The Quality of Physician-Patient Relationships

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