Original Research

Evidence-based medicine (EBM) involves decision-making based on the systematic identification and critical appraisal of research evidence in combination with clinical expertise and patient values.¹ Two important EBM tools are systematic reviews and an activity known as systematic literature surveillance.

Surveillance complements commonly used resources. Systematic reviews answer a precisely defined question using explicit methods to search for, select, evaluate, and synthesize available evidence. Though extremely valuable, new systematic reviews cannot be produced at a rate that keeps pace with new research information.²

Systematic literature surveillance, by contrast, starts with the evidence and uses explicit, protocol-based methods to select, evaluate, and synthesize new research information. It is an efficient way to find answers to numerous clinical questions, and thus complements systematic reviews for supporting point-of-care clinical references. Both tools should be an indispensable part of supporting clinical practice.

It can dramatically change knowledge. Imagine being faced with a patient who has a clinically significant head injury and not having immediate specialty backup. Steroid administration has been promoted to reduce cerebral edema. You search the Cochrane Library and find a systematic review of 19 randomized trials with 2295 patients. The review concludes that evidence is insufficient to rule out moderate benefits or moderate harms.³ A source complementing systematic reviews with systematic literature surveillance would include a more recent randomized trial with 10,008 patients showing that steroids significantly increase mortality at 2 weeks.⁴

And it’s efficient. Because each article can be identified and evaluated “once”—rather than repeatedly for separate questions posed in systematic reviews—systematic literature surveillance is a more efficient means for answering a large number of questions. It may be used for clinician alerting/newsletter services^5,6 or for updating knowledge syntheses in a clinical reference.⁷

To find the best available evidence during clinical practice, the evidence-based clinician should use references that synthesize the results of systematic literature surveillance and systematic reviews.

CORRESPONDENCE
Brian S. Alper, MD, MSPH, DynaMed, 3610 Buttonwood Drive, Suite 200, Columbia, MO 65201. E-mail: [email protected]

Evidence-based medicine (EBM) involves decision-making based on the systematic identification and critical appraisal of research evidence in combination with clinical expertise and patient values.¹ Two important EBM tools are systematic reviews and an activity known as systematic literature surveillance.

Surveillance complements commonly used resources. Systematic reviews answer a precisely defined question using explicit methods to search for, select, evaluate, and synthesize available evidence. Though extremely valuable, new systematic reviews cannot be produced at a rate that keeps pace with new research information.²

Systematic literature surveillance, by contrast, starts with the evidence and uses explicit, protocol-based methods to select, evaluate, and synthesize new research information. It is an efficient way to find answers to numerous clinical questions, and thus complements systematic reviews for supporting point-of-care clinical references. Both tools should be an indispensable part of supporting clinical practice.

It can dramatically change knowledge. Imagine being faced with a patient who has a clinically significant head injury and not having immediate specialty backup. Steroid administration has been promoted to reduce cerebral edema. You search the Cochrane Library and find a systematic review of 19 randomized trials with 2295 patients. The review concludes that evidence is insufficient to rule out moderate benefits or moderate harms.³ A source complementing systematic reviews with systematic literature surveillance would include a more recent randomized trial with 10,008 patients showing that steroids significantly increase mortality at 2 weeks.⁴

And it’s efficient. Because each article can be identified and evaluated “once”—rather than repeatedly for separate questions posed in systematic reviews—systematic literature surveillance is a more efficient means for answering a large number of questions. It may be used for clinician alerting/newsletter services^5,6 or for updating knowledge syntheses in a clinical reference.⁷

To find the best available evidence during clinical practice, the evidence-based clinician should use references that synthesize the results of systematic literature surveillance and systematic reviews.

CORRESPONDENCE
Brian S. Alper, MD, MSPH, DynaMed, 3610 Buttonwood Drive, Suite 200, Columbia, MO 65201. E-mail: [email protected]

Evidence-based medicine (EBM) involves decision-making based on the systematic identification and critical appraisal of research evidence in combination with clinical expertise and patient values.¹ Two important EBM tools are systematic reviews and an activity known as systematic literature surveillance.

Surveillance complements commonly used resources. Systematic reviews answer a precisely defined question using explicit methods to search for, select, evaluate, and synthesize available evidence. Though extremely valuable, new systematic reviews cannot be produced at a rate that keeps pace with new research information.²

Systematic literature surveillance, by contrast, starts with the evidence and uses explicit, protocol-based methods to select, evaluate, and synthesize new research information. It is an efficient way to find answers to numerous clinical questions, and thus complements systematic reviews for supporting point-of-care clinical references. Both tools should be an indispensable part of supporting clinical practice.

It can dramatically change knowledge. Imagine being faced with a patient who has a clinically significant head injury and not having immediate specialty backup. Steroid administration has been promoted to reduce cerebral edema. You search the Cochrane Library and find a systematic review of 19 randomized trials with 2295 patients. The review concludes that evidence is insufficient to rule out moderate benefits or moderate harms.³ A source complementing systematic reviews with systematic literature surveillance would include a more recent randomized trial with 10,008 patients showing that steroids significantly increase mortality at 2 weeks.⁴

And it’s efficient. Because each article can be identified and evaluated “once”—rather than repeatedly for separate questions posed in systematic reviews—systematic literature surveillance is a more efficient means for answering a large number of questions. It may be used for clinician alerting/newsletter services^5,6 or for updating knowledge syntheses in a clinical reference.⁷

To find the best available evidence during clinical practice, the evidence-based clinician should use references that synthesize the results of systematic literature surveillance and systematic reviews.

CORRESPONDENCE
Brian S. Alper, MD, MSPH, DynaMed, 3610 Buttonwood Drive, Suite 200, Columbia, MO 65201. E-mail: [email protected]

Levels of evidence are assigned to studies based on the quality of their design, validity, and applicability to patient care. The Agency for Health Care Quality and Research (AHRQ) has proposed that any system assigning levels of evidence should incorporate quality, quantity, and consistency of the evidence.

Leading family medicine journals have adopted a uniform grading system known as the Strength of Recommendation Taxonomy¹ (SORT), which includes these key elements and 3 levels of evidence. SORT is one among several different methods of grading levels of evidence that make use of similar principles. SORT’s primary advantage is its simplicity.

The randomized controlled trial (RCT) is the most rigorous study design. According to SORT, RCTs that deal with patient-oriented outcomes and include concealment, double-blinding, intention-to-treat analysis, and complete follow-up (and meta-analyses or systematic reviews of such randomized trials) provide Level 1 evidence. Observational studies, such as cohort and case-control studies (and systematic reviews that include them), are less rigorous in their design. They are assigned a Level of Evidence of 2. Level 3 evidence, the lowest level, is assigned to consensus guidelines, expert opinion, usual practice, and so forth, or to studies that look at intermediate or disease-ori-ented outcomes.

How it applies to recent findings

Although the Nurses Health Study,² a large cohort trial involving nearly 88,000 women, and other observational studies (SORT LOE: 2) suggested a cardiovascular benefit from vitamin E; the Finnish Alpha-Tocopherol, Beta-Carotene Cancer Prevention study,³ a well-designed RCT (LOE: 1), proved the opposite. A recently published Italian study⁴ provided Level 3 evidence, demonstrating that vitamin E prevented an oxidation-induced reduction in coronary blood flow. Therefore, based on the highest level of evidence available, vitamin E to prevent cardiovascular disease is not recommended.

