic information was also collected, including gravity, parity, and gestational age at delivery, as well as documented previous affective disorders and previous postpartum depression.

Data Analysis

We calculated simple descriptive statistics. Comparison of depression-related evaluations, treatments, and diagnoses for those with EPDS scores of 12 or higher, scores of 10 or 11, and scores lower than 10 with and without suicidal ideation were completed using Mantel-Haenzel chi-square testing and tests for trends. The number of diagnoses of depression for the entire population of the 909 subjects screened with the EPDS was estimated by applying the rate of diagnosed depression in the 171 women with EPDS scores lower than 10 to the other 558 women with scores of lower than 10. This estimate was based on the assumption that the 558 women with EPDS scores less than 10 whose medical records were not reviewed had similar rates of diagnosed depression as the women with EPDS scores less than 10 whose medical records were reviewed. This assumption appeared justified, since both groups had similar demographic characteristics and similar distributions of EPDS scores from 0 to 9. We compared the post-EPDS screening rates of PPD diagnosis with the prescreening rates obtained from a previous study of the same community²⁸ using the chi-squared statistic.

The institutional review boards of the Olmsted Medical Center and the Mayo Clinic approved our study design.

Results

The mean age at delivery of the 342 women (171 with normal EPDS scores and 171 women with elevated scores) whose medical records were reviewed was 29 years (range=16-46 years). On average this was the second pregnancy for these women, and most (94%) delivered at more than 36 weeks’ gestation. Ninety-two percent (315) of women made a postpartum visit, while 8% (27) did not and received the EPDS by mail. Eighty-two percent of the women saw a physician, and 18% saw a nurse practitioner or nurse midwife for the postpartum visit. The demographic data for the women in this study is similar to that for the entire group of 909 who completed the EPDS during the 9-month study.

Overall, 68 women were diagnosed with postpartum depression Figure 1. The rate of diagnosis of PPD varied by the EPDS score and was highest in women with scores of Ž12 compared with scores of 10 or 11 and <10 (P for trend=.01). When weighted for the whole population of women screened, the community rate of diagnosed PPD was estimated to be 10.7%.

Documentation of mental health evaluations and referrals was not universal and differed between those with normal and elevated EPDS scores Table 1. More than three fourths (77%) of the women with some level of suicidal ideation indicated on the EPDS had no documentation of further immediate evaluation or scheduled follow-up concerning the risk for suicide. This included 5 women whose EPDS scores indicated “sometimes” thinking about suicide and another 28 who “occasionally” thought about suicide.

In the 3 women with documented clinician concern regarding risk of self-harm, immediate action was also documented. All 3 of these women had indicated that they had experienced suicidal ideation during the previous week, according to their EPDS sheets. One of these women was admitted to an inpatient mental health unit for short-term evaluation and initiation of therapy. The others were started on outpatient medical therapy. Two suicide attempts were recorded in the medical records of the study cohort. One woman who expressed sometimes thinking of self-harm but had no documentation of further evaluation made a suicide attempt (by overdose of over-the-counter medications) approximately 1 month after her postpartum visit and EPDS screening. She was hospitalized in the intensive care unit (ICU) for medical stabilization and was later transferred to an inpatient mental health unit. Another suicide attempt in this cohort involved a woman with no thoughts of suicide reproted on the EPDS at 6 weeks postpartum.

Follow-up appointments to monitor confirmed or probable depression were suggested for 57 of the women, including 52 with EPDS scores of 10 or higher. In approximately a third of the cases (21, 37%) the follow-up appointment was with the same clinician. The other two thirds were scheduled to see a psychologist or psychiatrist. Follow-up visits were encouraged for 2.9% (5 of 171) of the women with EPDS scores lower than 10, for 23.5% (16 of 68) of the women with EPDS scores of 10 or 11, and for 45.3% (43 of 95) of the women with EPDS scores of 12 or higher (P for trend <.001).

Postpartum depression was diagnosed in 16 women at follow-up appointments initiated by the postpartum care provider. Altogether, 58 women were diagnosed with postpartum depression at visits clearly related to the 6-week postpartum visit. Most diagnoses of postpartum depression occurred within 90 days of delivery (65%).

An additional 46 subjects had later evaluation for postpartum depression which did not appear to be initiated by their postnatal care clinician. Only 10 of these women were given a diagnosis of depression. Sixteen of these women self-referred directly to a psychiatrist or psychologist, and the others were evaluated for depression during the course of a visit for another reason. The specialty of the other clinicians included family medicine (16), obstetrician/gynecologist or certified nurse-midwife (8), emergency department physician (2), and 1 each by a physiatrist, an endocrinologist, a nurse practitioner, and a physician’s assistant.

Treatment for women with diagnosed postpartum depression was universally documented. Antidepressant medications were prescribed for 49% of these women and counseling was given to 78%; many women received both (39%). In addition, one woman with a history of recurrent depression was started on an antidepressant immediately following delivery. She had no documented recurrence of depression in the postpartum period. None of the subjects in this study underwent electroconvulsive therapy during the first year postpartum. Three women were hospitalized for specific diagnoses of depression and 2 have been described previously. Another woman was hospitalized on a medical service at 4 months postpartum for fatigue, arthralgias, and other nonspecific symptoms that were eventually diagnosed as an unusual presentation of postpartum depression. Her EPDS score was 13 near 6 weeks postpartum, and she had a history of depression, including a pre-pregnancy attempted suicide.

Discussion

Routine screening for postpartum depression with the EPDS was associated with more-than-doubling the rate of physician-diagnosed postpartum depression in this community-based population. Many of the diagnoses of depression (85%) were made at a visit that could be directly linked to the 6-week postpartum visit during which the screening was completed. Depression-related care was offered in all women with the diagnosis of PPD. Consistent with other work,^15,17,18 women with an elevated EPDS score were 7 times more likely to be diagnosed with PPD. Although only an intermediate outcome measure, receiving treatment for PPD is the first step in effecting more patient-oriented outcomes, such as improved ability to carry out usual activities, ability to care for the new infant, and prevention of suicide.¹³

Most of the diagnoses of postpartum depression were made by the physician or midlevel practitioner who cared for the woman at her 6-week postpartum visit, and most were made within 3 months of delivery. These primary care physicians and obstetrical care providers both diagnosed the condition and provided care for many of the women. The importance of primary care physicians in the recognition and treatment of all types of depression has previously been confirmed.^13,14,29,30

The pattern of diagnosis early in the postpartum period is similar to that reported in other studies^2,15,12 with most women receiving the diagnosis within 6 months of delivery. During evaluation for their depression, many women with PPD reported that symptoms began within weeks of delivery and were simply tolerated until the diagnosis was made. Screening for depression at the 6-week postpartum visit is most likely to identify these women with early onset of symptoms.

EPDS screening is done at a single point in time, and not all postpartum depression is evident at or before this time. It is therefore important to continue to consider PPD as a diagnosis for women who have no signs or symptoms at the 6-week postpartum visit but present at a later time with findings that may be consistent with depression.¹⁷ In our study, it is impossible to determine whether the women ultimately diagnosed with PPD but had low EPDS scores near 6 weeks postpartum represent false-negative depression screens or whether these women were not symptomatic at the time of the EPDS screening.

The information documented in the medical records suggests that for some of the women with elevated EPDS scores, at the postpartum visits may have been missed opportunities to diagnose depression. Some women who had a first diagnosis of PPD at 3 to 9 months after delivery mentioned that symptoms had been present since the baby was aged younger than 1 month and had elevated EPDS screening scores. These women may represent the enhanced clarity of hindsight, the failure of the physician to address EPDS scores, the limited ability of the clinician to adequately evaluate depression,^5,31-33 or the failure of the women to disclose the severity of their symptoms.¹² The importance of reducing missed opportunities is exemplified by the woman with no documented response to a high EPDS score followed by a suicide attempt at approximately 3 months postpartum. The ICU record completed at the time of hospitalization for treatment of an attempted suicide by overdose states she had been symptomatic since shortly after the birth of the baby.

The lack of documented response to suicidal ideation indicated on the EPDS of several women is disturbing. It is not clear if the clinicians did not see the response, did not respond, or did not document their response (ie, unreported telephone follow-up). All clinicians received the same information about the program including written material and a presentation at a meeting of each department providing postnatal care. Each clinician was notified of any EPDS indication of thoughts of self-harm.

Other studies of psychiatric screening tools in primary care have found similar results. In their evaluation of the Primary Care Evaluation of Mental Disorders (PRIME-MD), Spitzer and colleagues³⁴ reported that although 80% of clinicians introduced to this diagnostic screening tool supported routine psychiatric screening in primary care settings, only 32% of patients given new diagnoses by screening had new management actions initiated or planned. Among 74 patients in their study with previously unrecognized major depression, 22% were scheduled for follow-up visits, 10% received antidepressant prescriptions, and 5% were referred to a mental health care provider.³⁴ Routine use of the EPDS at 6 weeks postpartum can help to diagnose depression, but it is clearly not a sufficient intervention by itself.

Antidepressant therapy was not universally documented for this group of women. This may reflect the available spectrum of treatment choices and patient and physician preferences noted in the medical literature.⁹ In addition, antidepressant therapy may be discouraged if women are breastfeeding.³⁵ We were unable to make this distinction in most of the women with depression; however, the issue of medication crossing into breast milk was raised in at least 5 medical records and on at least 2 occasions breastfeeding was listed as a reason not to use antidepressant therapy.

Limitations

Because we followed practice as it occurs, it is not possible to benchmark our results against those of clinical intervention trials in which all patients are assessed for the outcome. However, we can provide unique data on the changes in clinical practice following the institution of screening for all women at the 6-week postpartum visit. Women were considered to have PPD on the basis of diagnoses recorded in the medical record. These diagnoses reflect the physicians’ judgment and may not exactly reflect the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, diagnostic criteria for depression. However, it is the diagnoses that physicians and other clinicians make that are the basis for treatment provided to women. Therefore, this type of study offers important information regarding the clinical effectiveness of universal screening with the EPDS. When added to studies of the psychometric properties and the efficacy of the instrument, effectiveness data can help identify barriers that occur in the practice-based implementation of trial programs.

Olmsted County women represent a diversity of socioeconomic status with 22% of pregnancies being covered by Medicaid insurance. Although the screening tool has been validated in multiple racial groups,^17-19 racially diverse groups may respond differently to their physician’s discussion of signs and symptoms of depression. Therefore, our results may not be generalizable to all women in the United States. However, middle-class white women are often considered at low risk for psychosocial problems and may therefore fail to be evaluated for PPD, making this an important group in which to assess this mass screening program.

Conclusions

Universal screening for PPD using the EPDS can be successfully implemented in primary care practices and may be associated with a significant increase in the rate of recognition, diagnosis, and treatment of postpartum depression.

Related Resources

• WebMD
• National Institute of Mental Health (NIMH)
• National Mental Health Association (NMHA)
• Mental Health Online

The family as the unit of care has been a philosophical underpinning of family practice since its inception.^1-6 It is common for individual family members to see the same physician and for a family member to be present during a patient’s visit.^7-10 Flocke and colleagues¹¹ used direct observation to identify the frequency of care of a secondary patient. They defined a secondary patient as a family member other than the identified patient for an outpatient visit and found that care was provided to a second family member during 18% of the visits. Knishkowry and coworkers¹² used self-report by a group of Israeli family physicians to identify a 12% frequency of encounters where 1 or more visitors were simultaneously present. Although Flocke and colleagues focused on the assessment of the effects of these encounters on the primary patient and Knishkowry and coworkers described the number and characteristics of the visitors, neither set of authors described the actual content of care provided to the secondary patient. Our study was designed to describe the profile of services provided to secondary patients during visits to family physicians.

Methods

Sample Selection

Our study was conducted in the Ambulatory Sentinel Practice Network (ASPN), a network of 752 community-based primary care clinicians established in 1982 to conduct practice-based research.¹³ ASPN’s 122 practices in 34 states and 6 Canadian provinces have been shown to serve a patient population similar to the population of the United States.¹⁴ In addition, ASPN clinicians demonstrate practice patterns similar to those reported in the National Ambulatory Medical Survey,¹⁵ a national probability sample of visits to office-based physicians in the United States.

All ASPN clinicians were invited to participate in the study by a mailing that briefly described the study and its requirements. A total of 170 clinicians (23% of the total) from 50 member practices (41%) volunteered and completed the data collection.

Study Variables

A secondary patient was defined as another individual (a family member or friend of the primary patient who was either present or absent and was not scheduled for the visit) to whom the clinician offered a discernible service. The primary patient was the patient in the office who registered or signed in for the visit.

The clinician made the determination of whether a discernible service was provided to a secondary patient and reported the type of service using categories that included advice, providing a prescription, assessment or explanation of symptoms, follow-up on previous care, referral to another provider, general discussion of the secondary patient’s condition, and other. More than one service could be checked for a given visit. The categories were developed through input from ASPN clinicians at the network’s annual meeting and through subsequent discussion on the ASPN electronic mailing list.

Participating clinicians reported whether it was the primary patient’s first visit to the practice, who initiated the discussion about the secondary patient, whether the secondary patient was present, the estimated time required to discuss the secondary patient, whether the billing code reflected additional care, and an estimate of whether the care provided to the secondary patient could have substituted for a separate visit. The clinician also reported the age and sex of the primary and secondary patients and their relationship (spouse, parent, son or daughter, sibling, other relative, friend).

Data Collection and Analysis

The clinicians who agreed to participate in the study were sent protocol instructions and study materials. They were asked to complete data collection for each patient visit in which care was provided to a secondary patient during a 1-week period. They also reported the total number of all patient visits during the study period.

The data forms were sent to the ASPN central office where they were manually checked for completeness and key entered. The data reported by the clinicians were merged with information on the characteristics of the clinician (age, sex, years in practice) and practice characteristics (rural, urban, suburban) obtained from the ASPN member database.

Descriptive statistics are reported for primary patients, secondary patients, visits, and clinicians. We used chi-square tests for comparisons involving categorical variables and Student t tests to compare means for continuous variables. Significance was reported at P <.05.

Results

The 170 clinicians in 50 ASPN practices who participated in the study reported a total of 6957 patient visits during the 1-week reporting period. Ninety-five of the clinicians (56%) reported 1 or more instances of providing secondary care, yielding a total of 422 (6.1%) visits involving secondary care. Seventy-five clinicians reported no secondary care. The secondary encounter was most often initiated by the primary patient (55%) and least often by the clinician (15%). Secondary patents were present in the office 39% of the time and initiated the secondary care during 30% of the visits.

Clinicians estimated that the secondary care required an average of 4.9 minutes to deliver (range=1-60 minutes). They also reported that 64% of the secondary encounters were likely to have substituted for a separate office visit, while additional billing for the care was reported in only 5.2% of secondary encounters.

Categories of Service to the Secondary Patient

Advice was the discernable service provided in more than half the visits Table 1. Approximately 30% were accounted for equally by prescription, assessment or explanation of symptoms, and general discussion of condition. In addition, advice was also the most frequent service when the secondary patient encounter was judged by the clinician to have substituted for a separate office visit. In fact, advice was the most common in almost every secondary patient category, except secondary patients who were aged 65 years or older, where follow-up of a previous problem was the service category most likely to occur (data not shown).

Finally, certain services were more likely to be initiated by clinicians than patients Table 1. Follow-up and general discussion of a condition were associated with clinician-initiated secondary care, while advice, assessment or explanation of symptoms, and prescriptions were associated with patient-initiated secondary care.

Characteristics Associated with Secondary Care

There were few differences between clinicians who reported secondary care and those who did not. Physicians reporting secondary care were older (P <.05) and more likely to practice in a rural area (P <.05). Clinician sex and years in practice were not remarkable.

Table 2 shows the characteristics of primary and secondary patients. There were a greater percentage of women than men in the primary patient group (64%). The secondary patient was most often a spouse, parent, or child of the primary patient. Eighty-seven percent of the secondary patients were enrolled as patients in the practice.

Discussion

ASPN clinicians reported providing secondary care during approximately 6% of primary care visits and rarely billed for the service. Secondary care was provided primarily in the form of advice to another family member. An episode with a secondary patient was reported to take an average of 5 minutes and to substitute for a visit more than 60% of the time.

This is the first study to examine the content of care given to a secondary patient in community primary care practices. Although arranging a referral, dispensing a prescription (perhaps a renewal), or providing follow-up care might not be unexpected, ASPN clinicians reported more instances of the provision of more time-intensive and complex services, such as advice, assessment or explanation of symptoms, and general discussion of condition. The fact that clinicians reported this secondary care could substitute for an actual office visit 60% of the time further suggests some complexity of the service provided.