Levels of evidence can make it easier for busy physicians to apply the results of clinical research to their practice and to incorporate evidence-based medicine into patient care.

Levels of evidence are assigned to studies based on the quality of their design, validity, and applicability to patient care. The Agency for Health Care Quality and Research (AHRQ) has proposed that any system assigning levels of evidence should incorporate quality, quantity, and consistency of the evidence.

Leading family medicine journals have adopted a uniform grading system known as the Strength of Recommendation Taxonomy¹ (SORT), which includes these key elements and 3 levels of evidence. SORT is one among several different methods of grading levels of evidence that make use of similar principles. SORT’s primary advantage is its simplicity.

The randomized controlled trial (RCT) is the most rigorous study design. According to SORT, RCTs that deal with patient-oriented outcomes and include concealment, double-blinding, intention-to-treat analysis, and complete follow-up (and meta-analyses or systematic reviews of such randomized trials) provide Level 1 evidence. Observational studies, such as cohort and case-control studies (and systematic reviews that include them), are less rigorous in their design. They are assigned a Level of Evidence of 2. Level 3 evidence, the lowest level, is assigned to consensus guidelines, expert opinion, usual practice, and so forth, or to studies that look at intermediate or disease-ori-ented outcomes.

How it applies to recent findings

Although the Nurses Health Study,² a large cohort trial involving nearly 88,000 women, and other observational studies (SORT LOE: 2) suggested a cardiovascular benefit from vitamin E; the Finnish Alpha-Tocopherol, Beta-Carotene Cancer Prevention study,³ a well-designed RCT (LOE: 1), proved the opposite. A recently published Italian study⁴ provided Level 3 evidence, demonstrating that vitamin E prevented an oxidation-induced reduction in coronary blood flow. Therefore, based on the highest level of evidence available, vitamin E to prevent cardiovascular disease is not recommended.

Levels of evidence can make it easier for busy physicians to apply the results of clinical research to their practice and to incorporate evidence-based medicine into patient care.

Levels of evidence are assigned to studies based on the quality of their design, validity, and applicability to patient care. The Agency for Health Care Quality and Research (AHRQ) has proposed that any system assigning levels of evidence should incorporate quality, quantity, and consistency of the evidence.

Leading family medicine journals have adopted a uniform grading system known as the Strength of Recommendation Taxonomy¹ (SORT), which includes these key elements and 3 levels of evidence. SORT is one among several different methods of grading levels of evidence that make use of similar principles. SORT’s primary advantage is its simplicity.

The randomized controlled trial (RCT) is the most rigorous study design. According to SORT, RCTs that deal with patient-oriented outcomes and include concealment, double-blinding, intention-to-treat analysis, and complete follow-up (and meta-analyses or systematic reviews of such randomized trials) provide Level 1 evidence. Observational studies, such as cohort and case-control studies (and systematic reviews that include them), are less rigorous in their design. They are assigned a Level of Evidence of 2. Level 3 evidence, the lowest level, is assigned to consensus guidelines, expert opinion, usual practice, and so forth, or to studies that look at intermediate or disease-ori-ented outcomes.

How it applies to recent findings

Although the Nurses Health Study,² a large cohort trial involving nearly 88,000 women, and other observational studies (SORT LOE: 2) suggested a cardiovascular benefit from vitamin E; the Finnish Alpha-Tocopherol, Beta-Carotene Cancer Prevention study,³ a well-designed RCT (LOE: 1), proved the opposite. A recently published Italian study⁴ provided Level 3 evidence, demonstrating that vitamin E prevented an oxidation-induced reduction in coronary blood flow. Therefore, based on the highest level of evidence available, vitamin E to prevent cardiovascular disease is not recommended.

Levels of evidence can make it easier for busy physicians to apply the results of clinical research to their practice and to incorporate evidence-based medicine into patient care.

It is widely perceived and documented that both of the following are valuable attributes of treatment relationships: 1) patients’ trust in physicians and in the medical profession,¹ and 2) patients’ active involvement in treatment seeking, decision making, and adherence.^2-8 Both patient trust and active patient involvement are desirable in their own right and because they are associated with improved health outcomes. Paradoxically, however, it might be thought that these 2 attributes are in sharp conflict.

Patient trust might be more consistent with a deferential style of patient-physician interaction in which patients are passive, in contrast to assertive patient questioning or limitation of physician authority which might be indicative of patient distrust. If so, then pursuing active patient involvement might lead to lower trust, or promoting trust might lead to more passive patients, either of which might compromise optimal treatment relationships and health outcomes. At a minimum, it is a conceptual puzzle how these 2 views of desirable attributes of medical relationships can coexist without each taking account of the other view. There certainly are skeptics of patient trust who warn that, contrary to conventional wisdom, too much trust might be negative and that patients, for their own good, should be encouraged to trust less to avoid the dangers of paternalistic medicine,⁹ especially in managed care settings. ^10,11

Numerous studies examine either patient trust or patients’ roles in seeking care, level of participation in medical decisions, and adherence to treatment. However, few studies examine both halves of these 2 sets of attitudes and behaviors, and none examine the full cluster. None of the leading studies of patients’ attitudes toward seeking care or participating in medical decisions include measures of trust. A few studies of care-seeking also examine attitudes similar to or overlapping with trust, such as confidentiality or competency,^12,13 but none of these use the trust concept itself or any of the validated instruments that measure trust. Among studies of trust, some have explored trust’s connections with adherence,¹⁴ preferred role in decision making,¹⁵ patients’ requests for specific services,¹⁶ or willingness to seek care,^17,18 but few studies simultaneously explore these measures,^17,18 and only 1 study examines how they relate to trust in the medical profession, rather than trust in a specific physician.¹⁸ In general, these studies report that trust in a specific physician is associated with greater adherence to treatment recommendations, more willingness to seek care, and giving physicians more control over medical decisions.

Absent in this literature is any concurrent examination of how different types of trust relate to various aspects of patients’ views about their involvement in medical care. To advance understanding of these important connections, this study was designed to investigate whether patients’ trust in their primary physician and in the medical profession are related to their attitudes toward seeking care, preferred roles in medical decision making, and reported adherence to treatment recommendations.

Methods

How the sample was collected

A random national sample of 2637 households was selected in 1999 from a proprietary database of working residential telephone exchanges in the continental United States. A minimum of 15 attempts was made to reach those numbers that were not answered. Respondent selection within eligible households was done using the next birthday method.¹⁹

Inclusion criteria for the study included being at least 21 years of age and the ability to speak and understand English. Because this survey was part of a larger study of recent experiences with physicians and health insurers,^20-22 respondents were further screened to select only those with some type of public or private health care coverage and those who had been to a physician or other health professional at least twice during the past 2 years. Health care coverage broadly includes any type of public or private insurance, or access to other type of government or indigent care program or facility.

Contacts with 2172 potentially eligible individuals resulted in the following dispositions: 1117 (51.4%) were interviewed; 571 (26.3%) refused; 484 (22.3%) were unable to participate (not home, ill, non-English-speaking). To reduce respondent burden, the sample was randomly divided and only half (N=553) were asked the battery of questions about trust in the medical profession and following physicians’ recommendations. (The other half were asked other questions unrelated to this analysis. There were no statistically significant differences between the 2 halves on age, race, gender, health status, education, or income.)