The observation that certain services were more likely associated with clinician—rather than patient-initiated—secondary care might relate to how comfortable a clinician was with a particular service. However, the strength of the association for follow-up of a previous episode of care supports the Institute of Medicine definition of primary care as continuous and accountable.¹⁶

A limitation of our study is the reliance on physician self-report, which might vary from the report of the patient or an objective observer. The lower frequency of secondary care than reported in the direct observational study by Flocke and colleagues¹¹ is likely due to the lower sensitivity of physician self-report versus direct observation of service delivery.^17-19

Although our study does not provide data to identify the reasons that secondary care was provided, it is interesting to speculate that access to care might be involved. Access issues related to the clinician or practice might include the ease of scheduling a visit or phone contact.²⁰ Access issues related to the patient might include transportation, work responsibilities, or child or elder care.²⁰ For example, when the secondary patient was aged 65 years or older, follow-up was the service more likely to occur. Perhaps this represents an accommodation to this age group, thereby possibly obviating the arrangement of transportation for another visit. Transportation, or the lack thereof, might explain why secondary care occurred more often in rural settings. The finding that secondary care tended to be provided more often by older clinicians might be explained by their more comprehensive knowledge of the patient and family. Although the proportion of women was higher for both primary and secondary patients, the finding that the secondary patient was less likely to be a woman is consistent with previous research demonstrating the central role of women in accessing medical care for the family.^21,22

Future research should examine the reasons why secondary care is provided, from the perspectives of the physician and primary and secondary patients. In addition, the effects of other factors on the frequency and content of secondary care, such as health insurance, employment, access to care, and family structure must be elucidated. Such studies would provide useful information on the extent to which secondary care is an expression of barriers to access of care or an added value of family practice responding rationally to competing opportunities.^23,24 Also, studies need to assess whether the quality of care including the clinical outcomes, patient satisfaction, and cost of care for both primary and secondary patients is affected by the substitution of secondary care for a separate visit.

Conclusions

A physician’s care of a secondary patient includes the provision of potentially time-consuming and complex services that are largely not reimbursed or recognized by current measures of primary care. This provision of secondary care seems to facilitate access to care and represents an added value provided by family physicians.

Related Resources

• American Academy of Family Physicians www.aafp.org
• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/crfppc

The family as the unit of care has been a philosophical underpinning of family practice since its inception.^1-6 It is common for individual family members to see the same physician and for a family member to be present during a patient’s visit.^7-10 Flocke and colleagues¹¹ used direct observation to identify the frequency of care of a secondary patient. They defined a secondary patient as a family member other than the identified patient for an outpatient visit and found that care was provided to a second family member during 18% of the visits. Knishkowry and coworkers¹² used self-report by a group of Israeli family physicians to identify a 12% frequency of encounters where 1 or more visitors were simultaneously present. Although Flocke and colleagues focused on the assessment of the effects of these encounters on the primary patient and Knishkowry and coworkers described the number and characteristics of the visitors, neither set of authors described the actual content of care provided to the secondary patient. Our study was designed to describe the profile of services provided to secondary patients during visits to family physicians.

Methods

Sample Selection

Our study was conducted in the Ambulatory Sentinel Practice Network (ASPN), a network of 752 community-based primary care clinicians established in 1982 to conduct practice-based research.¹³ ASPN’s 122 practices in 34 states and 6 Canadian provinces have been shown to serve a patient population similar to the population of the United States.¹⁴ In addition, ASPN clinicians demonstrate practice patterns similar to those reported in the National Ambulatory Medical Survey,¹⁵ a national probability sample of visits to office-based physicians in the United States.

All ASPN clinicians were invited to participate in the study by a mailing that briefly described the study and its requirements. A total of 170 clinicians (23% of the total) from 50 member practices (41%) volunteered and completed the data collection.

Study Variables

A secondary patient was defined as another individual (a family member or friend of the primary patient who was either present or absent and was not scheduled for the visit) to whom the clinician offered a discernible service. The primary patient was the patient in the office who registered or signed in for the visit.

The clinician made the determination of whether a discernible service was provided to a secondary patient and reported the type of service using categories that included advice, providing a prescription, assessment or explanation of symptoms, follow-up on previous care, referral to another provider, general discussion of the secondary patient’s condition, and other. More than one service could be checked for a given visit. The categories were developed through input from ASPN clinicians at the network’s annual meeting and through subsequent discussion on the ASPN electronic mailing list.

Participating clinicians reported whether it was the primary patient’s first visit to the practice, who initiated the discussion about the secondary patient, whether the secondary patient was present, the estimated time required to discuss the secondary patient, whether the billing code reflected additional care, and an estimate of whether the care provided to the secondary patient could have substituted for a separate visit. The clinician also reported the age and sex of the primary and secondary patients and their relationship (spouse, parent, son or daughter, sibling, other relative, friend).

Data Collection and Analysis

The clinicians who agreed to participate in the study were sent protocol instructions and study materials. They were asked to complete data collection for each patient visit in which care was provided to a secondary patient during a 1-week period. They also reported the total number of all patient visits during the study period.

The data forms were sent to the ASPN central office where they were manually checked for completeness and key entered. The data reported by the clinicians were merged with information on the characteristics of the clinician (age, sex, years in practice) and practice characteristics (rural, urban, suburban) obtained from the ASPN member database.

Descriptive statistics are reported for primary patients, secondary patients, visits, and clinicians. We used chi-square tests for comparisons involving categorical variables and Student t tests to compare means for continuous variables. Significance was reported at P <.05.

Results

The 170 clinicians in 50 ASPN practices who participated in the study reported a total of 6957 patient visits during the 1-week reporting period. Ninety-five of the clinicians (56%) reported 1 or more instances of providing secondary care, yielding a total of 422 (6.1%) visits involving secondary care. Seventy-five clinicians reported no secondary care. The secondary encounter was most often initiated by the primary patient (55%) and least often by the clinician (15%). Secondary patents were present in the office 39% of the time and initiated the secondary care during 30% of the visits.

Clinicians estimated that the secondary care required an average of 4.9 minutes to deliver (range=1-60 minutes). They also reported that 64% of the secondary encounters were likely to have substituted for a separate office visit, while additional billing for the care was reported in only 5.2% of secondary encounters.

Categories of Service to the Secondary Patient

Advice was the discernable service provided in more than half the visits Table 1. Approximately 30% were accounted for equally by prescription, assessment or explanation of symptoms, and general discussion of condition. In addition, advice was also the most frequent service when the secondary patient encounter was judged by the clinician to have substituted for a separate office visit. In fact, advice was the most common in almost every secondary patient category, except secondary patients who were aged 65 years or older, where follow-up of a previous problem was the service category most likely to occur (data not shown).

Finally, certain services were more likely to be initiated by clinicians than patients Table 1. Follow-up and general discussion of a condition were associated with clinician-initiated secondary care, while advice, assessment or explanation of symptoms, and prescriptions were associated with patient-initiated secondary care.

Characteristics Associated with Secondary Care

There were few differences between clinicians who reported secondary care and those who did not. Physicians reporting secondary care were older (P <.05) and more likely to practice in a rural area (P <.05). Clinician sex and years in practice were not remarkable.

Table 2 shows the characteristics of primary and secondary patients. There were a greater percentage of women than men in the primary patient group (64%). The secondary patient was most often a spouse, parent, or child of the primary patient. Eighty-seven percent of the secondary patients were enrolled as patients in the practice.

Discussion

ASPN clinicians reported providing secondary care during approximately 6% of primary care visits and rarely billed for the service. Secondary care was provided primarily in the form of advice to another family member. An episode with a secondary patient was reported to take an average of 5 minutes and to substitute for a visit more than 60% of the time.

This is the first study to examine the content of care given to a secondary patient in community primary care practices. Although arranging a referral, dispensing a prescription (perhaps a renewal), or providing follow-up care might not be unexpected, ASPN clinicians reported more instances of the provision of more time-intensive and complex services, such as advice, assessment or explanation of symptoms, and general discussion of condition. The fact that clinicians reported this secondary care could substitute for an actual office visit 60% of the time further suggests some complexity of the service provided.

The observation that certain services were more likely associated with clinician—rather than patient-initiated—secondary care might relate to how comfortable a clinician was with a particular service. However, the strength of the association for follow-up of a previous episode of care supports the Institute of Medicine definition of primary care as continuous and accountable.¹⁶

A limitation of our study is the reliance on physician self-report, which might vary from the report of the patient or an objective observer. The lower frequency of secondary care than reported in the direct observational study by Flocke and colleagues¹¹ is likely due to the lower sensitivity of physician self-report versus direct observation of service delivery.^17-19

Although our study does not provide data to identify the reasons that secondary care was provided, it is interesting to speculate that access to care might be involved. Access issues related to the clinician or practice might include the ease of scheduling a visit or phone contact.²⁰ Access issues related to the patient might include transportation, work responsibilities, or child or elder care.²⁰ For example, when the secondary patient was aged 65 years or older, follow-up was the service more likely to occur. Perhaps this represents an accommodation to this age group, thereby possibly obviating the arrangement of transportation for another visit. Transportation, or the lack thereof, might explain why secondary care occurred more often in rural settings. The finding that secondary care tended to be provided more often by older clinicians might be explained by their more comprehensive knowledge of the patient and family. Although the proportion of women was higher for both primary and secondary patients, the finding that the secondary patient was less likely to be a woman is consistent with previous research demonstrating the central role of women in accessing medical care for the family.^21,22

Future research should examine the reasons why secondary care is provided, from the perspectives of the physician and primary and secondary patients. In addition, the effects of other factors on the frequency and content of secondary care, such as health insurance, employment, access to care, and family structure must be elucidated. Such studies would provide useful information on the extent to which secondary care is an expression of barriers to access of care or an added value of family practice responding rationally to competing opportunities.^23,24 Also, studies need to assess whether the quality of care including the clinical outcomes, patient satisfaction, and cost of care for both primary and secondary patients is affected by the substitution of secondary care for a separate visit.

Conclusions

A physician’s care of a secondary patient includes the provision of potentially time-consuming and complex services that are largely not reimbursed or recognized by current measures of primary care. This provision of secondary care seems to facilitate access to care and represents an added value provided by family physicians.

Related Resources

• American Academy of Family Physicians www.aafp.org
• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/crfppc

The family as the unit of care has been a philosophical underpinning of family practice since its inception.^1-6 It is common for individual family members to see the same physician and for a family member to be present during a patient’s visit.^7-10 Flocke and colleagues¹¹ used direct observation to identify the frequency of care of a secondary patient. They defined a secondary patient as a family member other than the identified patient for an outpatient visit and found that care was provided to a second family member during 18% of the visits. Knishkowry and coworkers¹² used self-report by a group of Israeli family physicians to identify a 12% frequency of encounters where 1 or more visitors were simultaneously present. Although Flocke and colleagues focused on the assessment of the effects of these encounters on the primary patient and Knishkowry and coworkers described the number and characteristics of the visitors, neither set of authors described the actual content of care provided to the secondary patient. Our study was designed to describe the profile of services provided to secondary patients during visits to family physicians.

Methods

Sample Selection

Our study was conducted in the Ambulatory Sentinel Practice Network (ASPN), a network of 752 community-based primary care clinicians established in 1982 to conduct practice-based research.¹³ ASPN’s 122 practices in 34 states and 6 Canadian provinces have been shown to serve a patient population similar to the population of the United States.¹⁴ In addition, ASPN clinicians demonstrate practice patterns similar to those reported in the National Ambulatory Medical Survey,¹⁵ a national probability sample of visits to office-based physicians in the United States.

All ASPN clinicians were invited to participate in the study by a mailing that briefly described the study and its requirements. A total of 170 clinicians (23% of the total) from 50 member practices (41%) volunteered and completed the data collection.

Study Variables

A secondary patient was defined as another individual (a family member or friend of the primary patient who was either present or absent and was not scheduled for the visit) to whom the clinician offered a discernible service. The primary patient was the patient in the office who registered or signed in for the visit.

The clinician made the determination of whether a discernible service was provided to a secondary patient and reported the type of service using categories that included advice, providing a prescription, assessment or explanation of symptoms, follow-up on previous care, referral to another provider, general discussion of the secondary patient’s condition, and other. More than one service could be checked for a given visit. The categories were developed through input from ASPN clinicians at the network’s annual meeting and through subsequent discussion on the ASPN electronic mailing list.

Participating clinicians reported whether it was the primary patient’s first visit to the practice, who initiated the discussion about the secondary patient, whether the secondary patient was present, the estimated time required to discuss the secondary patient, whether the billing code reflected additional care, and an estimate of whether the care provided to the secondary patient could have substituted for a separate visit. The clinician also reported the age and sex of the primary and secondary patients and their relationship (spouse, parent, son or daughter, sibling, other relative, friend).

Data Collection and Analysis

The clinicians who agreed to participate in the study were sent protocol instructions and study materials. They were asked to complete data collection for each patient visit in which care was provided to a secondary patient during a 1-week period. They also reported the total number of all patient visits during the study period.

The data forms were sent to the ASPN central office where they were manually checked for completeness and key entered. The data reported by the clinicians were merged with information on the characteristics of the clinician (age, sex, years in practice) and practice characteristics (rural, urban, suburban) obtained from the ASPN member database.

Descriptive statistics are reported for primary patients, secondary patients, visits, and clinicians. We used chi-square tests for comparisons involving categorical variables and Student t tests to compare means for continuous variables. Significance was reported at P <.05.

Results

The 170 clinicians in 50 ASPN practices who participated in the study reported a total of 6957 patient visits during the 1-week reporting period. Ninety-five of the clinicians (56%) reported 1 or more instances of providing secondary care, yielding a total of 422 (6.1%) visits involving secondary care. Seventy-five clinicians reported no secondary care. The secondary encounter was most often initiated by the primary patient (55%) and least often by the clinician (15%). Secondary patents were present in the office 39% of the time and initiated the secondary care during 30% of the visits.

Clinicians estimated that the secondary care required an average of 4.9 minutes to deliver (range=1-60 minutes). They also reported that 64% of the secondary encounters were likely to have substituted for a separate office visit, while additional billing for the care was reported in only 5.2% of secondary encounters.

Categories of Service to the Secondary Patient

Advice was the discernable service provided in more than half the visits Table 1. Approximately 30% were accounted for equally by prescription, assessment or explanation of symptoms, and general discussion of condition. In addition, advice was also the most frequent service when the secondary patient encounter was judged by the clinician to have substituted for a separate office visit. In fact, advice was the most common in almost every secondary patient category, except secondary patients who were aged 65 years or older, where follow-up of a previous problem was the service category most likely to occur (data not shown).

Finally, certain services were more likely to be initiated by clinicians than patients Table 1. Follow-up and general discussion of a condition were associated with clinician-initiated secondary care, while advice, assessment or explanation of symptoms, and prescriptions were associated with patient-initiated secondary care.

Characteristics Associated with Secondary Care

There were few differences between clinicians who reported secondary care and those who did not. Physicians reporting secondary care were older (P <.05) and more likely to practice in a rural area (P <.05). Clinician sex and years in practice were not remarkable.

Table 2 shows the characteristics of primary and secondary patients. There were a greater percentage of women than men in the primary patient group (64%). The secondary patient was most often a spouse, parent, or child of the primary patient. Eighty-seven percent of the secondary patients were enrolled as patients in the practice.

Discussion

ASPN clinicians reported providing secondary care during approximately 6% of primary care visits and rarely billed for the service. Secondary care was provided primarily in the form of advice to another family member. An episode with a secondary patient was reported to take an average of 5 minutes and to substitute for a visit more than 60% of the time.

This is the first study to examine the content of care given to a secondary patient in community primary care practices. Although arranging a referral, dispensing a prescription (perhaps a renewal), or providing follow-up care might not be unexpected, ASPN clinicians reported more instances of the provision of more time-intensive and complex services, such as advice, assessment or explanation of symptoms, and general discussion of condition. The fact that clinicians reported this secondary care could substitute for an actual office visit 60% of the time further suggests some complexity of the service provided.

The observation that certain services were more likely associated with clinician—rather than patient-initiated—secondary care might relate to how comfortable a clinician was with a particular service. However, the strength of the association for follow-up of a previous episode of care supports the Institute of Medicine definition of primary care as continuous and accountable.¹⁶

A limitation of our study is the reliance on physician self-report, which might vary from the report of the patient or an objective observer. The lower frequency of secondary care than reported in the direct observational study by Flocke and colleagues¹¹ is likely due to the lower sensitivity of physician self-report versus direct observation of service delivery.^17-19

Although our study does not provide data to identify the reasons that secondary care was provided, it is interesting to speculate that access to care might be involved. Access issues related to the clinician or practice might include the ease of scheduling a visit or phone contact.²⁰ Access issues related to the patient might include transportation, work responsibilities, or child or elder care.²⁰ For example, when the secondary patient was aged 65 years or older, follow-up was the service more likely to occur. Perhaps this represents an accommodation to this age group, thereby possibly obviating the arrangement of transportation for another visit. Transportation, or the lack thereof, might explain why secondary care occurred more often in rural settings. The finding that secondary care tended to be provided more often by older clinicians might be explained by their more comprehensive knowledge of the patient and family. Although the proportion of women was higher for both primary and secondary patients, the finding that the secondary patient was less likely to be a woman is consistent with previous research demonstrating the central role of women in accessing medical care for the family.^21,22

Future research should examine the reasons why secondary care is provided, from the perspectives of the physician and primary and secondary patients. In addition, the effects of other factors on the frequency and content of secondary care, such as health insurance, employment, access to care, and family structure must be elucidated. Such studies would provide useful information on the extent to which secondary care is an expression of barriers to access of care or an added value of family practice responding rationally to competing opportunities.^23,24 Also, studies need to assess whether the quality of care including the clinical outcomes, patient satisfaction, and cost of care for both primary and secondary patients is affected by the substitution of secondary care for a separate visit.

Conclusions

A physician’s care of a secondary patient includes the provision of potentially time-consuming and complex services that are largely not reimbursed or recognized by current measures of primary care. This provision of secondary care seems to facilitate access to care and represents an added value provided by family physicians.

Related Resources

• American Academy of Family Physicians www.aafp.org
• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/crfppc

It’s one thing to say that we have evidence that something works. It’s far more important to know how well it works. —David M. Eddy¹

In previous articles in this series on information mastery we outlined the importance of finding, evaluating, and implementing POEMs (Patient-Oriented Evidence that Matters) to maximize patient outcome at the point-of-care. Clinicians practicing as “information masters” will have the information they need when they need it, allowing them to offer their patients the best care.

In this article we take the concept of using POEMs one large and significant step further, and apply it not only to making decisions about individual patients, but also within the context of the entire community and population. Information mastery can improve the value of health delivery systems by increasing quality and controlling costs. By improving the value of health care, physicians should be able to provide universal and equitable health care access for all.

The problem of cost

Our collective complacency for 44 million uninsured is a national disgrace.²

The amount of money spent yearly in the United States for health care continues to rise at a rate faster than the rate of inflation. Whereas in 1960, when 5% of the gross national product was consumed by health care costs, this proportion has increased to 15% in the year 2000.³

Translating this number into actual dollars, the average family of 4 pays at least $10,000 per year in direct and indirect health care costs.⁴ Direct health care costs include insurance premiums and co-pays, and out-of-pocket expenses for medicines and devices. Additional, indirect, health care costs come in the form of higher costs of purchased goods as a result of the burden of paying for the health care of the workers who manufacture and sell the products.

More money is spent, per person, in the United States on health care than in any other country in the world. Approximately 50% more is spent on health care in the United States than is spent in Canada. The United Kingdom spends only about one-third of this amount of money.⁵ Despite this increased spending, average life expectancy is not substantially longer here than in other industrialized countries.^5,6

Socioeconomic status plays a larger role in the United States than health care spending in determining the length and quality of life.^7-12 In a recent study comparing survival rates for 15 “curable” cancers in Toronto, Ontario, and Detroit, Michigan, researchers found that socioeconomic status had no effect on survival for 12 of the 15 cancers occurring in the Canadians. However, patients who were in the lowest economic strata in Detroit had survival rates 40% lower than patients who had greater income did.⁹ Similar results have been found with heart disease,⁷ breast cancer,⁸ and HIV infection,¹¹ and for mortality rates in general across socioeconomic strata.¹³

It is a common assumption among many US lay persons and clinicians that the increase in mortality among the poor is due to an increase in high risk health behaviors, such as smoking, alcohol and drug abuse, obesity, and sedentary lifestyle. However, controlling for age, sex, race, urbanicity, education level, and health risk behaviors, people in the lowest-income group (family income $30,000 per year) have a mortality rate almost three-fold higher than those in the highest income group.¹⁴ This risk is especially high for low-income women, presumably because of inadequate prenatal care.

Despite an unemployment level that is at a 30-year low, more than 44 million people are uninsured, including 11 million children.¹⁵ The number of uninsured people grows at a rate of 100,000 people per month. These uninsured people are termed the “working poor”: persons who work in jobs with an income that makes them ineligible for public assistance programs but is insufficient to allow them to afford health care. These are the people who sell us our shirts, our shoes, our fast food, and those who cut our hair. The icon of the middle class—the shopping mall—is staffed largely by the uninsured.

We have a hard time “seeing” these people since they do not walk into our offices. Those who get sick either self-treat or overload our emergency departments. As a result, they become almost invisible to a health care industry in which, despite advances in community medicine, care begins at the time of an office visit.

And so, medical care in America has a seeming incongruity: Americans spend more money on health care than any other people in the world, yet 25% of them do not have adequate care. On the surface we seem to have a free and open system, unlike other countries in which health care is rationed. As we delve below the surface, however, we find that instead of rationing health care, we limit it to those who can afford it.

Can we open the doors to everyone?

We are in a tailspin: Individual patients drive up costs, which are passed on to other people, who try to recover their ‘fair share’ by overusing services when their turn comes around. —David M. Eddy⁴

The easiest course of action is to simply do nothing and allow US society to continue to devote more resources to health care. This choice, however, is likely not acceptable to that family of 4 that already devotes more than $10,000 per year for this care in direct and indirect costs.

In addition, it may not be financially feasible in the world economy. Managed care organizations pass on their costs to the companies, large and small, that ultimately pay for health care. Most clinicians and lay-persons are all too familiar with the problem of high business costs leading to many US businesses relocating their manufacturing plants in other countries where the costs are lower. One of the leading determinants of the costs of doing business in the United States is the cost of health care for the workers.

Historically, the costs of health care have generally risen at a rate of approximately 3% above the yearly rate of inflation. Eliminating many costs of health care services Table 1 which would be unrealistic—would produce a reduction in health care spending for about 5 years until the continued outpacing of inflation by health care costs would return us to the steady rise we currently are experiencing.

Another cost-sparing approach is to eliminate coverage for potentially beneficial health care services that are not essential. Patients would have the option of obtaining these services, but only if they choose to pay for them at full price. This approach takes away a major incentive that drives up medical costs; patients who pay insurance premiums often want to get their money’s worth, whether or not they need the care. Patients, not physicians, may therefore make decisions concerning whether they would like to pay for beneficial but not absolutely necessary services.

Rationing

I think it’s clear that future generations will marvel at our capacity to invent and document effective health services; let’s hope they will not marvel equally at our failure to deliver access to these services. —Mark Chassin¹

Deciding where the split occurs between necessary and beneficial is not as easy as it sounds. For example, if we had to choose between paying for mammograms for all women starting at age 50 years, or paying for bone marrow transplants for metastatic breast cancer, how would we decide? Would it be fair to ask a 50-year-old woman with metastatic breast cancer, her family, or her doctor? Of course not.

Instead, what would happen if we were able to ask the same 50-year-old woman with breast cancer when she was only 20 years old and cancer free? Which option would she have chosen at that time in her life: mammogram screening starting at age 50 or bone marrow transplant for metastatic cancer? Chances are good that she would have picked periodic mammography screening, since the likelihood of benefit would appear to her to be greater. More likely, though, a woman, her family, and her doctor would want both.

Faced with limited resources, paying for both and not making a choice leaves us in our present position: We don’t ration services in the United States, we ration people.

The R word—rationing—seems to induce the ire of most of us in health care. To many, rationing is defined as “denying necessary health care to persons who need it,” “not allowing people to receive expensive services,” or “interference by government or business entities in the practice of medicine.” Whatever the definition, explicit debate about methods of rationing health care is emotional and seems to focus on issues of a moral nature.

Yet clinicians already ration health care based on need. The patient with crushing substernal chest pain is given more time and effort than the hypochondriacal patient who comes in every month for a reassurance visit. Clinicians frequently make decisions about how to deliver health care based on a comparison of individual need—rationing in its purest form.

Understanding Rationing

This type of rationing is justifiable because it does not seem to violate the patient’s best interest—although patients might derive additional benefit from a few minutes of your time, this benefit would be small and not essential. When discussing rationing of services, one needs to make this crucial distinction between beneficial and necessary services, especially when resources are limited.¹⁶

Several other misunderstandings cloud the concept of rationing.¹⁷ The more-is-better fallacy stipulates that more care is synonymous with better care, and, since rationing limits care, it must be wrong. Research and common sense do not bear out this assumption. The common build-it-and-they-will-come approach to offering new health care services offers many examples of increased care without better outcomes.^18,19

The good-old-days fallacy occurs when we remember fondly those times when we did not have to face the endless frustrations of insurance forms, authorizations and peer-review forms. Unfortunately, getting paid in direct proportion to what services a clinician delivers also directly rewards unnecessary and even harmful interventions.

The Marcus Welby fallacy particularly applies to family physicians and is the most important one to correct. Named after the TV doctor who cared for only 1 patient per week, this fallacy refuses to let us acknowledge that (1) patients have a life outside of our offices, and (2) there are patients outside of our practice who are nonetheless affected by what goes on within our 4 walls.

All clinicians must recognize that always choosing to maximize care for individual patients places these patients, not only in conflict with society, but, ultimately, in conflict with themselves. For example, even though the incremental cost of an expensive versus inexpensive antibiotic for a respiratory infection seems minimal at the time, each of these decisions takes away money in the system that could be used by the same patients later in their life for truly life-threatening infections. In essence, beneficial yet unnecessary care mortgages the patient’s—and society’s—future.

The True Mission

If we fix overuse or misuse problems, we improve quality and reduce costs at the same time. Overuse is ubiquitous in American medicine.¹

Evidence-based medicine, and our derivation, information mastery, evolved as a way to make sense of the incredible amount of information available to practicing physicians so that they might improve their delivery of medical care. Lately the use of evidence-based/outcomes-based medicine techniques have been met with suspicion, especially because nonmedical professionals have embraced this approach.

The true goal of evidence-based medicine and information mastery is to provide effective and efficient care to patients via a health care system that allows all people to receive basic care. To meet this goal, this system has to be reconfigured so that existing resources are used in a way that is fair and equitable to all persons (and not just patients). Costs must be considered.

Improving quality and decreasing costs

The value of health care services can be improved either by improving quality or decreasing costs. This relationship can be conceptualized by the following equation:

Value = Quality Cost

If we decrease cost and compromise quality in the process, we gain nothing and may lose value. This is many clinicians’ greatest concern regarding cost-cutting efforts. If we can raise quality and decrease costs, however, we can significantly improve value.

Improving quality can be accomplished by reducing underuse, overuse, and misuse of medical care. Most current efforts to improve the quality of health care are focused on reducing underuse and are aimed at ways (practice guidelines, peer-review reports, and so forth) to get clinicians to do things they should be doing but are not. The problem, however, is that doing more is expensive and raises costs, thus reducing the amount of value gained as a result of the increase in quality. As a result, the gains are usually minimal and do nothing to lower the costs of health care or to make access universal.

However, it is estimated that we can safely eliminate almost 20% of the things we do in medicine and no one will be harmed as a result.20 The question is: which 20%? The best way to improve quality in the system is to address misuse and overuse of resources by focusing on using interventions that are less expensive, more effective, or both. Protecting our patients from overuse of services prevents them from being exposed to the risks of unnecessary interventions. This process starts by paying attention to what the evidence is telling us about our care. The concept of Patient-Oriented Evidence that Matters grew out of a need to identify information that tells us what treatments allow patients to live longer or better. As valid POEMs accumulate, practices must be changed according to this new and better information Table 2.

Using POEMs as our guide to which services to provide and which to leave out can eliminate waste in medicine, and in so doing, may result in a more fair distribution of resources. This will occur only if free market competition resulting from industry payers or legislation prevents the additional savings from becoming more profits for shareholders. A focus on POEMs will fit into any health care system concerned with proportioning limited resources. Many clinicians and laypersons in the United States connect the idea of rationing with the long waiting lists for health services in Canada and the United Kingdom. Even countries with universal health care access would benefit from eliminating useless or marginally helpful services.

Think Globally, Act Locally: What Each Clinician Should Do

Every dollar that is spent unnecessarily in the care of a healthy person potentially leads to further restrictions on reimbursement, further increases in health insurance premiums, loss of health insurance in borderline cases and, ultimately, fewer available resources for the care of the sicker patients who desperately need them. —Raymond J. Gibbons, MD²¹

Medicine has some very crucial decisions to make in the immediate future involving the allocation of resources, including the appropriate use of antibiotics, screening diabetics for microalbuminuria, screening for osteoporosis using bone densiometry, screening for prostate cancer using the prostate specific antigen test, and the use of routine obstetrical ultrasound. By continuing to provide services that do not improve patient outcomes, we add to the rising costs of health care, which results in fewer patients being able to receive the health care they need.

What can the individual clinician do? Each of us needs to learn about the benefits, harms, and costs of important interventions.²² We need to identify both the unnecessary and underused services and determine with patients if those services are worth the costs. More basic, applied, and practice-based research is needed to determine patient preferences about what information they want or need and how they would like to be included in the decision-making process. In addition, we must take responsibility for incorporating valid POEMs and guidelines into our everyday practice. Finally, we must accept that resources are limited and we can either continue to limit people who receive services, or limit the services themselves.

Practice behaviors this year have a direct effect on the health care budget for next year, both in a fee-for-service and capitated system. Excess spending this year results in fewer patients being insured next year and has a direct impact on how many people can afford coverage or how many individuals a specific company can afford to employ. The money saved by increasing the value of the services we provide (by limiting costs or increasing quality using valid POEMs as a guide to delivering these services) may not result in a direct decrease in the overall cost of health care. It will, however, reduce the yearly increase in health care spending occurring above and beyond the inflation rate. Figure

The role of family medicine

The only thing necessary for the triumph of evil is for good men to do nothing.—attributed to Edmund Burke

The survival of family medicine as an independent specialty is being challenged by competition from other providers and increased control by insurance organizations. These challenges are reflected in the lower number of medical students attracted to the specialty.

What does family medicine have to offer the students who have only the dictatorial dons of medicine or the young, brash hero-docs on “ER” as role models? By comparison, family medicine does not look challenging or sexy. Family medicine must be seen as cutting-edge and patient-centered. To achieve these goals, the specialty must embrace patient-oriented evidence that matters and balance the needs of each individual with the needs of the family and the entire community.

It’s one thing to say that we have evidence that something works. It’s far more important to know how well it works. —David M. Eddy¹

In previous articles in this series on information mastery we outlined the importance of finding, evaluating, and implementing POEMs (Patient-Oriented Evidence that Matters) to maximize patient outcome at the point-of-care. Clinicians practicing as “information masters” will have the information they need when they need it, allowing them to offer their patients the best care.

In this article we take the concept of using POEMs one large and significant step further, and apply it not only to making decisions about individual patients, but also within the context of the entire community and population. Information mastery can improve the value of health delivery systems by increasing quality and controlling costs. By improving the value of health care, physicians should be able to provide universal and equitable health care access for all.

The problem of cost

Our collective complacency for 44 million uninsured is a national disgrace.²

The amount of money spent yearly in the United States for health care continues to rise at a rate faster than the rate of inflation. Whereas in 1960, when 5% of the gross national product was consumed by health care costs, this proportion has increased to 15% in the year 2000.³

Translating this number into actual dollars, the average family of 4 pays at least $10,000 per year in direct and indirect health care costs.⁴ Direct health care costs include insurance premiums and co-pays, and out-of-pocket expenses for medicines and devices. Additional, indirect, health care costs come in the form of higher costs of purchased goods as a result of the burden of paying for the health care of the workers who manufacture and sell the products.

More money is spent, per person, in the United States on health care than in any other country in the world. Approximately 50% more is spent on health care in the United States than is spent in Canada. The United Kingdom spends only about one-third of this amount of money.⁵ Despite this increased spending, average life expectancy is not substantially longer here than in other industrialized countries.^5,6

Socioeconomic status plays a larger role in the United States than health care spending in determining the length and quality of life.^7-12 In a recent study comparing survival rates for 15 “curable” cancers in Toronto, Ontario, and Detroit, Michigan, researchers found that socioeconomic status had no effect on survival for 12 of the 15 cancers occurring in the Canadians. However, patients who were in the lowest economic strata in Detroit had survival rates 40% lower than patients who had greater income did.⁹ Similar results have been found with heart disease,⁷ breast cancer,⁸ and HIV infection,¹¹ and for mortality rates in general across socioeconomic strata.¹³

It is a common assumption among many US lay persons and clinicians that the increase in mortality among the poor is due to an increase in high risk health behaviors, such as smoking, alcohol and drug abuse, obesity, and sedentary lifestyle. However, controlling for age, sex, race, urbanicity, education level, and health risk behaviors, people in the lowest-income group (family income $30,000 per year) have a mortality rate almost three-fold higher than those in the highest income group.¹⁴ This risk is especially high for low-income women, presumably because of inadequate prenatal care.

Despite an unemployment level that is at a 30-year low, more than 44 million people are uninsured, including 11 million children.¹⁵ The number of uninsured people grows at a rate of 100,000 people per month. These uninsured people are termed the “working poor”: persons who work in jobs with an income that makes them ineligible for public assistance programs but is insufficient to allow them to afford health care. These are the people who sell us our shirts, our shoes, our fast food, and those who cut our hair. The icon of the middle class—the shopping mall—is staffed largely by the uninsured.

We have a hard time “seeing” these people since they do not walk into our offices. Those who get sick either self-treat or overload our emergency departments. As a result, they become almost invisible to a health care industry in which, despite advances in community medicine, care begins at the time of an office visit.

And so, medical care in America has a seeming incongruity: Americans spend more money on health care than any other people in the world, yet 25% of them do not have adequate care. On the surface we seem to have a free and open system, unlike other countries in which health care is rationed. As we delve below the surface, however, we find that instead of rationing health care, we limit it to those who can afford it.