Due to the particular selection criteria and survey method, this random sample is not fully representative of national norms. There is a somewhat greater representation of whites (84.3%) and females (68.9%) because they are more likely to have insurance and have gone to a physician recently. Otherwise, the majority of our sample was between 30 to 60 years old (58%) and college educated (58%), with a median income of about $40,000. Thus, the sample composition is sufficient to analyze most major demographic and socioeconomic groups.

Telephone interviews lasted approximately 25 minutes and were conducted by trained interviewers at the Survey Research Center of the University of South Carolina using computer assisted telephone interviewing. Verbal informed consent was obtained at the start of the telephone interviews and the study protocols were reviewed and approved by Wake Forest University Medical Center’s Institutional Review Board.

Measures

Attitudes toward and preferred roles in medical care were measured by the 5 questions listed in TABLE 1 . Data were also collected on a range of topics, including demographic characteristics, trust in the subject’s regular physician and the medical profession, satisfaction with care, and physical and mental health. Trust in the specific physician and in the medical profession were measured by 5-item scales, whose validation and properties are reported elsewhere (Cronbach’s alpha ≥0.77 for all).²³ Satisfaction with medical care was measured with a previously validated 12-item scale asking about health care received from all sources over the past few years.²⁴

TABLE 1
Items assessing patients’ roles in medical relationships

Statistical analysis

To assess patients’ involvement in medical care, frequencies of response categories were calculated for the general study population, as well as key subgroups— men, African Americans, and the elderly (age ≥65 years). Chi-square tests were used to determine whether views varied by sex, race, and the elderly. Spearman correlation was used to assess the relationship of the 5 views to each other, as well as to test whether views varied by trust in the physician and the medical profession, satisfaction with care, age, education, income, physical and mental health, number of visits and years with physician, and past dispute with a physician. Finally, multivariate regressions were performed to determine the most significant predictors of involvement in medical care. In all analyses, a P-value of .05 was considered statistically significant. However, in recognition of the multiple testing environment, consistency over the 5 items was considered as well.

Results

TABLE 2 summarizes the response frequencies for the 5 questions and TABLE 3 shows the correlations of the 5 items with each other. All items are significantly correlated at level 0.005 or below. As one might expect, the questions on relying on physicians’ judgment and seeking professional help are most correlated, as are physician/patient control and roles.

TABLE 2
Views of the general population and subgroups about involvement in medical care

TABLE 3
Correlation between views on involvement in medical care

Trust and satisfaction

As shown in TABLE 4 , trust in a specific physician is associated only with always following recommendations. Trust in the medical profession and satisfaction with care are both highly associated with all 5 views towards involvement in medical care. More trust in the medical profession, and greater satisfaction with care, are associated with following recommendations, relying on the judgment of physicians, seeking professional medical help, and granting increased control and decision making to the physician.

TABLE 4
Correlation between trust and satisfaction and medical views

Patient demographics

A number of patient demographic characteristics are significantly associated with patients’ involvement in medical care. Women are less likely to want to rely on the judgment of physicians and are more likely to want to share control and make joint decisions. There is also an extremely strong effect of age. The elderly (age ≥65 years) are more compliant, deferential, and passive under each of our measures. Interestingly, race has no significant effects on these measures of involvement.

Patients with less education are more likely to want to follow physicians’ recommendations, rely on their physician’s judgment, and seek professional medical help (P=.0015, <0.001, 0.027, respectively). They are also more likely to give the physician more control and authority to make decisions for them (P<.001 for both). People with less income (on a 9-point scale) are also more likely to want to follow physicians’ recommendations, rely on their judgment, and give the physician more control and authority (P<.007 for all). Patients in worse physical health (on a 5-point scale) are more likely to want to rely on the judgment of physicians and to seek professional medical help (correlation=0.11; P=.014, .017). There is no significant association between mental health and any of these views.

Patients’ relationship with physicians

Several aspects of the patient/physician relationship were significantly associated with patients’ involvement in medical care. Not surprisingly, willingness to seek professional medical help increases with number of visits to the physician (P=.039), and granting increased control and decision making to the physician increases with continuity of care, as measured by number of years with the physician (P=.001, .037). Additionally, having had a past dispute with a physician was significantly related to patients’ involvement in medical care. Patients with a past dispute were less likely to always follow physicians’ recommendations, rely on the judgment of physicians, or seek professional medical help. They were also more likely to take control and make medical decisions themselves.

Multivariate regression models

Multivariate regression models were used to determine the most significant predictors of patients’ involvement in medical care. Multivariate linear regression was used for the outcomes of following physicians’ recommendations and physician/patient control and roles. Logistic regression was used for the binary outcome variables— rely on physician’s judgment and seek professional help.

The predictor variables used in the model were sex, age, education, physical health, number of physician visits, number of years with physician, past dispute, trust, and satisfaction with care. Income and elderly were not used as they are correlated with education and age, respectively, which were more predictive of patient involvement. Significant predictors are shown in TABLE 5 . Notably, the patient’s trust in their specific physician did not predict any of these views, and satisfaction with care predicted only following physicians’ recommendations. Demographic, health status, and other variables were also nonsignificant in many or most regressions. Only trust in the medical profession predicted all 5 views.

TABLE 5
P-values for significant predictors in the multivariate analysis

Discussion

What predicts involvement in medical care?

Trust a key predictor. Trust in the medical profession is a key predictor of people’s involvement in medical care. It is a significant predictor of self reports of: 1) following physicians’ treatment recommendations, willingness to 2) seek care and to 3) rely on physicians’ judgment, and wanting to 4) give physicians more control and 5) let them make decisions for patients. Most other relationship factors (trust in a specific physician, length and continuity of treatment relationship, past dispute with physician) predict fewer of these variables and are no longer significant in multivariate analyses. In bivariate analyses, satisfaction with care is also a consistent predictor of all 5 measures of involvement in medical care, but it usually no longer remains significant in regression analyses once trust in the medical profession is added to the models.

Demographics not significant. Similarly, demographic characteristics such as age, sex, or education often are no longer significant in regression models that control for trust in the medical profession, suggesting that, in some instances, these factors may influence views about involvement through their effect on trust in the medical profession.

Patient roles and control of medical care.

In general, increased trust in the medical profession is associated with a more deferential patient role in medical relationships. Higher trust is associated with greater willingness to give control to physicians and allow them to make decisions for the patient. However, other aspects of patient involvement are not as easily classified as deferential in the sense of the term that connotes a passive patient role. Trust in the medical profession is also associated with greater willingness to seek care and to comply with treatment recommendations.

To the extent that trust is associated with deferential or passive patient roles, it is notable that this association exists primarily only with the measures that refer to the medical system as a whole (including satisfaction with care generally), and not with the measures that are specific to particular physicians. The latter include trust in the patient’s personal physician, past disputes with that physician, number of visits with that physician, and length of relationship with that physician. Each of these is much less predictive of patients’ involvement in medical care than is trust in the medical profession or satisfaction with care generally.

What does this say about patient relationships in general?

This suggests, consistent with prior research,^18,22 that patients’ views about particular physicians are substantially (but not entirely) independent from their views about the medical system in general. This finding is also consistent with prior explanations that the nature of interpersonal physician trust evolves over the course of an ongoing treatment relationship to accommodate both more active and more passive patient roles.²⁵ In fact, in our sample, trust in the specific physician was generally quite high regardless of preferred involvement in medical care.

Throughout all of these associations, the direction of possible causality is not established by this study. A person’s attitudes about involvement in medical care may be determined by their trust in the medical profession, or their trust may be determined by the types of involvement they have had in medical care. Most likely, there is a cyclical relationship between the 2.³⁵

These findings provide reassurance that promoting trust will not likely cause a reversion to excessively paternalistic medical relationships. Trust in specific physicians is only weakly related to patients’ views about active vs passive roles. Trust in the medical profession is much more pertinent to these views, but that type of trust is distinct from trust in specific physicians. Moreover, trust in the medical profession is consistent with some desirable forms of patient involvement.

These findings also suggest that trust is able to form in relationships where patients are either active or passive. Assertive patient involvement should not be seen as indicative of distrust. Instead, other studies suggest that trust is promoted by communicating effectively with patients such as by listening carefully, answering questions clearly, giving them as much information as they want, and involving them in medical decisions.^26,27

Attitudes towards involvement in medical care. Apart from the relationship to trust, these findings shed important light on general attitudes toward involvement in medical care. Overall, the majority of people report following their physicians’ recommendations and think that it is better to rely on the expert judgment of physicians and seek professional help. Most people think the physician and patient should share control equally and make decisions together. More patients give control to their physicians than to themselves. These findings are consistent with previous studies.^8,28-36

Also, consistent with prior studies, we found that younger and more educated patients prefer more assertive roles, as do women. We found no racial differences, but this may be due to the under representation of minorities in our sample. Finally, the small amount of variability explained in our models indicates that while we have identified predictive factors, many other factors affect people’s views towards medical care. Physicians need to be aware that patient desires for participation vary, and communication about such desires is necessary during visits.

Limitations of this study

Several study limitations should be noted. First, our measures of involvement in medical care are self-reported and do not necessarily reflect patients’ actual behaviors. However, most measures have been previously validated to some extent. Second, the selection criteria for the study do not allow for generalization to populations that less routinely seek care or are uninsured. Lastly, this is an exploratory study that was not driven by specific hypotheses derived from prior studies or firm theory. The empirical study of trust and its connection with other attitudes and relationship characteristics is still in its infancy,³⁷ which calls for more exploratory approaches that identify areas of focus for future research. The connection between trust and patients’ involvement in medical care is one such area deserving further study.

Acknowledgments

Research supported by the Robert Wood Johnson Foundation, the National Eye Institute (EY012443-02), and the National Institute on Aging (AG015248-03).

CORRESPONDING AUTHOR
Felicia Trachtenberg, PhD, New England Research Institutes, 9 Galen Street, Watertown, MA 02472. E-mail: [email protected]

It is widely perceived and documented that both of the following are valuable attributes of treatment relationships: 1) patients’ trust in physicians and in the medical profession,¹ and 2) patients’ active involvement in treatment seeking, decision making, and adherence.^2-8 Both patient trust and active patient involvement are desirable in their own right and because they are associated with improved health outcomes. Paradoxically, however, it might be thought that these 2 attributes are in sharp conflict.

Patient trust might be more consistent with a deferential style of patient-physician interaction in which patients are passive, in contrast to assertive patient questioning or limitation of physician authority which might be indicative of patient distrust. If so, then pursuing active patient involvement might lead to lower trust, or promoting trust might lead to more passive patients, either of which might compromise optimal treatment relationships and health outcomes. At a minimum, it is a conceptual puzzle how these 2 views of desirable attributes of medical relationships can coexist without each taking account of the other view. There certainly are skeptics of patient trust who warn that, contrary to conventional wisdom, too much trust might be negative and that patients, for their own good, should be encouraged to trust less to avoid the dangers of paternalistic medicine,⁹ especially in managed care settings. ^10,11

Numerous studies examine either patient trust or patients’ roles in seeking care, level of participation in medical decisions, and adherence to treatment. However, few studies examine both halves of these 2 sets of attitudes and behaviors, and none examine the full cluster. None of the leading studies of patients’ attitudes toward seeking care or participating in medical decisions include measures of trust. A few studies of care-seeking also examine attitudes similar to or overlapping with trust, such as confidentiality or competency,^12,13 but none of these use the trust concept itself or any of the validated instruments that measure trust. Among studies of trust, some have explored trust’s connections with adherence,¹⁴ preferred role in decision making,¹⁵ patients’ requests for specific services,¹⁶ or willingness to seek care,^17,18 but few studies simultaneously explore these measures,^17,18 and only 1 study examines how they relate to trust in the medical profession, rather than trust in a specific physician.¹⁸ In general, these studies report that trust in a specific physician is associated with greater adherence to treatment recommendations, more willingness to seek care, and giving physicians more control over medical decisions.

Absent in this literature is any concurrent examination of how different types of trust relate to various aspects of patients’ views about their involvement in medical care. To advance understanding of these important connections, this study was designed to investigate whether patients’ trust in their primary physician and in the medical profession are related to their attitudes toward seeking care, preferred roles in medical decision making, and reported adherence to treatment recommendations.

Methods

How the sample was collected

A random national sample of 2637 households was selected in 1999 from a proprietary database of working residential telephone exchanges in the continental United States. A minimum of 15 attempts was made to reach those numbers that were not answered. Respondent selection within eligible households was done using the next birthday method.¹⁹

Inclusion criteria for the study included being at least 21 years of age and the ability to speak and understand English. Because this survey was part of a larger study of recent experiences with physicians and health insurers,^20-22 respondents were further screened to select only those with some type of public or private health care coverage and those who had been to a physician or other health professional at least twice during the past 2 years. Health care coverage broadly includes any type of public or private insurance, or access to other type of government or indigent care program or facility.

Contacts with 2172 potentially eligible individuals resulted in the following dispositions: 1117 (51.4%) were interviewed; 571 (26.3%) refused; 484 (22.3%) were unable to participate (not home, ill, non-English-speaking). To reduce respondent burden, the sample was randomly divided and only half (N=553) were asked the battery of questions about trust in the medical profession and following physicians’ recommendations. (The other half were asked other questions unrelated to this analysis. There were no statistically significant differences between the 2 halves on age, race, gender, health status, education, or income.)

Due to the particular selection criteria and survey method, this random sample is not fully representative of national norms. There is a somewhat greater representation of whites (84.3%) and females (68.9%) because they are more likely to have insurance and have gone to a physician recently. Otherwise, the majority of our sample was between 30 to 60 years old (58%) and college educated (58%), with a median income of about $40,000. Thus, the sample composition is sufficient to analyze most major demographic and socioeconomic groups.

Telephone interviews lasted approximately 25 minutes and were conducted by trained interviewers at the Survey Research Center of the University of South Carolina using computer assisted telephone interviewing. Verbal informed consent was obtained at the start of the telephone interviews and the study protocols were reviewed and approved by Wake Forest University Medical Center’s Institutional Review Board.