Can we open the doors to everyone?

We are in a tailspin: Individual patients drive up costs, which are passed on to other people, who try to recover their ‘fair share’ by overusing services when their turn comes around. —David M. Eddy⁴

The easiest course of action is to simply do nothing and allow US society to continue to devote more resources to health care. This choice, however, is likely not acceptable to that family of 4 that already devotes more than $10,000 per year for this care in direct and indirect costs.

In addition, it may not be financially feasible in the world economy. Managed care organizations pass on their costs to the companies, large and small, that ultimately pay for health care. Most clinicians and lay-persons are all too familiar with the problem of high business costs leading to many US businesses relocating their manufacturing plants in other countries where the costs are lower. One of the leading determinants of the costs of doing business in the United States is the cost of health care for the workers.

Historically, the costs of health care have generally risen at a rate of approximately 3% above the yearly rate of inflation. Eliminating many costs of health care services Table 1 which would be unrealistic—would produce a reduction in health care spending for about 5 years until the continued outpacing of inflation by health care costs would return us to the steady rise we currently are experiencing.

Another cost-sparing approach is to eliminate coverage for potentially beneficial health care services that are not essential. Patients would have the option of obtaining these services, but only if they choose to pay for them at full price. This approach takes away a major incentive that drives up medical costs; patients who pay insurance premiums often want to get their money’s worth, whether or not they need the care. Patients, not physicians, may therefore make decisions concerning whether they would like to pay for beneficial but not absolutely necessary services.

Rationing

I think it’s clear that future generations will marvel at our capacity to invent and document effective health services; let’s hope they will not marvel equally at our failure to deliver access to these services. —Mark Chassin¹

Deciding where the split occurs between necessary and beneficial is not as easy as it sounds. For example, if we had to choose between paying for mammograms for all women starting at age 50 years, or paying for bone marrow transplants for metastatic breast cancer, how would we decide? Would it be fair to ask a 50-year-old woman with metastatic breast cancer, her family, or her doctor? Of course not.

Instead, what would happen if we were able to ask the same 50-year-old woman with breast cancer when she was only 20 years old and cancer free? Which option would she have chosen at that time in her life: mammogram screening starting at age 50 or bone marrow transplant for metastatic cancer? Chances are good that she would have picked periodic mammography screening, since the likelihood of benefit would appear to her to be greater. More likely, though, a woman, her family, and her doctor would want both.

Faced with limited resources, paying for both and not making a choice leaves us in our present position: We don’t ration services in the United States, we ration people.

The R word—rationing—seems to induce the ire of most of us in health care. To many, rationing is defined as “denying necessary health care to persons who need it,” “not allowing people to receive expensive services,” or “interference by government or business entities in the practice of medicine.” Whatever the definition, explicit debate about methods of rationing health care is emotional and seems to focus on issues of a moral nature.

Yet clinicians already ration health care based on need. The patient with crushing substernal chest pain is given more time and effort than the hypochondriacal patient who comes in every month for a reassurance visit. Clinicians frequently make decisions about how to deliver health care based on a comparison of individual need—rationing in its purest form.

Understanding Rationing

This type of rationing is justifiable because it does not seem to violate the patient’s best interest—although patients might derive additional benefit from a few minutes of your time, this benefit would be small and not essential. When discussing rationing of services, one needs to make this crucial distinction between beneficial and necessary services, especially when resources are limited.¹⁶

Several other misunderstandings cloud the concept of rationing.¹⁷ The more-is-better fallacy stipulates that more care is synonymous with better care, and, since rationing limits care, it must be wrong. Research and common sense do not bear out this assumption. The common build-it-and-they-will-come approach to offering new health care services offers many examples of increased care without better outcomes.^18,19

The good-old-days fallacy occurs when we remember fondly those times when we did not have to face the endless frustrations of insurance forms, authorizations and peer-review forms. Unfortunately, getting paid in direct proportion to what services a clinician delivers also directly rewards unnecessary and even harmful interventions.

The Marcus Welby fallacy particularly applies to family physicians and is the most important one to correct. Named after the TV doctor who cared for only 1 patient per week, this fallacy refuses to let us acknowledge that (1) patients have a life outside of our offices, and (2) there are patients outside of our practice who are nonetheless affected by what goes on within our 4 walls.

All clinicians must recognize that always choosing to maximize care for individual patients places these patients, not only in conflict with society, but, ultimately, in conflict with themselves. For example, even though the incremental cost of an expensive versus inexpensive antibiotic for a respiratory infection seems minimal at the time, each of these decisions takes away money in the system that could be used by the same patients later in their life for truly life-threatening infections. In essence, beneficial yet unnecessary care mortgages the patient’s—and society’s—future.

The True Mission

If we fix overuse or misuse problems, we improve quality and reduce costs at the same time. Overuse is ubiquitous in American medicine.¹

Evidence-based medicine, and our derivation, information mastery, evolved as a way to make sense of the incredible amount of information available to practicing physicians so that they might improve their delivery of medical care. Lately the use of evidence-based/outcomes-based medicine techniques have been met with suspicion, especially because nonmedical professionals have embraced this approach.

The true goal of evidence-based medicine and information mastery is to provide effective and efficient care to patients via a health care system that allows all people to receive basic care. To meet this goal, this system has to be reconfigured so that existing resources are used in a way that is fair and equitable to all persons (and not just patients). Costs must be considered.

Improving quality and decreasing costs

The value of health care services can be improved either by improving quality or decreasing costs. This relationship can be conceptualized by the following equation:

Value = Quality Cost

If we decrease cost and compromise quality in the process, we gain nothing and may lose value. This is many clinicians’ greatest concern regarding cost-cutting efforts. If we can raise quality and decrease costs, however, we can significantly improve value.

Improving quality can be accomplished by reducing underuse, overuse, and misuse of medical care. Most current efforts to improve the quality of health care are focused on reducing underuse and are aimed at ways (practice guidelines, peer-review reports, and so forth) to get clinicians to do things they should be doing but are not. The problem, however, is that doing more is expensive and raises costs, thus reducing the amount of value gained as a result of the increase in quality. As a result, the gains are usually minimal and do nothing to lower the costs of health care or to make access universal.

However, it is estimated that we can safely eliminate almost 20% of the things we do in medicine and no one will be harmed as a result.20 The question is: which 20%? The best way to improve quality in the system is to address misuse and overuse of resources by focusing on using interventions that are less expensive, more effective, or both. Protecting our patients from overuse of services prevents them from being exposed to the risks of unnecessary interventions. This process starts by paying attention to what the evidence is telling us about our care. The concept of Patient-Oriented Evidence that Matters grew out of a need to identify information that tells us what treatments allow patients to live longer or better. As valid POEMs accumulate, practices must be changed according to this new and better information Table 2.

Using POEMs as our guide to which services to provide and which to leave out can eliminate waste in medicine, and in so doing, may result in a more fair distribution of resources. This will occur only if free market competition resulting from industry payers or legislation prevents the additional savings from becoming more profits for shareholders. A focus on POEMs will fit into any health care system concerned with proportioning limited resources. Many clinicians and laypersons in the United States connect the idea of rationing with the long waiting lists for health services in Canada and the United Kingdom. Even countries with universal health care access would benefit from eliminating useless or marginally helpful services.

Think Globally, Act Locally: What Each Clinician Should Do

Every dollar that is spent unnecessarily in the care of a healthy person potentially leads to further restrictions on reimbursement, further increases in health insurance premiums, loss of health insurance in borderline cases and, ultimately, fewer available resources for the care of the sicker patients who desperately need them. —Raymond J. Gibbons, MD²¹

Medicine has some very crucial decisions to make in the immediate future involving the allocation of resources, including the appropriate use of antibiotics, screening diabetics for microalbuminuria, screening for osteoporosis using bone densiometry, screening for prostate cancer using the prostate specific antigen test, and the use of routine obstetrical ultrasound. By continuing to provide services that do not improve patient outcomes, we add to the rising costs of health care, which results in fewer patients being able to receive the health care they need.

What can the individual clinician do? Each of us needs to learn about the benefits, harms, and costs of important interventions.²² We need to identify both the unnecessary and underused services and determine with patients if those services are worth the costs. More basic, applied, and practice-based research is needed to determine patient preferences about what information they want or need and how they would like to be included in the decision-making process. In addition, we must take responsibility for incorporating valid POEMs and guidelines into our everyday practice. Finally, we must accept that resources are limited and we can either continue to limit people who receive services, or limit the services themselves.

Practice behaviors this year have a direct effect on the health care budget for next year, both in a fee-for-service and capitated system. Excess spending this year results in fewer patients being insured next year and has a direct impact on how many people can afford coverage or how many individuals a specific company can afford to employ. The money saved by increasing the value of the services we provide (by limiting costs or increasing quality using valid POEMs as a guide to delivering these services) may not result in a direct decrease in the overall cost of health care. It will, however, reduce the yearly increase in health care spending occurring above and beyond the inflation rate. Figure

The role of family medicine

The only thing necessary for the triumph of evil is for good men to do nothing.—attributed to Edmund Burke

The survival of family medicine as an independent specialty is being challenged by competition from other providers and increased control by insurance organizations. These challenges are reflected in the lower number of medical students attracted to the specialty.

What does family medicine have to offer the students who have only the dictatorial dons of medicine or the young, brash hero-docs on “ER” as role models? By comparison, family medicine does not look challenging or sexy. Family medicine must be seen as cutting-edge and patient-centered. To achieve these goals, the specialty must embrace patient-oriented evidence that matters and balance the needs of each individual with the needs of the family and the entire community.

It’s one thing to say that we have evidence that something works. It’s far more important to know how well it works. —David M. Eddy¹

In previous articles in this series on information mastery we outlined the importance of finding, evaluating, and implementing POEMs (Patient-Oriented Evidence that Matters) to maximize patient outcome at the point-of-care. Clinicians practicing as “information masters” will have the information they need when they need it, allowing them to offer their patients the best care.

In this article we take the concept of using POEMs one large and significant step further, and apply it not only to making decisions about individual patients, but also within the context of the entire community and population. Information mastery can improve the value of health delivery systems by increasing quality and controlling costs. By improving the value of health care, physicians should be able to provide universal and equitable health care access for all.

The problem of cost

Our collective complacency for 44 million uninsured is a national disgrace.²

The amount of money spent yearly in the United States for health care continues to rise at a rate faster than the rate of inflation. Whereas in 1960, when 5% of the gross national product was consumed by health care costs, this proportion has increased to 15% in the year 2000.³

Translating this number into actual dollars, the average family of 4 pays at least $10,000 per year in direct and indirect health care costs.⁴ Direct health care costs include insurance premiums and co-pays, and out-of-pocket expenses for medicines and devices. Additional, indirect, health care costs come in the form of higher costs of purchased goods as a result of the burden of paying for the health care of the workers who manufacture and sell the products.

More money is spent, per person, in the United States on health care than in any other country in the world. Approximately 50% more is spent on health care in the United States than is spent in Canada. The United Kingdom spends only about one-third of this amount of money.⁵ Despite this increased spending, average life expectancy is not substantially longer here than in other industrialized countries.^5,6

Socioeconomic status plays a larger role in the United States than health care spending in determining the length and quality of life.^7-12 In a recent study comparing survival rates for 15 “curable” cancers in Toronto, Ontario, and Detroit, Michigan, researchers found that socioeconomic status had no effect on survival for 12 of the 15 cancers occurring in the Canadians. However, patients who were in the lowest economic strata in Detroit had survival rates 40% lower than patients who had greater income did.⁹ Similar results have been found with heart disease,⁷ breast cancer,⁸ and HIV infection,¹¹ and for mortality rates in general across socioeconomic strata.¹³

It is a common assumption among many US lay persons and clinicians that the increase in mortality among the poor is due to an increase in high risk health behaviors, such as smoking, alcohol and drug abuse, obesity, and sedentary lifestyle. However, controlling for age, sex, race, urbanicity, education level, and health risk behaviors, people in the lowest-income group (family income $30,000 per year) have a mortality rate almost three-fold higher than those in the highest income group.¹⁴ This risk is especially high for low-income women, presumably because of inadequate prenatal care.

Despite an unemployment level that is at a 30-year low, more than 44 million people are uninsured, including 11 million children.¹⁵ The number of uninsured people grows at a rate of 100,000 people per month. These uninsured people are termed the “working poor”: persons who work in jobs with an income that makes them ineligible for public assistance programs but is insufficient to allow them to afford health care. These are the people who sell us our shirts, our shoes, our fast food, and those who cut our hair. The icon of the middle class—the shopping mall—is staffed largely by the uninsured.

We have a hard time “seeing” these people since they do not walk into our offices. Those who get sick either self-treat or overload our emergency departments. As a result, they become almost invisible to a health care industry in which, despite advances in community medicine, care begins at the time of an office visit.

And so, medical care in America has a seeming incongruity: Americans spend more money on health care than any other people in the world, yet 25% of them do not have adequate care. On the surface we seem to have a free and open system, unlike other countries in which health care is rationed. As we delve below the surface, however, we find that instead of rationing health care, we limit it to those who can afford it.

Can we open the doors to everyone?

We are in a tailspin: Individual patients drive up costs, which are passed on to other people, who try to recover their ‘fair share’ by overusing services when their turn comes around. —David M. Eddy⁴

The easiest course of action is to simply do nothing and allow US society to continue to devote more resources to health care. This choice, however, is likely not acceptable to that family of 4 that already devotes more than $10,000 per year for this care in direct and indirect costs.

In addition, it may not be financially feasible in the world economy. Managed care organizations pass on their costs to the companies, large and small, that ultimately pay for health care. Most clinicians and lay-persons are all too familiar with the problem of high business costs leading to many US businesses relocating their manufacturing plants in other countries where the costs are lower. One of the leading determinants of the costs of doing business in the United States is the cost of health care for the workers.

Historically, the costs of health care have generally risen at a rate of approximately 3% above the yearly rate of inflation. Eliminating many costs of health care services Table 1 which would be unrealistic—would produce a reduction in health care spending for about 5 years until the continued outpacing of inflation by health care costs would return us to the steady rise we currently are experiencing.

Another cost-sparing approach is to eliminate coverage for potentially beneficial health care services that are not essential. Patients would have the option of obtaining these services, but only if they choose to pay for them at full price. This approach takes away a major incentive that drives up medical costs; patients who pay insurance premiums often want to get their money’s worth, whether or not they need the care. Patients, not physicians, may therefore make decisions concerning whether they would like to pay for beneficial but not absolutely necessary services.

Rationing

I think it’s clear that future generations will marvel at our capacity to invent and document effective health services; let’s hope they will not marvel equally at our failure to deliver access to these services. —Mark Chassin¹

Deciding where the split occurs between necessary and beneficial is not as easy as it sounds. For example, if we had to choose between paying for mammograms for all women starting at age 50 years, or paying for bone marrow transplants for metastatic breast cancer, how would we decide? Would it be fair to ask a 50-year-old woman with metastatic breast cancer, her family, or her doctor? Of course not.

Instead, what would happen if we were able to ask the same 50-year-old woman with breast cancer when she was only 20 years old and cancer free? Which option would she have chosen at that time in her life: mammogram screening starting at age 50 or bone marrow transplant for metastatic cancer? Chances are good that she would have picked periodic mammography screening, since the likelihood of benefit would appear to her to be greater. More likely, though, a woman, her family, and her doctor would want both.

Faced with limited resources, paying for both and not making a choice leaves us in our present position: We don’t ration services in the United States, we ration people.

The R word—rationing—seems to induce the ire of most of us in health care. To many, rationing is defined as “denying necessary health care to persons who need it,” “not allowing people to receive expensive services,” or “interference by government or business entities in the practice of medicine.” Whatever the definition, explicit debate about methods of rationing health care is emotional and seems to focus on issues of a moral nature.

Yet clinicians already ration health care based on need. The patient with crushing substernal chest pain is given more time and effort than the hypochondriacal patient who comes in every month for a reassurance visit. Clinicians frequently make decisions about how to deliver health care based on a comparison of individual need—rationing in its purest form.

Understanding Rationing

This type of rationing is justifiable because it does not seem to violate the patient’s best interest—although patients might derive additional benefit from a few minutes of your time, this benefit would be small and not essential. When discussing rationing of services, one needs to make this crucial distinction between beneficial and necessary services, especially when resources are limited.¹⁶

Several other misunderstandings cloud the concept of rationing.¹⁷ The more-is-better fallacy stipulates that more care is synonymous with better care, and, since rationing limits care, it must be wrong. Research and common sense do not bear out this assumption. The common build-it-and-they-will-come approach to offering new health care services offers many examples of increased care without better outcomes.^18,19

The good-old-days fallacy occurs when we remember fondly those times when we did not have to face the endless frustrations of insurance forms, authorizations and peer-review forms. Unfortunately, getting paid in direct proportion to what services a clinician delivers also directly rewards unnecessary and even harmful interventions.

The Marcus Welby fallacy particularly applies to family physicians and is the most important one to correct. Named after the TV doctor who cared for only 1 patient per week, this fallacy refuses to let us acknowledge that (1) patients have a life outside of our offices, and (2) there are patients outside of our practice who are nonetheless affected by what goes on within our 4 walls.