Measures

Attitudes toward and preferred roles in medical care were measured by the 5 questions listed in TABLE 1 . Data were also collected on a range of topics, including demographic characteristics, trust in the subject’s regular physician and the medical profession, satisfaction with care, and physical and mental health. Trust in the specific physician and in the medical profession were measured by 5-item scales, whose validation and properties are reported elsewhere (Cronbach’s alpha ≥0.77 for all).²³ Satisfaction with medical care was measured with a previously validated 12-item scale asking about health care received from all sources over the past few years.²⁴

TABLE 1
Items assessing patients’ roles in medical relationships

Statistical analysis

To assess patients’ involvement in medical care, frequencies of response categories were calculated for the general study population, as well as key subgroups— men, African Americans, and the elderly (age ≥65 years). Chi-square tests were used to determine whether views varied by sex, race, and the elderly. Spearman correlation was used to assess the relationship of the 5 views to each other, as well as to test whether views varied by trust in the physician and the medical profession, satisfaction with care, age, education, income, physical and mental health, number of visits and years with physician, and past dispute with a physician. Finally, multivariate regressions were performed to determine the most significant predictors of involvement in medical care. In all analyses, a P-value of .05 was considered statistically significant. However, in recognition of the multiple testing environment, consistency over the 5 items was considered as well.

Results

TABLE 2 summarizes the response frequencies for the 5 questions and TABLE 3 shows the correlations of the 5 items with each other. All items are significantly correlated at level 0.005 or below. As one might expect, the questions on relying on physicians’ judgment and seeking professional help are most correlated, as are physician/patient control and roles.

TABLE 2
Views of the general population and subgroups about involvement in medical care

TABLE 3
Correlation between views on involvement in medical care

Trust and satisfaction

As shown in TABLE 4 , trust in a specific physician is associated only with always following recommendations. Trust in the medical profession and satisfaction with care are both highly associated with all 5 views towards involvement in medical care. More trust in the medical profession, and greater satisfaction with care, are associated with following recommendations, relying on the judgment of physicians, seeking professional medical help, and granting increased control and decision making to the physician.

TABLE 4
Correlation between trust and satisfaction and medical views

Patient demographics

A number of patient demographic characteristics are significantly associated with patients’ involvement in medical care. Women are less likely to want to rely on the judgment of physicians and are more likely to want to share control and make joint decisions. There is also an extremely strong effect of age. The elderly (age ≥65 years) are more compliant, deferential, and passive under each of our measures. Interestingly, race has no significant effects on these measures of involvement.

Patients with less education are more likely to want to follow physicians’ recommendations, rely on their physician’s judgment, and seek professional medical help (P=.0015, <0.001, 0.027, respectively). They are also more likely to give the physician more control and authority to make decisions for them (P<.001 for both). People with less income (on a 9-point scale) are also more likely to want to follow physicians’ recommendations, rely on their judgment, and give the physician more control and authority (P<.007 for all). Patients in worse physical health (on a 5-point scale) are more likely to want to rely on the judgment of physicians and to seek professional medical help (correlation=0.11; P=.014, .017). There is no significant association between mental health and any of these views.

Patients’ relationship with physicians

Several aspects of the patient/physician relationship were significantly associated with patients’ involvement in medical care. Not surprisingly, willingness to seek professional medical help increases with number of visits to the physician (P=.039), and granting increased control and decision making to the physician increases with continuity of care, as measured by number of years with the physician (P=.001, .037). Additionally, having had a past dispute with a physician was significantly related to patients’ involvement in medical care. Patients with a past dispute were less likely to always follow physicians’ recommendations, rely on the judgment of physicians, or seek professional medical help. They were also more likely to take control and make medical decisions themselves.

Multivariate regression models

Multivariate regression models were used to determine the most significant predictors of patients’ involvement in medical care. Multivariate linear regression was used for the outcomes of following physicians’ recommendations and physician/patient control and roles. Logistic regression was used for the binary outcome variables— rely on physician’s judgment and seek professional help.

The predictor variables used in the model were sex, age, education, physical health, number of physician visits, number of years with physician, past dispute, trust, and satisfaction with care. Income and elderly were not used as they are correlated with education and age, respectively, which were more predictive of patient involvement. Significant predictors are shown in TABLE 5 . Notably, the patient’s trust in their specific physician did not predict any of these views, and satisfaction with care predicted only following physicians’ recommendations. Demographic, health status, and other variables were also nonsignificant in many or most regressions. Only trust in the medical profession predicted all 5 views.

TABLE 5
P-values for significant predictors in the multivariate analysis

Discussion

What predicts involvement in medical care?

Trust a key predictor. Trust in the medical profession is a key predictor of people’s involvement in medical care. It is a significant predictor of self reports of: 1) following physicians’ treatment recommendations, willingness to 2) seek care and to 3) rely on physicians’ judgment, and wanting to 4) give physicians more control and 5) let them make decisions for patients. Most other relationship factors (trust in a specific physician, length and continuity of treatment relationship, past dispute with physician) predict fewer of these variables and are no longer significant in multivariate analyses. In bivariate analyses, satisfaction with care is also a consistent predictor of all 5 measures of involvement in medical care, but it usually no longer remains significant in regression analyses once trust in the medical profession is added to the models.

Demographics not significant. Similarly, demographic characteristics such as age, sex, or education often are no longer significant in regression models that control for trust in the medical profession, suggesting that, in some instances, these factors may influence views about involvement through their effect on trust in the medical profession.

Patient roles and control of medical care.

In general, increased trust in the medical profession is associated with a more deferential patient role in medical relationships. Higher trust is associated with greater willingness to give control to physicians and allow them to make decisions for the patient. However, other aspects of patient involvement are not as easily classified as deferential in the sense of the term that connotes a passive patient role. Trust in the medical profession is also associated with greater willingness to seek care and to comply with treatment recommendations.

To the extent that trust is associated with deferential or passive patient roles, it is notable that this association exists primarily only with the measures that refer to the medical system as a whole (including satisfaction with care generally), and not with the measures that are specific to particular physicians. The latter include trust in the patient’s personal physician, past disputes with that physician, number of visits with that physician, and length of relationship with that physician. Each of these is much less predictive of patients’ involvement in medical care than is trust in the medical profession or satisfaction with care generally.

What does this say about patient relationships in general?

This suggests, consistent with prior research,^18,22 that patients’ views about particular physicians are substantially (but not entirely) independent from their views about the medical system in general. This finding is also consistent with prior explanations that the nature of interpersonal physician trust evolves over the course of an ongoing treatment relationship to accommodate both more active and more passive patient roles.²⁵ In fact, in our sample, trust in the specific physician was generally quite high regardless of preferred involvement in medical care.

Throughout all of these associations, the direction of possible causality is not established by this study. A person’s attitudes about involvement in medical care may be determined by their trust in the medical profession, or their trust may be determined by the types of involvement they have had in medical care. Most likely, there is a cyclical relationship between the 2.³⁵

These findings provide reassurance that promoting trust will not likely cause a reversion to excessively paternalistic medical relationships. Trust in specific physicians is only weakly related to patients’ views about active vs passive roles. Trust in the medical profession is much more pertinent to these views, but that type of trust is distinct from trust in specific physicians. Moreover, trust in the medical profession is consistent with some desirable forms of patient involvement.

These findings also suggest that trust is able to form in relationships where patients are either active or passive. Assertive patient involvement should not be seen as indicative of distrust. Instead, other studies suggest that trust is promoted by communicating effectively with patients such as by listening carefully, answering questions clearly, giving them as much information as they want, and involving them in medical decisions.^26,27

Attitudes towards involvement in medical care. Apart from the relationship to trust, these findings shed important light on general attitudes toward involvement in medical care. Overall, the majority of people report following their physicians’ recommendations and think that it is better to rely on the expert judgment of physicians and seek professional help. Most people think the physician and patient should share control equally and make decisions together. More patients give control to their physicians than to themselves. These findings are consistent with previous studies.^8,28-36

Also, consistent with prior studies, we found that younger and more educated patients prefer more assertive roles, as do women. We found no racial differences, but this may be due to the under representation of minorities in our sample. Finally, the small amount of variability explained in our models indicates that while we have identified predictive factors, many other factors affect people’s views towards medical care. Physicians need to be aware that patient desires for participation vary, and communication about such desires is necessary during visits.

Limitations of this study

Several study limitations should be noted. First, our measures of involvement in medical care are self-reported and do not necessarily reflect patients’ actual behaviors. However, most measures have been previously validated to some extent. Second, the selection criteria for the study do not allow for generalization to populations that less routinely seek care or are uninsured. Lastly, this is an exploratory study that was not driven by specific hypotheses derived from prior studies or firm theory. The empirical study of trust and its connection with other attitudes and relationship characteristics is still in its infancy,³⁷ which calls for more exploratory approaches that identify areas of focus for future research. The connection between trust and patients’ involvement in medical care is one such area deserving further study.

Acknowledgments

Research supported by the Robert Wood Johnson Foundation, the National Eye Institute (EY012443-02), and the National Institute on Aging (AG015248-03).

CORRESPONDING AUTHOR
Felicia Trachtenberg, PhD, New England Research Institutes, 9 Galen Street, Watertown, MA 02472. E-mail: [email protected]

It is widely perceived and documented that both of the following are valuable attributes of treatment relationships: 1) patients’ trust in physicians and in the medical profession,¹ and 2) patients’ active involvement in treatment seeking, decision making, and adherence.^2-8 Both patient trust and active patient involvement are desirable in their own right and because they are associated with improved health outcomes. Paradoxically, however, it might be thought that these 2 attributes are in sharp conflict.

Patient trust might be more consistent with a deferential style of patient-physician interaction in which patients are passive, in contrast to assertive patient questioning or limitation of physician authority which might be indicative of patient distrust. If so, then pursuing active patient involvement might lead to lower trust, or promoting trust might lead to more passive patients, either of which might compromise optimal treatment relationships and health outcomes. At a minimum, it is a conceptual puzzle how these 2 views of desirable attributes of medical relationships can coexist without each taking account of the other view. There certainly are skeptics of patient trust who warn that, contrary to conventional wisdom, too much trust might be negative and that patients, for their own good, should be encouraged to trust less to avoid the dangers of paternalistic medicine,⁹ especially in managed care settings. ^10,11

Numerous studies examine either patient trust or patients’ roles in seeking care, level of participation in medical decisions, and adherence to treatment. However, few studies examine both halves of these 2 sets of attitudes and behaviors, and none examine the full cluster. None of the leading studies of patients’ attitudes toward seeking care or participating in medical decisions include measures of trust. A few studies of care-seeking also examine attitudes similar to or overlapping with trust, such as confidentiality or competency,^12,13 but none of these use the trust concept itself or any of the validated instruments that measure trust. Among studies of trust, some have explored trust’s connections with adherence,¹⁴ preferred role in decision making,¹⁵ patients’ requests for specific services,¹⁶ or willingness to seek care,^17,18 but few studies simultaneously explore these measures,^17,18 and only 1 study examines how they relate to trust in the medical profession, rather than trust in a specific physician.¹⁸ In general, these studies report that trust in a specific physician is associated with greater adherence to treatment recommendations, more willingness to seek care, and giving physicians more control over medical decisions.

Absent in this literature is any concurrent examination of how different types of trust relate to various aspects of patients’ views about their involvement in medical care. To advance understanding of these important connections, this study was designed to investigate whether patients’ trust in their primary physician and in the medical profession are related to their attitudes toward seeking care, preferred roles in medical decision making, and reported adherence to treatment recommendations.

Methods

How the sample was collected

A random national sample of 2637 households was selected in 1999 from a proprietary database of working residential telephone exchanges in the continental United States. A minimum of 15 attempts was made to reach those numbers that were not answered. Respondent selection within eligible households was done using the next birthday method.¹⁹

Inclusion criteria for the study included being at least 21 years of age and the ability to speak and understand English. Because this survey was part of a larger study of recent experiences with physicians and health insurers,^20-22 respondents were further screened to select only those with some type of public or private health care coverage and those who had been to a physician or other health professional at least twice during the past 2 years. Health care coverage broadly includes any type of public or private insurance, or access to other type of government or indigent care program or facility.

Contacts with 2172 potentially eligible individuals resulted in the following dispositions: 1117 (51.4%) were interviewed; 571 (26.3%) refused; 484 (22.3%) were unable to participate (not home, ill, non-English-speaking). To reduce respondent burden, the sample was randomly divided and only half (N=553) were asked the battery of questions about trust in the medical profession and following physicians’ recommendations. (The other half were asked other questions unrelated to this analysis. There were no statistically significant differences between the 2 halves on age, race, gender, health status, education, or income.)

Due to the particular selection criteria and survey method, this random sample is not fully representative of national norms. There is a somewhat greater representation of whites (84.3%) and females (68.9%) because they are more likely to have insurance and have gone to a physician recently. Otherwise, the majority of our sample was between 30 to 60 years old (58%) and college educated (58%), with a median income of about $40,000. Thus, the sample composition is sufficient to analyze most major demographic and socioeconomic groups.

Telephone interviews lasted approximately 25 minutes and were conducted by trained interviewers at the Survey Research Center of the University of South Carolina using computer assisted telephone interviewing. Verbal informed consent was obtained at the start of the telephone interviews and the study protocols were reviewed and approved by Wake Forest University Medical Center’s Institutional Review Board.

Measures

Attitudes toward and preferred roles in medical care were measured by the 5 questions listed in TABLE 1 . Data were also collected on a range of topics, including demographic characteristics, trust in the subject’s regular physician and the medical profession, satisfaction with care, and physical and mental health. Trust in the specific physician and in the medical profession were measured by 5-item scales, whose validation and properties are reported elsewhere (Cronbach’s alpha ≥0.77 for all).²³ Satisfaction with medical care was measured with a previously validated 12-item scale asking about health care received from all sources over the past few years.²⁴

TABLE 1
Items assessing patients’ roles in medical relationships

Statistical analysis

To assess patients’ involvement in medical care, frequencies of response categories were calculated for the general study population, as well as key subgroups— men, African Americans, and the elderly (age ≥65 years). Chi-square tests were used to determine whether views varied by sex, race, and the elderly. Spearman correlation was used to assess the relationship of the 5 views to each other, as well as to test whether views varied by trust in the physician and the medical profession, satisfaction with care, age, education, income, physical and mental health, number of visits and years with physician, and past dispute with a physician. Finally, multivariate regressions were performed to determine the most significant predictors of involvement in medical care. In all analyses, a P-value of .05 was considered statistically significant. However, in recognition of the multiple testing environment, consistency over the 5 items was considered as well.