All clinicians must recognize that always choosing to maximize care for individual patients places these patients, not only in conflict with society, but, ultimately, in conflict with themselves. For example, even though the incremental cost of an expensive versus inexpensive antibiotic for a respiratory infection seems minimal at the time, each of these decisions takes away money in the system that could be used by the same patients later in their life for truly life-threatening infections. In essence, beneficial yet unnecessary care mortgages the patient’s—and society’s—future.

The True Mission

If we fix overuse or misuse problems, we improve quality and reduce costs at the same time. Overuse is ubiquitous in American medicine.¹

Evidence-based medicine, and our derivation, information mastery, evolved as a way to make sense of the incredible amount of information available to practicing physicians so that they might improve their delivery of medical care. Lately the use of evidence-based/outcomes-based medicine techniques have been met with suspicion, especially because nonmedical professionals have embraced this approach.

The true goal of evidence-based medicine and information mastery is to provide effective and efficient care to patients via a health care system that allows all people to receive basic care. To meet this goal, this system has to be reconfigured so that existing resources are used in a way that is fair and equitable to all persons (and not just patients). Costs must be considered.

Improving quality and decreasing costs

The value of health care services can be improved either by improving quality or decreasing costs. This relationship can be conceptualized by the following equation:

Value = Quality Cost

If we decrease cost and compromise quality in the process, we gain nothing and may lose value. This is many clinicians’ greatest concern regarding cost-cutting efforts. If we can raise quality and decrease costs, however, we can significantly improve value.

Improving quality can be accomplished by reducing underuse, overuse, and misuse of medical care. Most current efforts to improve the quality of health care are focused on reducing underuse and are aimed at ways (practice guidelines, peer-review reports, and so forth) to get clinicians to do things they should be doing but are not. The problem, however, is that doing more is expensive and raises costs, thus reducing the amount of value gained as a result of the increase in quality. As a result, the gains are usually minimal and do nothing to lower the costs of health care or to make access universal.

However, it is estimated that we can safely eliminate almost 20% of the things we do in medicine and no one will be harmed as a result.20 The question is: which 20%? The best way to improve quality in the system is to address misuse and overuse of resources by focusing on using interventions that are less expensive, more effective, or both. Protecting our patients from overuse of services prevents them from being exposed to the risks of unnecessary interventions. This process starts by paying attention to what the evidence is telling us about our care. The concept of Patient-Oriented Evidence that Matters grew out of a need to identify information that tells us what treatments allow patients to live longer or better. As valid POEMs accumulate, practices must be changed according to this new and better information Table 2.

Using POEMs as our guide to which services to provide and which to leave out can eliminate waste in medicine, and in so doing, may result in a more fair distribution of resources. This will occur only if free market competition resulting from industry payers or legislation prevents the additional savings from becoming more profits for shareholders. A focus on POEMs will fit into any health care system concerned with proportioning limited resources. Many clinicians and laypersons in the United States connect the idea of rationing with the long waiting lists for health services in Canada and the United Kingdom. Even countries with universal health care access would benefit from eliminating useless or marginally helpful services.

Think Globally, Act Locally: What Each Clinician Should Do

Every dollar that is spent unnecessarily in the care of a healthy person potentially leads to further restrictions on reimbursement, further increases in health insurance premiums, loss of health insurance in borderline cases and, ultimately, fewer available resources for the care of the sicker patients who desperately need them. —Raymond J. Gibbons, MD²¹

Medicine has some very crucial decisions to make in the immediate future involving the allocation of resources, including the appropriate use of antibiotics, screening diabetics for microalbuminuria, screening for osteoporosis using bone densiometry, screening for prostate cancer using the prostate specific antigen test, and the use of routine obstetrical ultrasound. By continuing to provide services that do not improve patient outcomes, we add to the rising costs of health care, which results in fewer patients being able to receive the health care they need.

What can the individual clinician do? Each of us needs to learn about the benefits, harms, and costs of important interventions.²² We need to identify both the unnecessary and underused services and determine with patients if those services are worth the costs. More basic, applied, and practice-based research is needed to determine patient preferences about what information they want or need and how they would like to be included in the decision-making process. In addition, we must take responsibility for incorporating valid POEMs and guidelines into our everyday practice. Finally, we must accept that resources are limited and we can either continue to limit people who receive services, or limit the services themselves.

Practice behaviors this year have a direct effect on the health care budget for next year, both in a fee-for-service and capitated system. Excess spending this year results in fewer patients being insured next year and has a direct impact on how many people can afford coverage or how many individuals a specific company can afford to employ. The money saved by increasing the value of the services we provide (by limiting costs or increasing quality using valid POEMs as a guide to delivering these services) may not result in a direct decrease in the overall cost of health care. It will, however, reduce the yearly increase in health care spending occurring above and beyond the inflation rate. Figure

The role of family medicine

The only thing necessary for the triumph of evil is for good men to do nothing.—attributed to Edmund Burke

The survival of family medicine as an independent specialty is being challenged by competition from other providers and increased control by insurance organizations. These challenges are reflected in the lower number of medical students attracted to the specialty.

What does family medicine have to offer the students who have only the dictatorial dons of medicine or the young, brash hero-docs on “ER” as role models? By comparison, family medicine does not look challenging or sexy. Family medicine must be seen as cutting-edge and patient-centered. To achieve these goals, the specialty must embrace patient-oriented evidence that matters and balance the needs of each individual with the needs of the family and the entire community.

Poverty is bad for a person’s health,¹ diminishing physical, cognitive, and psychological² well-being.³ Attempts to understand why persons with low socioeconomic status have poor health point to psychosocial and behavioral variables,⁴ such as smoking, bad dietary habits, exposure to trauma and violence, sedentary lifestyles, hopelessness, hostility, and depression.⁵ Mental illnesses⁶ (particularly depression⁷) cause more disability⁸ and diminished functional status than most physical illnesses.⁹ It is no surprise, therefore, that people with untreated mental illness use a disproportionate amount of health care resources.¹⁰ Thus, the more than 11 million¹¹ people living in poverty who are uninsured are a particularly vulnerable sector of our society.¹²

In primary care settings the prevalence of mental illness^13,14 and its relationship to functional status⁹ and health care use¹⁰ is well studied. However, we know little about these issues in indigent primary care populations. Miranda and colleagues¹⁵ studied 205 women at an urban, public sector, gynecology clinic and found that 48% had at least 1 psychiatric disorder. Olfson and coworkers¹⁶ studied an urban, older, low-income, mostly Hispanic, general medical population and found a high prevalence of depression, anxiety, substance use, and suicidal ideation associated with decreased function. A recent study by Woolf and colleagues¹⁷ found the functional status of inner city, indigent, primary care patients to be lower than the general population and lower than a national sample of patients with common chronic illnesses. We found no studies examining the prevalence of mental disorders and their relationship with functional status and health care use in low-income uninsured patients in primary care settings.

Family physicians¹⁸ and the Surgeon General¹⁹ have advocated for the integration of mental health into medical settings to improve care and reduce the stigmatization of mental illness. Models of collaboration^20,21 between mental health and medical providers have been shown to be effective^22-25 and cost-effective.^26,27 Attending to patient preferences about therapeutic modality enhances the effectiveness of mental disorder treatments.²⁸ However, we found no studies examining indigent patient preferences regarding the separation versus integration of medical and mental health services.

We predicted that an underprivileged, uninsured sample of primary care patients would have higher levels of mental illness and more impaired function than has been reported in general primary care samples. We also suspected that heath care utilization would be higher for indigent patients with more psychiatric symptoms compared with indigent primary care patients without psychiatric distress. Because policy,¹⁹ provider,¹⁸ and research^23,25 recommendations endorse collaborative care designs, we wanted to initiate exploration of patient preferences for these service structures, since attending to patient wishes might enhance the effectiveness of future interventions.

Methods

Setting

We conducted our study at the Marillac Clinic in Grand Junction, Colorado. Marillac is a privately funded, nonprofit, primary care clinic serving Mesa County, Colorado (3313 square miles), with a population of 113,000 in 1999. Marillac serves only people without any form of health insurance (no Medicare or Medicaid) and with household incomes less than 150% of federal poverty guidelines. In 1998,14.5% of the Mesa County population lived below the poverty level, 16.6% lacked health insurance, 4.5% were unemployed, 90% were white, and 8% were Hispanic.²⁹ The Human Subjects Review committee of St. Mary’s Hospital, Grand Junction, Colorado, approved our study.

Selection of Subjects

During April and early May 1999, all consecutive patients aged 18 years and older with clinic medical appointments were invited to participate in our study. One of 4 medical office assistants explained the study to each patient. The participants could allow or not allow study findings to be shared with their care providers after the clinic visit. Trained readers were available for those patients unable to read the survey. This service was used on 3 occasions. Participants received a $5 coupon to a local grocery store. Of the 589 patients approached, 68 refused and 21 were missed, for an enrollment of 500 patients (85%), representing 19% of the patients seen at the clinic in 1999. Of those refusing participation, 18 were not interested, 17 were too sick, 23 were too busy, and 10 cited other reasons. There were no significant differences in age or sex between the study participants and those who refused. The mean age for those who refused was 40 years (standard deviation [SD]=11.2) versus 38 years (SD=12.1) for study participants, (t(588)=1.88). Of those who refused 59% were women; 68% of the study participants were women (c2=2.42; df=1).

Data Collection

Patients. The patients completed a questionnaire before being seen by their health care providers. For patients who agreed, providers were alerted when survey results indicated the presence of 1 or more mental disorders.

Study Instrument. The study instruments included the recently validated Primary Care Evaluation of Mental Disorders (PRIME-MD) Patient Health Questionnaire (PHQ)³⁰; the 20-item Medical Outcomes Study Short Form (SF-20), a validated³¹ tool to assess functional status; and other questions described below. The PRIME-MD PHQ is a self-report version of the original PRIME-MD³² that allows researchers to assess the presence of 7 psychiatric disorders. Like the original PRIME-MD, the PHQ assessed threshold disorders (major depression, panic disorder, other anxiety disorder, bulimia) and subthreshold disorders (minor depression, binge-eating disorder, probable alcohol abuse, somatoform disorder). Because providers were blind to patient response, diagnosis of somatoform disorder was not included. Some patients were classified as symptom screen positive because they indicated distress but failed to meet criteria for a subthreshold or a threshold diagnosis. Patients who screened positive reported depressed mood or low interest on more than half the days, a panic attack in the previous 4 weeks, feeling nervous more than half the days, often feeling “you cannot control what or how much you eat,” or being “bothered a lot” by more than 6 of the 13 PHQ physical symptoms. The SF-20 measures functional status in 6 dimensions. We defined the term “disability days” as the number of days in the past 3 months patients were kept from usual activities because of not feeling well. Health care use was defined by the number of separate times during the past 3 months that patients went to a medical physician in an office, clinic, or emergency room because of not feeling well, not counting the present visit.

Other questions added to the questionnaire included a 20-item list of current physical illnesses (medical comorbidity)²⁴ and demographic questions. A question to assess patient preferences for service design asked: “In the future, if you desire mental health care, please check your top 2 preferred designs:” (A) your medical provider and mental health provider work in the same setting and communicate with one another about your care; (B) communication between providers with service at separate settings; (C) providers do not communicate with service at the same setting; and (D) providers do not communicate, and service is provided at separate settings (alternatives B, C, and D are abbreviated).

Statistical Analyses

We examined descriptive data for the sample. To test the hypothesis that poor, uninsured primary care patients will have higher rates of mental illness compared with a general primary care sample, we compared prevalence data for psychiatric disorders in the study sample against a representative primary care PHQ sample of 3000 patients. Since the PHQ study is from a different population with different sociodemographic and medical characteristics and different sampling techniques, we could not directly compare its data with statistical tests to discern differences in the populations. To determine if there were more problems with functional status and disability days and higher health care use for patients with more psychiatric symptoms, we classified patients into 3 groups based on psychiatric diagnoses: symptom screen negative, screen positive/subthreshold, and threshold groups. Originally the screen-positive and subthreshold groups were analyzed separately. However, because of nearly identical means they were combined for subsequent analyses. Analyses of covariance were used to examine differences in functioning and disability days. The covariates we used were personal income and number of physical health problems. We performed a chi-square analysis to determine if the psychiatric patient groups differed in the proportion with 3 or more physician visits in the past 3 months.

Results

Prevalence of Psychiatric Diagnoses

Table 1 shows the study sample demographics. Table 2 presents a comparison of the prevalence of current disorders in the Marillac study with those in the PHQ 3000 study. The percentage of Marillac patients with at least 1 current psychiatric diagnosis is almost twice the prevalence of the PHQ study (51% vs 28%). Marillac patients had between 2 and 3 times as many of each of the current threshold diagnoses. The rates of each current subthreshold disorder are mildly higher than the PHQ study except for probable alcohol abuse, which is more than twice as high.

Functional Status, Disability Days, and Health Care Use

Figure 1 displays the adjusted means for the 6 scales of the SF-20 and shows that patients with one or more current threshold psychiatric disorders have significantly (P <.001) lower functional status on all SF-20 scales compared with the other 2 groups, which did not differ except for mental health. The percentage of patients in each of the 3 psychiatric symptom groups were: symptom screen negative, 31%; symptom screen positive/subthreshold diagnosis, 34%; and threshold diagnosis, 35%.

Screen-negative patients reported a mean (SD) of 4.3 (8.5) disability days; screen-positive/subthreshold patients reported 5.6 (12.2) days; and threshold diagnosis patients reported 18.9 (25.6) days. Controlling for physical comorbidity and personal income, patients with threshold psychiatric diagnoses had significantly more disability days than either of the 2 other groups, which did not differ from one another (F[2453]=30.20; P <.001). The 3 groups differed in number of physician visits in the previous 3 months. Controlling for physical comorbidity and personal income, percentages of patients within each diagnostic group with 3 or more physician visits were: screen negative, 15.7%; screen positive/subthreshold, 21.7%; and threshold, 34.5%. Patients in the threshold group were more likely to report 3 or more visits than patients in the other 2 groups (c2=16.27; df=2; P <.001). Differences between the screen-positive/subthreshold and the threshold group were also significant (c2=6.77; P <.009), but differences between screen-negative and screen-positive/subthreshold groups did not reach significance (c2=1.87; P <.17).

Patient Preferences for Medical and Mental Health Service Designs

Table 3 shows patient preferences. After choosing a first option patients were asked to make a second choice, which meant changing the location of service to maintain interprovider communication or eliminating communication to maintain service at a preferred location. Of the 284 patients who marked 2 votes, 246 (87%) chose the 2 options for providers to communicate with one another. The proportion of votes for the 2 communication options within each of the symptom groups was: threshold, 91%, subthreshold/screen positive, 86%, and screen negative, 90%.

Discussion

Using an instrument recently validated across 3000 primary care patients (the PRIME-MD PHQ) we found the proportion of the patients in this clinic with current major mental illnesses to be roughly twice the number in the general population (35% vs 15%). Overall, a larger proportion of patients in the Marillac population report some current psychiatric distress compared with the sample from Spitzer and coworkers³⁰ (51% vs 28%). Because the PHQ does not diagnose dysthymia, non-alcohol–related substance abuse, or other chronic mental illnesses such as bipolar disorder, these findings represent a conservative view of the prevalence of mental illness and addictive disorders in this sample. Also, because primary care providers did not evaluate whether physical symptoms were secondary to a medical illness, somatoform disorders were not diagnosed.

Consistent with other studies^9,33 patients with threshold disorders report significantly lower functional status compared with patients with subthreshold diagnoses or who are screen positive for psychiatric distress or without any psychiatric symptoms. However, these other studies have found a gradient of functional status inversely proportional to the degree of psychiatric impairment that was absent in the Marillac sample. The mean scores for Marillac symptom screen-negative patients are 7 to 15 points lower than the PHQ 3000 symptom screen-negative patients across all 6 SF-20 scales. These findings are consistent with findings reported by Woolf and colleagues¹⁷ who found mean scores on all the functional status indices for low-income patients to be significantly lower than their overall population means. It is unclear whether these findings are because of more severe mental disorders, a higher prevalence of physical disorders, or other characteristics of low-income populations.

Consistent with other studies,^10,30 patients at Marillac with higher levels of psychiatric symptoms report increasing numbers of disability days and physician visits. Comparing disability days in PHQ 3000 patients in the threshold diagnosis (17), subthreshold (6.6), and screen-positive (4.8) groups shows similar numbers to our sample. However, the number of disability days for the Marillac patients without any psychiatric symptoms is almost twice as high as that in the PHQ 3000 sample (2.4) and consistent with the lower levels of functional status in the Marillac screen-negative group.

Although the prevalence of virtually all biomedical, psychosocial, and psychiatric illnesses is greater in the underprivileged, special attention needs to be paid to addressing cognitive, psychosocial, and psychiatric issues. The high prevalence of mental disorders may lead to chronic disability,³⁴ perpetuating poverty. Mental illness complicates the management of chronic medical illness and increases risks for illness and death.³⁵ Diminished sense of control of life compromises self-care^36,37 and well-being.³⁸

The majority (90%) of Marillac patients preferred their medical providers and mental health providers to communicate with one another about their health care. These patient preferences combined with research supporting the use of collaborative designs represented a compelling argument for system redesign. The findings of this study helped secure 4 years of funding from the Robert Wood Johnson 2000 Local Initiative Funding Partners Program to match funding from local contributors lead by the Colorado Trust. These funds will pay for on-site counselors, case managers, psychiatric and substance abuse assessments, group treatments, and ongoing training to create stronger linkages with a variety of community agencies (The Mesa County Coalition on Health). Marillac has adopted Collaborative Family Health Care,³⁹ a model emphasizing teamwork between biomedical, nursing, and psychosocial providers, and that views the patient^40,41 and family^42-44 as crucial in treatment design and implementation. System adjustments emphasized the management of chronic illness⁴⁵ with a focus on the psychosocial needs of this population.⁴⁶ More details of these changes are described elsewhere.⁴⁷

Limitations

The major limitation of our study may be lack of generalizability to other indigent primary care populations. More studies are needed that examine the prevalence of mental illnesses and relationships with functional status and disability in poor, urban populations with and without health insurance. Most subjects in our study are white and speak English. The prevalence and nature of mental disorders among urban diverse primary care patients may differ from the profiles we have described. In our study the method used to assess medical comorbidity relied on patient report. Patients may have under-reported or over-reported physical illness. Some symptoms reported on the PHQ could be caused by medical illnesses, and many may be medically unexplained.⁴⁸ In the PHQ study, mental health professionals interviewed patients to validate survey findings. We assume that responses from this low-income population are valid, but future studies may want to further validate the PHQ in indigent samples. Our data probably underestimate overall prevalence of mental disorders in Marillac patients, because the number of disorders detected by the PHQ is limited. The Marillac population was younger (18-64 years) than the PHQ-3000 sample (19-99 years). An older population may have a different prevalence of mental disorders, levels of functional status, and service use. The relationships among these variables may also be different.

Conclusions

We found an indigent uninsured primary care adult population to have an extremely high prevalence of current mental disorders. Also, in addition to the expected decrease in functional status for those with severe mental disorders, the functional status of the entire clinic population was quite low. A sizable portion of the literature suggests that much of this diminished health-related quality of life might be the expression of an impoverished existence. Beyond financial poverty and limited education, the chronically poor person suffers from a higher prevalence of mental illness and a limited sense of being able to control the future. Patient preferences support provider and policy recommendations for the integration of mental health and primary care services. These health care designs may increase our potential to improve the health of those with the greatest need.⁴⁹

Acknowledgments

Between August 1998 and July 1999 Mr Mauksch was on leave from the University of Washington Department of Family Medicine as a consultant in collaborative care to the Marillac Clinic. Funding for his position came from the Brownson Memorial Fund, the Victim/Witness Assistance and Law Enforcement Fund—21st Judicial District of Colorado, the Sisters of Charity, and St. Mary’s Hospital.

We wish to thank the Marillac Clinic medical assistants, administrative staff, community volunteers, and clinicians who contributed to this study in many essential ways. Because no additional financial support was used to fund this study, the role of the entire clinic community was critical for its successful completion.

The authors thank Jurgen Unützer, MD, for help in selecting an instrument to measure medical comorbidity.

Related Resources:

• The Collaborative Family HealthCare Coalition www.cfhcc.org
• The Institute for Healthcare Improvement-Information on “Improving Care for People with Chronic Conditions,” a national congress with a focus on asthma and depression www.ihi.org
• Anxiety Disorders Association of America www.adaa.org
• National Depressive and Manic Depressive Association www.ndmda.org

Poverty is bad for a person’s health,¹ diminishing physical, cognitive, and psychological² well-being.³ Attempts to understand why persons with low socioeconomic status have poor health point to psychosocial and behavioral variables,⁴ such as smoking, bad dietary habits, exposure to trauma and violence, sedentary lifestyles, hopelessness, hostility, and depression.⁵ Mental illnesses⁶ (particularly depression⁷) cause more disability⁸ and diminished functional status than most physical illnesses.⁹ It is no surprise, therefore, that people with untreated mental illness use a disproportionate amount of health care resources.¹⁰ Thus, the more than 11 million¹¹ people living in poverty who are uninsured are a particularly vulnerable sector of our society.¹²

In primary care settings the prevalence of mental illness^13,14 and its relationship to functional status⁹ and health care use¹⁰ is well studied. However, we know little about these issues in indigent primary care populations. Miranda and colleagues¹⁵ studied 205 women at an urban, public sector, gynecology clinic and found that 48% had at least 1 psychiatric disorder. Olfson and coworkers¹⁶ studied an urban, older, low-income, mostly Hispanic, general medical population and found a high prevalence of depression, anxiety, substance use, and suicidal ideation associated with decreased function. A recent study by Woolf and colleagues¹⁷ found the functional status of inner city, indigent, primary care patients to be lower than the general population and lower than a national sample of patients with common chronic illnesses. We found no studies examining the prevalence of mental disorders and their relationship with functional status and health care use in low-income uninsured patients in primary care settings.

Family physicians¹⁸ and the Surgeon General¹⁹ have advocated for the integration of mental health into medical settings to improve care and reduce the stigmatization of mental illness. Models of collaboration^20,21 between mental health and medical providers have been shown to be effective^22-25 and cost-effective.^26,27 Attending to patient preferences about therapeutic modality enhances the effectiveness of mental disorder treatments.²⁸ However, we found no studies examining indigent patient preferences regarding the separation versus integration of medical and mental health services.

We predicted that an underprivileged, uninsured sample of primary care patients would have higher levels of mental illness and more impaired function than has been reported in general primary care samples. We also suspected that heath care utilization would be higher for indigent patients with more psychiatric symptoms compared with indigent primary care patients without psychiatric distress. Because policy,¹⁹ provider,¹⁸ and research^23,25 recommendations endorse collaborative care designs, we wanted to initiate exploration of patient preferences for these service structures, since attending to patient wishes might enhance the effectiveness of future interventions.

Methods

Setting

We conducted our study at the Marillac Clinic in Grand Junction, Colorado. Marillac is a privately funded, nonprofit, primary care clinic serving Mesa County, Colorado (3313 square miles), with a population of 113,000 in 1999. Marillac serves only people without any form of health insurance (no Medicare or Medicaid) and with household incomes less than 150% of federal poverty guidelines. In 1998,14.5% of the Mesa County population lived below the poverty level, 16.6% lacked health insurance, 4.5% were unemployed, 90% were white, and 8% were Hispanic.²⁹ The Human Subjects Review committee of St. Mary’s Hospital, Grand Junction, Colorado, approved our study.

Selection of Subjects

During April and early May 1999, all consecutive patients aged 18 years and older with clinic medical appointments were invited to participate in our study. One of 4 medical office assistants explained the study to each patient. The participants could allow or not allow study findings to be shared with their care providers after the clinic visit. Trained readers were available for those patients unable to read the survey. This service was used on 3 occasions. Participants received a $5 coupon to a local grocery store. Of the 589 patients approached, 68 refused and 21 were missed, for an enrollment of 500 patients (85%), representing 19% of the patients seen at the clinic in 1999. Of those refusing participation, 18 were not interested, 17 were too sick, 23 were too busy, and 10 cited other reasons. There were no significant differences in age or sex between the study participants and those who refused. The mean age for those who refused was 40 years (standard deviation [SD]=11.2) versus 38 years (SD=12.1) for study participants, (t(588)=1.88). Of those who refused 59% were women; 68% of the study participants were women (c2=2.42; df=1).

Data Collection

Patients. The patients completed a questionnaire before being seen by their health care providers. For patients who agreed, providers were alerted when survey results indicated the presence of 1 or more mental disorders.

Study Instrument. The study instruments included the recently validated Primary Care Evaluation of Mental Disorders (PRIME-MD) Patient Health Questionnaire (PHQ)³⁰; the 20-item Medical Outcomes Study Short Form (SF-20), a validated³¹ tool to assess functional status; and other questions described below. The PRIME-MD PHQ is a self-report version of the original PRIME-MD³² that allows researchers to assess the presence of 7 psychiatric disorders. Like the original PRIME-MD, the PHQ assessed threshold disorders (major depression, panic disorder, other anxiety disorder, bulimia) and subthreshold disorders (minor depression, binge-eating disorder, probable alcohol abuse, somatoform disorder). Because providers were blind to patient response, diagnosis of somatoform disorder was not included. Some patients were classified as symptom screen positive because they indicated distress but failed to meet criteria for a subthreshold or a threshold diagnosis. Patients who screened positive reported depressed mood or low interest on more than half the days, a panic attack in the previous 4 weeks, feeling nervous more than half the days, often feeling “you cannot control what or how much you eat,” or being “bothered a lot” by more than 6 of the 13 PHQ physical symptoms. The SF-20 measures functional status in 6 dimensions. We defined the term “disability days” as the number of days in the past 3 months patients were kept from usual activities because of not feeling well. Health care use was defined by the number of separate times during the past 3 months that patients went to a medical physician in an office, clinic, or emergency room because of not feeling well, not counting the present visit.

Other questions added to the questionnaire included a 20-item list of current physical illnesses (medical comorbidity)²⁴ and demographic questions. A question to assess patient preferences for service design asked: “In the future, if you desire mental health care, please check your top 2 preferred designs:” (A) your medical provider and mental health provider work in the same setting and communicate with one another about your care; (B) communication between providers with service at separate settings; (C) providers do not communicate with service at the same setting; and (D) providers do not communicate, and service is provided at separate settings (alternatives B, C, and D are abbreviated).

Statistical Analyses

We examined descriptive data for the sample. To test the hypothesis that poor, uninsured primary care patients will have higher rates of mental illness compared with a general primary care sample, we compared prevalence data for psychiatric disorders in the study sample against a representative primary care PHQ sample of 3000 patients. Since the PHQ study is from a different population with different sociodemographic and medical characteristics and different sampling techniques, we could not directly compare its data with statistical tests to discern differences in the populations. To determine if there were more problems with functional status and disability days and higher health care use for patients with more psychiatric symptoms, we classified patients into 3 groups based on psychiatric diagnoses: symptom screen negative, screen positive/subthreshold, and threshold groups. Originally the screen-positive and subthreshold groups were analyzed separately. However, because of nearly identical means they were combined for subsequent analyses. Analyses of covariance were used to examine differences in functioning and disability days. The covariates we used were personal income and number of physical health problems. We performed a chi-square analysis to determine if the psychiatric patient groups differed in the proportion with 3 or more physician visits in the past 3 months.

Results

Prevalence of Psychiatric Diagnoses

Table 1 shows the study sample demographics. Table 2 presents a comparison of the prevalence of current disorders in the Marillac study with those in the PHQ 3000 study. The percentage of Marillac patients with at least 1 current psychiatric diagnosis is almost twice the prevalence of the PHQ study (51% vs 28%). Marillac patients had between 2 and 3 times as many of each of the current threshold diagnoses. The rates of each current subthreshold disorder are mildly higher than the PHQ study except for probable alcohol abuse, which is more than twice as high.

Functional Status, Disability Days, and Health Care Use

Figure 1 displays the adjusted means for the 6 scales of the SF-20 and shows that patients with one or more current threshold psychiatric disorders have significantly (P <.001) lower functional status on all SF-20 scales compared with the other 2 groups, which did not differ except for mental health. The percentage of patients in each of the 3 psychiatric symptom groups were: symptom screen negative, 31%; symptom screen positive/subthreshold diagnosis, 34%; and threshold diagnosis, 35%.

Screen-negative patients reported a mean (SD) of 4.3 (8.5) disability days; screen-positive/subthreshold patients reported 5.6 (12.2) days; and threshold diagnosis patients reported 18.9 (25.6) days. Controlling for physical comorbidity and personal income, patients with threshold psychiatric diagnoses had significantly more disability days than either of the 2 other groups, which did not differ from one another (F[2453]=30.20; P <.001). The 3 groups differed in number of physician visits in the previous 3 months. Controlling for physical comorbidity and personal income, percentages of patients within each diagnostic group with 3 or more physician visits were: screen negative, 15.7%; screen positive/subthreshold, 21.7%; and threshold, 34.5%. Patients in the threshold group were more likely to report 3 or more visits than patients in the other 2 groups (c2=16.27; df=2; P <.001). Differences between the screen-positive/subthreshold and the threshold group were also significant (c2=6.77; P <.009), but differences between screen-negative and screen-positive/subthreshold groups did not reach significance (c2=1.87; P <.17).

Patient Preferences for Medical and Mental Health Service Designs

Table 3 shows patient preferences. After choosing a first option patients were asked to make a second choice, which meant changing the location of service to maintain interprovider communication or eliminating communication to maintain service at a preferred location. Of the 284 patients who marked 2 votes, 246 (87%) chose the 2 options for providers to communicate with one another. The proportion of votes for the 2 communication options within each of the symptom groups was: threshold, 91%, subthreshold/screen positive, 86%, and screen negative, 90%.

Discussion

Using an instrument recently validated across 3000 primary care patients (the PRIME-MD PHQ) we found the proportion of the patients in this clinic with current major mental illnesses to be roughly twice the number in the general population (35% vs 15%). Overall, a larger proportion of patients in the Marillac population report some current psychiatric distress compared with the sample from Spitzer and coworkers³⁰ (51% vs 28%). Because the PHQ does not diagnose dysthymia, non-alcohol–related substance abuse, or other chronic mental illnesses such as bipolar disorder, these findings represent a conservative view of the prevalence of mental illness and addictive disorders in this sample. Also, because primary care providers did not evaluate whether physical symptoms were secondary to a medical illness, somatoform disorders were not diagnosed.

Consistent with other studies^9,33 patients with threshold disorders report significantly lower functional status compared with patients with subthreshold diagnoses or who are screen positive for psychiatric distress or without any psychiatric symptoms. However, these other studies have found a gradient of functional status inversely proportional to the degree of psychiatric impairment that was absent in the Marillac sample. The mean scores for Marillac symptom screen-negative patients are 7 to 15 points lower than the PHQ 3000 symptom screen-negative patients across all 6 SF-20 scales. These findings are consistent with findings reported by Woolf and colleagues¹⁷ who found mean scores on all the functional status indices for low-income patients to be significantly lower than their overall population means. It is unclear whether these findings are because of more severe mental disorders, a higher prevalence of physical disorders, or other characteristics of low-income populations.

Consistent with other studies,^10,30 patients at Marillac with higher levels of psychiatric symptoms report increasing numbers of disability days and physician visits. Comparing disability days in PHQ 3000 patients in the threshold diagnosis (17), subthreshold (6.6), and screen-positive (4.8) groups shows similar numbers to our sample. However, the number of disability days for the Marillac patients without any psychiatric symptoms is almost twice as high as that in the PHQ 3000 sample (2.4) and consistent with the lower levels of functional status in the Marillac screen-negative group.

Although the prevalence of virtually all biomedical, psychosocial, and psychiatric illnesses is greater in the underprivileged, special attention needs to be paid to addressing cognitive, psychosocial, and psychiatric issues. The high prevalence of mental disorders may lead to chronic disability,³⁴ perpetuating poverty. Mental illness complicates the management of chronic medical illness and increases risks for illness and death.³⁵ Diminished sense of control of life compromises self-care^36,37 and well-being.³⁸

The majority (90%) of Marillac patients preferred their medical providers and mental health providers to communicate with one another about their health care. These patient preferences combined with research supporting the use of collaborative designs represented a compelling argument for system redesign. The findings of this study helped secure 4 years of funding from the Robert Wood Johnson 2000 Local Initiative Funding Partners Program to match funding from local contributors lead by the Colorado Trust. These funds will pay for on-site counselors, case managers, psychiatric and substance abuse assessments, group treatments, and ongoing training to create stronger linkages with a variety of community agencies (The Mesa County Coalition on Health). Marillac has adopted Collaborative Family Health Care,³⁹ a model emphasizing teamwork between biomedical, nursing, and psychosocial providers, and that views the patient^40,41 and family^42-44 as crucial in treatment design and implementation. System adjustments emphasized the management of chronic illness⁴⁵ with a focus on the psychosocial needs of this population.⁴⁶ More details of these changes are described elsewhere.⁴⁷

Limitations

The major limitation of our study may be lack of generalizability to other indigent primary care populations. More studies are needed that examine the prevalence of mental illnesses and relationships with functional status and disability in poor, urban populations with and without health insurance. Most subjects in our study are white and speak English. The prevalence and nature of mental disorders among urban diverse primary care patients may differ from the profiles we have described. In our study the method used to assess medical comorbidity relied on patient report. Patients may have under-reported or over-reported physical illness. Some symptoms reported on the PHQ could be caused by medical illnesses, and many may be medically unexplained.⁴⁸ In the PHQ study, mental health professionals interviewed patients to validate survey findings. We assume that responses from this low-income population are valid, but future studies may want to further validate the PHQ in indigent samples. Our data probably underestimate overall prevalence of mental disorders in Marillac patients, because the number of disorders detected by the PHQ is limited. The Marillac population was younger (18-64 years) than the PHQ-3000 sample (19-99 years). An older population may have a different prevalence of mental disorders, levels of functional status, and service use. The relationships among these variables may also be different.