Results

TABLE 2 summarizes the response frequencies for the 5 questions and TABLE 3 shows the correlations of the 5 items with each other. All items are significantly correlated at level 0.005 or below. As one might expect, the questions on relying on physicians’ judgment and seeking professional help are most correlated, as are physician/patient control and roles.

TABLE 2
Views of the general population and subgroups about involvement in medical care

TABLE 3
Correlation between views on involvement in medical care

Trust and satisfaction

As shown in TABLE 4 , trust in a specific physician is associated only with always following recommendations. Trust in the medical profession and satisfaction with care are both highly associated with all 5 views towards involvement in medical care. More trust in the medical profession, and greater satisfaction with care, are associated with following recommendations, relying on the judgment of physicians, seeking professional medical help, and granting increased control and decision making to the physician.

TABLE 4
Correlation between trust and satisfaction and medical views

Patient demographics

A number of patient demographic characteristics are significantly associated with patients’ involvement in medical care. Women are less likely to want to rely on the judgment of physicians and are more likely to want to share control and make joint decisions. There is also an extremely strong effect of age. The elderly (age ≥65 years) are more compliant, deferential, and passive under each of our measures. Interestingly, race has no significant effects on these measures of involvement.

Patients with less education are more likely to want to follow physicians’ recommendations, rely on their physician’s judgment, and seek professional medical help (P=.0015, <0.001, 0.027, respectively). They are also more likely to give the physician more control and authority to make decisions for them (P<.001 for both). People with less income (on a 9-point scale) are also more likely to want to follow physicians’ recommendations, rely on their judgment, and give the physician more control and authority (P<.007 for all). Patients in worse physical health (on a 5-point scale) are more likely to want to rely on the judgment of physicians and to seek professional medical help (correlation=0.11; P=.014, .017). There is no significant association between mental health and any of these views.

Patients’ relationship with physicians

Several aspects of the patient/physician relationship were significantly associated with patients’ involvement in medical care. Not surprisingly, willingness to seek professional medical help increases with number of visits to the physician (P=.039), and granting increased control and decision making to the physician increases with continuity of care, as measured by number of years with the physician (P=.001, .037). Additionally, having had a past dispute with a physician was significantly related to patients’ involvement in medical care. Patients with a past dispute were less likely to always follow physicians’ recommendations, rely on the judgment of physicians, or seek professional medical help. They were also more likely to take control and make medical decisions themselves.

Multivariate regression models

Multivariate regression models were used to determine the most significant predictors of patients’ involvement in medical care. Multivariate linear regression was used for the outcomes of following physicians’ recommendations and physician/patient control and roles. Logistic regression was used for the binary outcome variables— rely on physician’s judgment and seek professional help.

The predictor variables used in the model were sex, age, education, physical health, number of physician visits, number of years with physician, past dispute, trust, and satisfaction with care. Income and elderly were not used as they are correlated with education and age, respectively, which were more predictive of patient involvement. Significant predictors are shown in TABLE 5 . Notably, the patient’s trust in their specific physician did not predict any of these views, and satisfaction with care predicted only following physicians’ recommendations. Demographic, health status, and other variables were also nonsignificant in many or most regressions. Only trust in the medical profession predicted all 5 views.

TABLE 5
P-values for significant predictors in the multivariate analysis

Discussion

What predicts involvement in medical care?

Trust a key predictor. Trust in the medical profession is a key predictor of people’s involvement in medical care. It is a significant predictor of self reports of: 1) following physicians’ treatment recommendations, willingness to 2) seek care and to 3) rely on physicians’ judgment, and wanting to 4) give physicians more control and 5) let them make decisions for patients. Most other relationship factors (trust in a specific physician, length and continuity of treatment relationship, past dispute with physician) predict fewer of these variables and are no longer significant in multivariate analyses. In bivariate analyses, satisfaction with care is also a consistent predictor of all 5 measures of involvement in medical care, but it usually no longer remains significant in regression analyses once trust in the medical profession is added to the models.

Demographics not significant. Similarly, demographic characteristics such as age, sex, or education often are no longer significant in regression models that control for trust in the medical profession, suggesting that, in some instances, these factors may influence views about involvement through their effect on trust in the medical profession.

Patient roles and control of medical care.

In general, increased trust in the medical profession is associated with a more deferential patient role in medical relationships. Higher trust is associated with greater willingness to give control to physicians and allow them to make decisions for the patient. However, other aspects of patient involvement are not as easily classified as deferential in the sense of the term that connotes a passive patient role. Trust in the medical profession is also associated with greater willingness to seek care and to comply with treatment recommendations.

To the extent that trust is associated with deferential or passive patient roles, it is notable that this association exists primarily only with the measures that refer to the medical system as a whole (including satisfaction with care generally), and not with the measures that are specific to particular physicians. The latter include trust in the patient’s personal physician, past disputes with that physician, number of visits with that physician, and length of relationship with that physician. Each of these is much less predictive of patients’ involvement in medical care than is trust in the medical profession or satisfaction with care generally.

What does this say about patient relationships in general?

This suggests, consistent with prior research,^18,22 that patients’ views about particular physicians are substantially (but not entirely) independent from their views about the medical system in general. This finding is also consistent with prior explanations that the nature of interpersonal physician trust evolves over the course of an ongoing treatment relationship to accommodate both more active and more passive patient roles.²⁵ In fact, in our sample, trust in the specific physician was generally quite high regardless of preferred involvement in medical care.

Throughout all of these associations, the direction of possible causality is not established by this study. A person’s attitudes about involvement in medical care may be determined by their trust in the medical profession, or their trust may be determined by the types of involvement they have had in medical care. Most likely, there is a cyclical relationship between the 2.³⁵

These findings provide reassurance that promoting trust will not likely cause a reversion to excessively paternalistic medical relationships. Trust in specific physicians is only weakly related to patients’ views about active vs passive roles. Trust in the medical profession is much more pertinent to these views, but that type of trust is distinct from trust in specific physicians. Moreover, trust in the medical profession is consistent with some desirable forms of patient involvement.

These findings also suggest that trust is able to form in relationships where patients are either active or passive. Assertive patient involvement should not be seen as indicative of distrust. Instead, other studies suggest that trust is promoted by communicating effectively with patients such as by listening carefully, answering questions clearly, giving them as much information as they want, and involving them in medical decisions.^26,27

Attitudes towards involvement in medical care. Apart from the relationship to trust, these findings shed important light on general attitudes toward involvement in medical care. Overall, the majority of people report following their physicians’ recommendations and think that it is better to rely on the expert judgment of physicians and seek professional help. Most people think the physician and patient should share control equally and make decisions together. More patients give control to their physicians than to themselves. These findings are consistent with previous studies.^8,28-36

Also, consistent with prior studies, we found that younger and more educated patients prefer more assertive roles, as do women. We found no racial differences, but this may be due to the under representation of minorities in our sample. Finally, the small amount of variability explained in our models indicates that while we have identified predictive factors, many other factors affect people’s views towards medical care. Physicians need to be aware that patient desires for participation vary, and communication about such desires is necessary during visits.