Conclusions

We found an indigent uninsured primary care adult population to have an extremely high prevalence of current mental disorders. Also, in addition to the expected decrease in functional status for those with severe mental disorders, the functional status of the entire clinic population was quite low. A sizable portion of the literature suggests that much of this diminished health-related quality of life might be the expression of an impoverished existence. Beyond financial poverty and limited education, the chronically poor person suffers from a higher prevalence of mental illness and a limited sense of being able to control the future. Patient preferences support provider and policy recommendations for the integration of mental health and primary care services. These health care designs may increase our potential to improve the health of those with the greatest need.⁴⁹

Acknowledgments

Between August 1998 and July 1999 Mr Mauksch was on leave from the University of Washington Department of Family Medicine as a consultant in collaborative care to the Marillac Clinic. Funding for his position came from the Brownson Memorial Fund, the Victim/Witness Assistance and Law Enforcement Fund—21st Judicial District of Colorado, the Sisters of Charity, and St. Mary’s Hospital.

We wish to thank the Marillac Clinic medical assistants, administrative staff, community volunteers, and clinicians who contributed to this study in many essential ways. Because no additional financial support was used to fund this study, the role of the entire clinic community was critical for its successful completion.

The authors thank Jurgen Unützer, MD, for help in selecting an instrument to measure medical comorbidity.

Related Resources:

• The Collaborative Family HealthCare Coalition www.cfhcc.org
• The Institute for Healthcare Improvement-Information on “Improving Care for People with Chronic Conditions,” a national congress with a focus on asthma and depression www.ihi.org
• Anxiety Disorders Association of America www.adaa.org
• National Depressive and Manic Depressive Association www.ndmda.org

Poverty is bad for a person’s health,¹ diminishing physical, cognitive, and psychological² well-being.³ Attempts to understand why persons with low socioeconomic status have poor health point to psychosocial and behavioral variables,⁴ such as smoking, bad dietary habits, exposure to trauma and violence, sedentary lifestyles, hopelessness, hostility, and depression.⁵ Mental illnesses⁶ (particularly depression⁷) cause more disability⁸ and diminished functional status than most physical illnesses.⁹ It is no surprise, therefore, that people with untreated mental illness use a disproportionate amount of health care resources.¹⁰ Thus, the more than 11 million¹¹ people living in poverty who are uninsured are a particularly vulnerable sector of our society.¹²

In primary care settings the prevalence of mental illness^13,14 and its relationship to functional status⁹ and health care use¹⁰ is well studied. However, we know little about these issues in indigent primary care populations. Miranda and colleagues¹⁵ studied 205 women at an urban, public sector, gynecology clinic and found that 48% had at least 1 psychiatric disorder. Olfson and coworkers¹⁶ studied an urban, older, low-income, mostly Hispanic, general medical population and found a high prevalence of depression, anxiety, substance use, and suicidal ideation associated with decreased function. A recent study by Woolf and colleagues¹⁷ found the functional status of inner city, indigent, primary care patients to be lower than the general population and lower than a national sample of patients with common chronic illnesses. We found no studies examining the prevalence of mental disorders and their relationship with functional status and health care use in low-income uninsured patients in primary care settings.

Family physicians¹⁸ and the Surgeon General¹⁹ have advocated for the integration of mental health into medical settings to improve care and reduce the stigmatization of mental illness. Models of collaboration^20,21 between mental health and medical providers have been shown to be effective^22-25 and cost-effective.^26,27 Attending to patient preferences about therapeutic modality enhances the effectiveness of mental disorder treatments.²⁸ However, we found no studies examining indigent patient preferences regarding the separation versus integration of medical and mental health services.

We predicted that an underprivileged, uninsured sample of primary care patients would have higher levels of mental illness and more impaired function than has been reported in general primary care samples. We also suspected that heath care utilization would be higher for indigent patients with more psychiatric symptoms compared with indigent primary care patients without psychiatric distress. Because policy,¹⁹ provider,¹⁸ and research^23,25 recommendations endorse collaborative care designs, we wanted to initiate exploration of patient preferences for these service structures, since attending to patient wishes might enhance the effectiveness of future interventions.

Methods

Setting

We conducted our study at the Marillac Clinic in Grand Junction, Colorado. Marillac is a privately funded, nonprofit, primary care clinic serving Mesa County, Colorado (3313 square miles), with a population of 113,000 in 1999. Marillac serves only people without any form of health insurance (no Medicare or Medicaid) and with household incomes less than 150% of federal poverty guidelines. In 1998,14.5% of the Mesa County population lived below the poverty level, 16.6% lacked health insurance, 4.5% were unemployed, 90% were white, and 8% were Hispanic.²⁹ The Human Subjects Review committee of St. Mary’s Hospital, Grand Junction, Colorado, approved our study.

Selection of Subjects

During April and early May 1999, all consecutive patients aged 18 years and older with clinic medical appointments were invited to participate in our study. One of 4 medical office assistants explained the study to each patient. The participants could allow or not allow study findings to be shared with their care providers after the clinic visit. Trained readers were available for those patients unable to read the survey. This service was used on 3 occasions. Participants received a $5 coupon to a local grocery store. Of the 589 patients approached, 68 refused and 21 were missed, for an enrollment of 500 patients (85%), representing 19% of the patients seen at the clinic in 1999. Of those refusing participation, 18 were not interested, 17 were too sick, 23 were too busy, and 10 cited other reasons. There were no significant differences in age or sex between the study participants and those who refused. The mean age for those who refused was 40 years (standard deviation [SD]=11.2) versus 38 years (SD=12.1) for study participants, (t(588)=1.88). Of those who refused 59% were women; 68% of the study participants were women (c2=2.42; df=1).

Data Collection

Patients. The patients completed a questionnaire before being seen by their health care providers. For patients who agreed, providers were alerted when survey results indicated the presence of 1 or more mental disorders.

Study Instrument. The study instruments included the recently validated Primary Care Evaluation of Mental Disorders (PRIME-MD) Patient Health Questionnaire (PHQ)³⁰; the 20-item Medical Outcomes Study Short Form (SF-20), a validated³¹ tool to assess functional status; and other questions described below. The PRIME-MD PHQ is a self-report version of the original PRIME-MD³² that allows researchers to assess the presence of 7 psychiatric disorders. Like the original PRIME-MD, the PHQ assessed threshold disorders (major depression, panic disorder, other anxiety disorder, bulimia) and subthreshold disorders (minor depression, binge-eating disorder, probable alcohol abuse, somatoform disorder). Because providers were blind to patient response, diagnosis of somatoform disorder was not included. Some patients were classified as symptom screen positive because they indicated distress but failed to meet criteria for a subthreshold or a threshold diagnosis. Patients who screened positive reported depressed mood or low interest on more than half the days, a panic attack in the previous 4 weeks, feeling nervous more than half the days, often feeling “you cannot control what or how much you eat,” or being “bothered a lot” by more than 6 of the 13 PHQ physical symptoms. The SF-20 measures functional status in 6 dimensions. We defined the term “disability days” as the number of days in the past 3 months patients were kept from usual activities because of not feeling well. Health care use was defined by the number of separate times during the past 3 months that patients went to a medical physician in an office, clinic, or emergency room because of not feeling well, not counting the present visit.

Other questions added to the questionnaire included a 20-item list of current physical illnesses (medical comorbidity)²⁴ and demographic questions. A question to assess patient preferences for service design asked: “In the future, if you desire mental health care, please check your top 2 preferred designs:” (A) your medical provider and mental health provider work in the same setting and communicate with one another about your care; (B) communication between providers with service at separate settings; (C) providers do not communicate with service at the same setting; and (D) providers do not communicate, and service is provided at separate settings (alternatives B, C, and D are abbreviated).

Statistical Analyses

We examined descriptive data for the sample. To test the hypothesis that poor, uninsured primary care patients will have higher rates of mental illness compared with a general primary care sample, we compared prevalence data for psychiatric disorders in the study sample against a representative primary care PHQ sample of 3000 patients. Since the PHQ study is from a different population with different sociodemographic and medical characteristics and different sampling techniques, we could not directly compare its data with statistical tests to discern differences in the populations. To determine if there were more problems with functional status and disability days and higher health care use for patients with more psychiatric symptoms, we classified patients into 3 groups based on psychiatric diagnoses: symptom screen negative, screen positive/subthreshold, and threshold groups. Originally the screen-positive and subthreshold groups were analyzed separately. However, because of nearly identical means they were combined for subsequent analyses. Analyses of covariance were used to examine differences in functioning and disability days. The covariates we used were personal income and number of physical health problems. We performed a chi-square analysis to determine if the psychiatric patient groups differed in the proportion with 3 or more physician visits in the past 3 months.

Results

Prevalence of Psychiatric Diagnoses

Table 1 shows the study sample demographics. Table 2 presents a comparison of the prevalence of current disorders in the Marillac study with those in the PHQ 3000 study. The percentage of Marillac patients with at least 1 current psychiatric diagnosis is almost twice the prevalence of the PHQ study (51% vs 28%). Marillac patients had between 2 and 3 times as many of each of the current threshold diagnoses. The rates of each current subthreshold disorder are mildly higher than the PHQ study except for probable alcohol abuse, which is more than twice as high.

Functional Status, Disability Days, and Health Care Use

Figure 1 displays the adjusted means for the 6 scales of the SF-20 and shows that patients with one or more current threshold psychiatric disorders have significantly (P <.001) lower functional status on all SF-20 scales compared with the other 2 groups, which did not differ except for mental health. The percentage of patients in each of the 3 psychiatric symptom groups were: symptom screen negative, 31%; symptom screen positive/subthreshold diagnosis, 34%; and threshold diagnosis, 35%.

Screen-negative patients reported a mean (SD) of 4.3 (8.5) disability days; screen-positive/subthreshold patients reported 5.6 (12.2) days; and threshold diagnosis patients reported 18.9 (25.6) days. Controlling for physical comorbidity and personal income, patients with threshold psychiatric diagnoses had significantly more disability days than either of the 2 other groups, which did not differ from one another (F[2453]=30.20; P <.001). The 3 groups differed in number of physician visits in the previous 3 months. Controlling for physical comorbidity and personal income, percentages of patients within each diagnostic group with 3 or more physician visits were: screen negative, 15.7%; screen positive/subthreshold, 21.7%; and threshold, 34.5%. Patients in the threshold group were more likely to report 3 or more visits than patients in the other 2 groups (c2=16.27; df=2; P <.001). Differences between the screen-positive/subthreshold and the threshold group were also significant (c2=6.77; P <.009), but differences between screen-negative and screen-positive/subthreshold groups did not reach significance (c2=1.87; P <.17).

Patient Preferences for Medical and Mental Health Service Designs

Table 3 shows patient preferences. After choosing a first option patients were asked to make a second choice, which meant changing the location of service to maintain interprovider communication or eliminating communication to maintain service at a preferred location. Of the 284 patients who marked 2 votes, 246 (87%) chose the 2 options for providers to communicate with one another. The proportion of votes for the 2 communication options within each of the symptom groups was: threshold, 91%, subthreshold/screen positive, 86%, and screen negative, 90%.

Discussion

Using an instrument recently validated across 3000 primary care patients (the PRIME-MD PHQ) we found the proportion of the patients in this clinic with current major mental illnesses to be roughly twice the number in the general population (35% vs 15%). Overall, a larger proportion of patients in the Marillac population report some current psychiatric distress compared with the sample from Spitzer and coworkers³⁰ (51% vs 28%). Because the PHQ does not diagnose dysthymia, non-alcohol–related substance abuse, or other chronic mental illnesses such as bipolar disorder, these findings represent a conservative view of the prevalence of mental illness and addictive disorders in this sample. Also, because primary care providers did not evaluate whether physical symptoms were secondary to a medical illness, somatoform disorders were not diagnosed.

Consistent with other studies^9,33 patients with threshold disorders report significantly lower functional status compared with patients with subthreshold diagnoses or who are screen positive for psychiatric distress or without any psychiatric symptoms. However, these other studies have found a gradient of functional status inversely proportional to the degree of psychiatric impairment that was absent in the Marillac sample. The mean scores for Marillac symptom screen-negative patients are 7 to 15 points lower than the PHQ 3000 symptom screen-negative patients across all 6 SF-20 scales. These findings are consistent with findings reported by Woolf and colleagues¹⁷ who found mean scores on all the functional status indices for low-income patients to be significantly lower than their overall population means. It is unclear whether these findings are because of more severe mental disorders, a higher prevalence of physical disorders, or other characteristics of low-income populations.

Consistent with other studies,^10,30 patients at Marillac with higher levels of psychiatric symptoms report increasing numbers of disability days and physician visits. Comparing disability days in PHQ 3000 patients in the threshold diagnosis (17), subthreshold (6.6), and screen-positive (4.8) groups shows similar numbers to our sample. However, the number of disability days for the Marillac patients without any psychiatric symptoms is almost twice as high as that in the PHQ 3000 sample (2.4) and consistent with the lower levels of functional status in the Marillac screen-negative group.

Although the prevalence of virtually all biomedical, psychosocial, and psychiatric illnesses is greater in the underprivileged, special attention needs to be paid to addressing cognitive, psychosocial, and psychiatric issues. The high prevalence of mental disorders may lead to chronic disability,³⁴ perpetuating poverty. Mental illness complicates the management of chronic medical illness and increases risks for illness and death.³⁵ Diminished sense of control of life compromises self-care^36,37 and well-being.³⁸

The majority (90%) of Marillac patients preferred their medical providers and mental health providers to communicate with one another about their health care. These patient preferences combined with research supporting the use of collaborative designs represented a compelling argument for system redesign. The findings of this study helped secure 4 years of funding from the Robert Wood Johnson 2000 Local Initiative Funding Partners Program to match funding from local contributors lead by the Colorado Trust. These funds will pay for on-site counselors, case managers, psychiatric and substance abuse assessments, group treatments, and ongoing training to create stronger linkages with a variety of community agencies (The Mesa County Coalition on Health). Marillac has adopted Collaborative Family Health Care,³⁹ a model emphasizing teamwork between biomedical, nursing, and psychosocial providers, and that views the patient^40,41 and family^42-44 as crucial in treatment design and implementation. System adjustments emphasized the management of chronic illness⁴⁵ with a focus on the psychosocial needs of this population.⁴⁶ More details of these changes are described elsewhere.⁴⁷

Limitations

The major limitation of our study may be lack of generalizability to other indigent primary care populations. More studies are needed that examine the prevalence of mental illnesses and relationships with functional status and disability in poor, urban populations with and without health insurance. Most subjects in our study are white and speak English. The prevalence and nature of mental disorders among urban diverse primary care patients may differ from the profiles we have described. In our study the method used to assess medical comorbidity relied on patient report. Patients may have under-reported or over-reported physical illness. Some symptoms reported on the PHQ could be caused by medical illnesses, and many may be medically unexplained.⁴⁸ In the PHQ study, mental health professionals interviewed patients to validate survey findings. We assume that responses from this low-income population are valid, but future studies may want to further validate the PHQ in indigent samples. Our data probably underestimate overall prevalence of mental disorders in Marillac patients, because the number of disorders detected by the PHQ is limited. The Marillac population was younger (18-64 years) than the PHQ-3000 sample (19-99 years). An older population may have a different prevalence of mental disorders, levels of functional status, and service use. The relationships among these variables may also be different.

Conclusions

We found an indigent uninsured primary care adult population to have an extremely high prevalence of current mental disorders. Also, in addition to the expected decrease in functional status for those with severe mental disorders, the functional status of the entire clinic population was quite low. A sizable portion of the literature suggests that much of this diminished health-related quality of life might be the expression of an impoverished existence. Beyond financial poverty and limited education, the chronically poor person suffers from a higher prevalence of mental illness and a limited sense of being able to control the future. Patient preferences support provider and policy recommendations for the integration of mental health and primary care services. These health care designs may increase our potential to improve the health of those with the greatest need.⁴⁹

Acknowledgments

Between August 1998 and July 1999 Mr Mauksch was on leave from the University of Washington Department of Family Medicine as a consultant in collaborative care to the Marillac Clinic. Funding for his position came from the Brownson Memorial Fund, the Victim/Witness Assistance and Law Enforcement Fund—21st Judicial District of Colorado, the Sisters of Charity, and St. Mary’s Hospital.

We wish to thank the Marillac Clinic medical assistants, administrative staff, community volunteers, and clinicians who contributed to this study in many essential ways. Because no additional financial support was used to fund this study, the role of the entire clinic community was critical for its successful completion.

The authors thank Jurgen Unützer, MD, for help in selecting an instrument to measure medical comorbidity.

Related Resources:

• The Collaborative Family HealthCare Coalition www.cfhcc.org
• The Institute for Healthcare Improvement-Information on “Improving Care for People with Chronic Conditions,” a national congress with a focus on asthma and depression www.ihi.org
• Anxiety Disorders Association of America www.adaa.org
• National Depressive and Manic Depressive Association www.ndmda.org

Infections of the upper respiratory tract account for some of the most common acute illnesses seen in primary care settings. The term “upper respiratory infection” (URI) covers any infectious disease process that involves the respiratory system, starting with the nose and ending just before the lungs. Our study dealt exclusively with the common cold.

Most patients with typical URI syndromes can be treated symptomatically. Although antimicrobial therapy is indicated in the presence of bacterial infection, it is believed that most cases are viral in nature.¹ However, of those patients who seek a physician’s care for colds and bronchitis, 50% to 70% receive an antibiotic prescription.² Ten percent of all antibiotics prescribed are for the common cold and other URIs.² The percentage is even higher in the pediatric population where antibiotic prescriptions for colds, URIs, and bronchitis (ie, conditions not affected by antibiotics) accounted for more than 20% of all antibiotics prescribed to US children in 1992.³

Prescribing antibiotics for URIs does not improve patient outcome, and this practice does not benefit physicians by reducing return visits or increasing patient satisfaction.^4,5 It is also not a cost-effective strategy. Evidence from a Medicaid population suggests that the antibiotics used to treat colds account for 23% of the total cost of managing URIs and add more than $11 to the cost of managing every URI episode.⁶ Nevertheless, a 1996 study conducted on a Medicaid population concluded that a majority of individuals receiving medical care for the common cold are still given prescriptions for unnecessary antibiotics.⁷

Clinicians often report that they are motivated to prescribe antibiotics by patient expectations.⁸ For example, parents frequently have misconceptions about which illnesses warrant antibiotic therapy leading them to request these drugs for their children.⁹ A survey conducted in Kentucky showed that when patients do not recognize the normal presentation of a URI or understand the effectiveness of antibiotics, inappropriate use and expectations may arise.¹⁰ Also questionable is whether physicians are able to accurately identify situations for which antibiotics are appropriate. Even without the influence of misinformed patients, physicians may be prescribing antibiotics inappropriately because of misdiagnosis.¹¹

The purpose of our study was to determine what a select segment of the population (undergraduate college students) knows about URIs and the perceptions of antibiotic therapy held by members of that segment. The information provided can contribute to our understanding of what types of interventions are required to change patients’ perceptions about the appropriateness of antibiotic therapy.

Methods

Participants (n=425) were students aged 18 years and older on 3 college campuses in Louisiana and Indiana. Two of the colleges were public institutions; the third was private. Data were collected in November 1999 in accordance with human subject guidelines after approval by the appropriate institutional review boards. Research assistants distributed the surveys in public areas on each of the campuses.

Survey Instrument

Participants completed a self-report survey primarily composed of 3 symptom scenarios. Two of these scenarios were employed in previous studies.^10,12 The scenarios represented variations in presentation of an uncomplicated URI along 3 dimensions: duration of symptoms, color of nasal discharge, and the presence of a low-grade fever. The scenarios were: (1) “You have had an illness for 5 days with the following symptoms: sore throat, cough, and runny nose with clear nasal discharge”; (2) “You have had an illness for 5 days with the following symptoms: sore throat, cough, and runny nose with discolored (yellow, green, brown) discharge”; and (3) “You have had an illness for 3 days with the following symptoms: sore throat, cough, runny nose with clear discharge, and low grade fever (less than 101ÞF).”

Following each scenario’s presentation, a participant was asked to indicate on a 5-point Likert-type scale (1=very likely; 5=very unlikely) how likely he or she was to seek care from a physician for the illness and the likelihood of using several treatment modalities (eg, antibiotics, antihistamines, pain relievers, vitamin C) for the presented condition.

Participants were also queried about a variety of demographic variables, current and past smoking status, and their belief in the effectiveness of antibiotics against the common cold. Finally, participants were asked to indicate whether they would see a physician if they had a cold (no specific symptoms were provided to define “cold”). The questionnaire was designed for self-administration and required less than 5 minutes to complete.

Analysis

Likert-based responses on the questionnaire were dichotomized by combining “very likely” and “somewhat likely” into 1 group and “neutral,” “somewhat unlikely,” and “very unlikely” responses into another. These responses and the yes or no responses were analyzed using chi-square tests.

The likelihood of seeking care and using antibiotics to treat a cold were averaged across scenarios. We used bivariate analyses to examine the relationship of demographic characteristics to averaged likelihood of seeking care and antibiotic use.

Stepwise multiple regression analyses were employed to examine the effects of participant characteristics on likelihood of seeking care and likelihood of using antibiotics. Because the regression analysis for each individual scenario produced similar results, we only report the analysis employing averaged likelihood of seeking care and averaged likelihood of using antibiotics. Participant demographic characteristics (continuous and discrete) that were included as independent variables in these analyses were sex, race, age, type of health insurance, and year of college. In analyzing likelihood of seeking care we included the additional variable of belief in the effectiveness of antibiotics in treating the common cold as an independent variable. We performed an additional stepwise multiple regression analysis using belief in antibiotic effectiveness as the dependent variable and participant demographic characteristics as the independent variables.

Results

The demographic characteristics of the total sample and each recruitment site are provided in Table 1. In response to a free-format question on the survey, 24% of students in the total sample reported enrollment in a science-related field of study.

Responses to the questionnaire were first examined by campus. Although the campus samples differed significantly in some ways (eg, sex and racial distributions, average age, year in college, type of health insurance), no significant differences existed between campuses in terms of the study’s variables of focus. Therefore, we conducted all analyses on the total sample.

Antibiotic Effectiveness

Forty-one percent of the total sample believed that antibiotics were effective for treating the common cold Table 2. Of those who reported a belief that antibiotics were effective for cold treatment, 24% would see a physician for a cold (10% of the total sample). Of those who did not believe in the effectiveness of antibiotics for treating colds, 12% would still seek a physician’s care for a cold (7% of the total sample).

Symptom Complex Analysis

Analysis of antibiotic use by symptom complex Table 3 revealed that 63% of all students would be “somewhat likely “or “very likely” to use antibiotics in the scenario of 5 days with discolored discharge, compared with 53% in the 3 days with clear discharge with low-grade fever scenario, and 44% in the scenario of 5 days with clear discharge (P <.001 for each comparison). Percentages were higher for those students who believe antibiotics are effective in treating the common cold: 77%, 73%, and 64%, respectively. Even among those students reporting that antibiotics are not effective in treating colds, a high number of students endorsed “somewhat likely” or “very likely” antibiotic use when presented with the 3 scenarios (32% for 5 days with clear discharge; 42% for 3 days with clear discharge and low-grade fever; 55% for 5 days with discolored discharge).

Participants’ beliefs about whether they would seek a physician’s care were also analyzed by symptom complex Table 4. Fifty-six percent of all students surveyed reported being “somewhat likely” or “very likely” to see a physician in the scenario of 5 days with discolored nasal discharge, which was a significantly greater percentage than those who would see a physician for 3 days with clear discharge and a low-grade fever (44%) and those who would see a physician if they experienced 5 days of clear discharge (29%). Percentages of those seeking medical help followed the same pattern among those students who believe antibiotics are effective in treating the common cold and those who do not believe in the effectiveness of antibiotics.

Care-seeking and antibiotic use were also analyzed separately for students who would seek a physician’s care for a common cold and those who would not Table 5. Not surprisingly, more than 50% of the students who reported that they would seek a physician’s care for a common cold endorsed seeking care in each of the presented scenarios. Likelihood of antibiotic use did not differ by symptom complex within this group, with more than 70% reporting they would use antibiotics in each of the scenarios. Among students who would not seek a physician’s help for a common cold, a significant number reported that they would still see a physician when faced with the described sets of symptoms (23% to 49% based on the scenario). High numbers of students who reported that they would not seek treatment for a common cold still reported that they were likely to use antibiotics in the presented scenarios (39% to 59% based on symptom complex).

Comparative Data

We examined the relationship of demographic characteristics to average likelihood of seeking care and average likelihood of using antibiotics (across scenarios). Reported likelihood of using antibiotics was not significantly related to any of the demographic characteristics (sex, race/ethnicity, age, college year, smoking status, or type of health insurance). Women (mean=2.89±1.14) reported being more likely than men (mean=3.17±1.19) to seek care across scenarios (P <.017). Age was also significantly related to care seeking (r=17; P <.001). The relationships between college year and care seeking and between type of health insurance and care seeking are depicted in Table 6. Freshmen and sophomores were less likely to seek care than juniors and seniors. Those using university health services were more likely than those with a private physician or other health arrangement to seek care (P <.001). Smoking status and race/ethnicity were not found to be related to reported likelihood of seeking care within this sample.

We conducted stepwise multiple regression analyses to examine predictors of belief in the effectiveness of antibiotics for cold symptoms, reported likelihood of antibiotic use across scenarios, and likelihood of seeking care across scenarios. Year of college was the only significant predictor of belief in the effectiveness of antibiotics for cold symptoms (b=0.18; P=.001). Belief in antibiotic effectiveness decreased with increasing years of higher education. Age was the only significant predictor of antibiotic use (b=0.16; P=.001). Greater age was associated with increased likelihood of using antibiotics. Average reported likelihood of seeking care was predicted by type of health insurance (b=0.22; P=.001) and by belief in the effectiveness of antibiotics to treat common colds (b=0.11; P=.03).

Discussion

A majority of students surveyed in this study would use antibiotics for the symptoms of a common cold, especially when accompanied by low-grade fever or discolored nasal discharge. This belief persisted in a significant portion even for the scenario with 5 days of clear nasal discharge and no fever. This is commensurate with other studies of different populations in the literature.¹³ A majority of students who reported they would not see a physician for a common cold still thought they would seek care for the presented scenarios, indicating a tendency toward inaccurate self-diagnosis especially when faced with symptoms perceived to be indicators of greater illness (eg, fever, discolored nasal discharge). Thick and opaque nasal discharge is part of the natural course of common cold and is not an indication to use antimicrobial therapy unless the symptoms persist longer than 10 to 14 days without improvement,¹⁴ but this message does not appear to be reaching patients. The use of antibiotics for viral illness and for uncomplicated URIs will lead to resistance and is discouraged by researchers and infectious disease experts.¹⁵

Our study provides evidence that even educated individuals may not recognize common URI presentations and appropriate treatment. A significant portion of this sample was unable to link their stated beliefs about colds with symptom scenarios consistent with URIs. In this sample, demographic variables did not alter the likelihood of using antibiotics, indicating the broad-based nature of the misconceptions. Our findings are consistent with previous studies showing that patients in family practice, internal medicine clinics, or emergency department waiting rooms do not recognize symptoms of the common presentation of URIs.¹⁰ The same study hypothesized that education about normal presentation could reduce visits to the physician and the misuse of antibiotics. Another recent study showed that members of ethnically diverse communities believe in the effectiveness of antibiotics for treatment of colds and often obtain them without a prescription.¹²

In our study, women reported a greater likelihood of care seeking across all scenarios, a finding consistent with the literature.16 Another variable predictive of seeking care was college year. Freshmen and sophomores were less likely to seek care than students in later periods of their education. This finding may have been due to the relationship between health care use and increasing age.

Our study emphasizes the need for patient education to: (1) minimize the use of health services or self-limiting illnesses; (2) decrease the inappropriate use of antibiotics; and (3) promote increased selectivity in patients’ choices of nonprescription remedies (eg, promote the use of evidence-based remedies). The central role of the physician as educator and opinion leader in this process should not be overlooked. From previous research it remains unclear whether physicians really believe in the usefulness of antibiotics in treating URIs or whether they are responding to patient expectations. However, some findings indicate a continued need for physician education about the natural history of URIs. For example, Schwartz and colleagues¹¹ surveyed family physicians and pediatricians in Virginia with a questionnaire describing case scenarios that involve a single-day history of greenish-yellowish discharge and low-grade fever. A majority of the physicians surveyed endorsed prescribing antibiotics. Another study revealed that physicians confronted with colored nasal discharge, lack of response to decongestants, and fever (among other symptoms) diagnosed sinusitis and prescribed antibiotics in more than 98% of cases and prescribed the same for more than 13% of patients with URIs.¹⁷ In a 1999 study, pediatricians and family physicians in Georgia reported prescribing practices that were inconsistent with published recommendations for judicious antibiotic use in the treatment of URIs, again overprescribing anitibiotics.¹⁸ Given such study results can we blame our patients’ for their enduring confidence in antibiotic effectiveness? Targeted educational information provided by trusted physicians and a refusal by physicians to prescribe unnecessary antibiotics are vital components in influencing our patients’ beliefs. Although some patients will obtain antibiotics without prescriptions,¹² we can have a significant impact on the beliefs encouraging these practices. The ultimate responsibility for changing patient attitudes and prescribing habits of physicians remains with us.

Limitations

Participants in this self-report survey were volunteers, which may introduce a selection bias. Also, because current college students composed the study population, the ability to extrapolate these findings to other young adults, college-educated older adults, or adults in general is unknown.

Conclusions

Our results show that health care education is needed among college students, especially in the area of symptoms and treatments of common illnesses. Educational interventions for patients and physicians concerning the presentation of the common cold and the appropriate use of antibiotics should be the focus of continuing research. This may result in more medical resources and less-resistant organisms.

Infections of the upper respiratory tract account for some of the most common acute illnesses seen in primary care settings. The term “upper respiratory infection” (URI) covers any infectious disease process that involves the respiratory system, starting with the nose and ending just before the lungs. Our study dealt exclusively with the common cold.

Most patients with typical URI syndromes can be treated symptomatically. Although antimicrobial therapy is indicated in the presence of bacterial infection, it is believed that most cases are viral in nature.¹ However, of those patients who seek a physician’s care for colds and bronchitis, 50% to 70% receive an antibiotic prescription.² Ten percent of all antibiotics prescribed are for the common cold and other URIs.² The percentage is even higher in the pediatric population where antibiotic prescriptions for colds, URIs, and bronchitis (ie, conditions not affected by antibiotics) accounted for more than 20% of all antibiotics prescribed to US children in 1992.³

Prescribing antibiotics for URIs does not improve patient outcome, and this practice does not benefit physicians by reducing return visits or increasing patient satisfaction.^4,5 It is also not a cost-effective strategy. Evidence from a Medicaid population suggests that the antibiotics used to treat colds account for 23% of the total cost of managing URIs and add more than $11 to the cost of managing every URI episode.⁶ Nevertheless, a 1996 study conducted on a Medicaid population concluded that a majority of individuals receiving medical care for the common cold are still given prescriptions for unnecessary antibiotics.⁷

Clinicians often report that they are motivated to prescribe antibiotics by patient expectations.⁸ For example, parents frequently have misconceptions about which illnesses warrant antibiotic therapy leading them to request these drugs for their children.⁹ A survey conducted in Kentucky showed that when patients do not recognize the normal presentation of a URI or understand the effectiveness of antibiotics, inappropriate use and expectations may arise.¹⁰ Also questionable is whether physicians are able to accurately identify situations for which antibiotics are appropriate. Even without the influence of misinformed patients, physicians may be prescribing antibiotics inappropriately because of misdiagnosis.¹¹

The purpose of our study was to determine what a select segment of the population (undergraduate college students) knows about URIs and the perceptions of antibiotic therapy held by members of that segment. The information provided can contribute to our understanding of what types of interventions are required to change patients’ perceptions about the appropriateness of antibiotic therapy.

Methods

Participants (n=425) were students aged 18 years and older on 3 college campuses in Louisiana and Indiana. Two of the colleges were public institutions; the third was private. Data were collected in November 1999 in accordance with human subject guidelines after approval by the appropriate institutional review boards. Research assistants distributed the surveys in public areas on each of the campuses.

Survey Instrument

Participants completed a self-report survey primarily composed of 3 symptom scenarios. Two of these scenarios were employed in previous studies.^10,12 The scenarios represented variations in presentation of an uncomplicated URI along 3 dimensions: duration of symptoms, color of nasal discharge, and the presence of a low-grade fever. The scenarios were: (1) “You have had an illness for 5 days with the following symptoms: sore throat, cough, and runny nose with clear nasal discharge”; (2) “You have had an illness for 5 days with the following symptoms: sore throat, cough, and runny nose with discolored (yellow, green, brown) discharge”; and (3) “You have had an illness for 3 days with the following symptoms: sore throat, cough, runny nose with clear discharge, and low grade fever (less than 101ÞF).”

Following each scenario’s presentation, a participant was asked to indicate on a 5-point Likert-type scale (1=very likely; 5=very unlikely) how likely he or she was to seek care from a physician for the illness and the likelihood of using several treatment modalities (eg, antibiotics, antihistamines, pain relievers, vitamin C) for the presented condition.

Participants were also queried about a variety of demographic variables, current and past smoking status, and their belief in the effectiveness of antibiotics against the common cold. Finally, participants were asked to indicate whether they would see a physician if they had a cold (no specific symptoms were provided to define “cold”). The questionnaire was designed for self-administration and required less than 5 minutes to complete.

Analysis

Likert-based responses on the questionnaire were dichotomized by combining “very likely” and “somewhat likely” into 1 group and “neutral,” “somewhat unlikely,” and “very unlikely” responses into another. These responses and the yes or no responses were analyzed using chi-square tests.