Limitations of this study

Several study limitations should be noted. First, our measures of involvement in medical care are self-reported and do not necessarily reflect patients’ actual behaviors. However, most measures have been previously validated to some extent. Second, the selection criteria for the study do not allow for generalization to populations that less routinely seek care or are uninsured. Lastly, this is an exploratory study that was not driven by specific hypotheses derived from prior studies or firm theory. The empirical study of trust and its connection with other attitudes and relationship characteristics is still in its infancy,³⁷ which calls for more exploratory approaches that identify areas of focus for future research. The connection between trust and patients’ involvement in medical care is one such area deserving further study.

Acknowledgments

Research supported by the Robert Wood Johnson Foundation, the National Eye Institute (EY012443-02), and the National Institute on Aging (AG015248-03).

CORRESPONDING AUTHOR
Felicia Trachtenberg, PhD, New England Research Institutes, 9 Galen Street, Watertown, MA 02472. E-mail: [email protected]

Published reports of non-consented contraceptive practices in Mexico,^1-3 including intrauterine device (IUD) placement, have been largely anecdotal and have not been systematically validated. The Family Planning Clinic of the Maricopa County Department of Public Health predominantly serves a Mexican immigrant population. Providers in the clinic have also reported hearing about non-consented IUD placement. To investigate this issue, 466 women between July 1, 2000, and July 1, 2002, were recruited to answer a survey. This sample of convenience represented 29% of new clients during this period. Informed consent was obtained from each participant and no woman refused to participate. The study was IRB-approved. The survey was translated into Spanish and delivered orally by bilingual interviewers.

The mean age of the participants was 27.2 years, and the mean number of prior pregnancies was 2.6. One hundred eighty-eight women (40%) reported receiving gynecologic health care in Mexico from 1 month to 20 years previously (mean of 6.1 years). One hundred four women reported having an IUD placed at some time, including 85 of those who had received care in Mexico (46%). Of these, 23 women reported having an IUD inserted without their knowledge or consent. All 23 reported that the IUD had been placed while receiving care in Mexico, representing 27 % of those who had received an IUD there. Twenty-one of the 23 women said the IUD had been placed immediately after they had given birth, and 2 while seeking family planning services. Sixteen reported that the IUD insertion occurred at a hospital, while 4 said it happened at a clinic; 3 did not respond to this question.

Five of the women realized within days that an IUD had been placed; 3 within weeks, 4 within months, and 9 did not find out for a year or more; 5 did not respond. Three felt a string, 3 said the IUD fell out, 7 reported that a second healthcare worker discovered it, 2 were informed later by the person who placed the IUD, and 4 had adverse symptoms (2 from infection). Of 19 responding, 2 said they had the IUD taken out the same day it was discovered, 4 about a month later, 5 about a year later, 5 one to five years later, 1 five to ten years later, and 2 still had the IUD in place. Sixteen provinces in Mexico were listed as the location where the IUD placement occurred.

A significant percentage of women of reproductive age from Mexico served by this border area family planning clinic reported that they had an IUD placed without their knowledge or consent in Mexico. It is possible that these women were actually informed about the procedure, and just did not fully understand it. We were not able to investigate if this practice occurs in other countries or if it affects women of all socioeconomic classes in Mexico.

It is not clear if the women who did not have the IUD removed chose to keep it as their contraceptive method of choice or because they lacked access to health care to have it removed. It is possible that many of them would have chosen an IUD had the option been presented to them.

Since our sample was one of convenience in a busy public clinic, the possibility for selection bias exists. Therefore these findings are very preliminary and need to be verified in larger, better controlled studies. However, all those who provide healthcare services to women immigrants from Mexico should add the possibility of complications from an unrecognized IUD to the differential diagnosis if patients present with pelvic or abdominal pain, pelvic infections, or infertility.

Corresponding author
 Doug Campos-Outcalt, MD, MPA, 4001 North 3rd Street, Phoenix, AZ 85012. E-mail: [email protected].

Published reports of non-consented contraceptive practices in Mexico,^1-3 including intrauterine device (IUD) placement, have been largely anecdotal and have not been systematically validated. The Family Planning Clinic of the Maricopa County Department of Public Health predominantly serves a Mexican immigrant population. Providers in the clinic have also reported hearing about non-consented IUD placement. To investigate this issue, 466 women between July 1, 2000, and July 1, 2002, were recruited to answer a survey. This sample of convenience represented 29% of new clients during this period. Informed consent was obtained from each participant and no woman refused to participate. The study was IRB-approved. The survey was translated into Spanish and delivered orally by bilingual interviewers.

The mean age of the participants was 27.2 years, and the mean number of prior pregnancies was 2.6. One hundred eighty-eight women (40%) reported receiving gynecologic health care in Mexico from 1 month to 20 years previously (mean of 6.1 years). One hundred four women reported having an IUD placed at some time, including 85 of those who had received care in Mexico (46%). Of these, 23 women reported having an IUD inserted without their knowledge or consent. All 23 reported that the IUD had been placed while receiving care in Mexico, representing 27 % of those who had received an IUD there. Twenty-one of the 23 women said the IUD had been placed immediately after they had given birth, and 2 while seeking family planning services. Sixteen reported that the IUD insertion occurred at a hospital, while 4 said it happened at a clinic; 3 did not respond to this question.

Five of the women realized within days that an IUD had been placed; 3 within weeks, 4 within months, and 9 did not find out for a year or more; 5 did not respond. Three felt a string, 3 said the IUD fell out, 7 reported that a second healthcare worker discovered it, 2 were informed later by the person who placed the IUD, and 4 had adverse symptoms (2 from infection). Of 19 responding, 2 said they had the IUD taken out the same day it was discovered, 4 about a month later, 5 about a year later, 5 one to five years later, 1 five to ten years later, and 2 still had the IUD in place. Sixteen provinces in Mexico were listed as the location where the IUD placement occurred.

A significant percentage of women of reproductive age from Mexico served by this border area family planning clinic reported that they had an IUD placed without their knowledge or consent in Mexico. It is possible that these women were actually informed about the procedure, and just did not fully understand it. We were not able to investigate if this practice occurs in other countries or if it affects women of all socioeconomic classes in Mexico.

It is not clear if the women who did not have the IUD removed chose to keep it as their contraceptive method of choice or because they lacked access to health care to have it removed. It is possible that many of them would have chosen an IUD had the option been presented to them.

Since our sample was one of convenience in a busy public clinic, the possibility for selection bias exists. Therefore these findings are very preliminary and need to be verified in larger, better controlled studies. However, all those who provide healthcare services to women immigrants from Mexico should add the possibility of complications from an unrecognized IUD to the differential diagnosis if patients present with pelvic or abdominal pain, pelvic infections, or infertility.

Corresponding author
 Doug Campos-Outcalt, MD, MPA, 4001 North 3rd Street, Phoenix, AZ 85012. E-mail: [email protected].

Published reports of non-consented contraceptive practices in Mexico,^1-3 including intrauterine device (IUD) placement, have been largely anecdotal and have not been systematically validated. The Family Planning Clinic of the Maricopa County Department of Public Health predominantly serves a Mexican immigrant population. Providers in the clinic have also reported hearing about non-consented IUD placement. To investigate this issue, 466 women between July 1, 2000, and July 1, 2002, were recruited to answer a survey. This sample of convenience represented 29% of new clients during this period. Informed consent was obtained from each participant and no woman refused to participate. The study was IRB-approved. The survey was translated into Spanish and delivered orally by bilingual interviewers.

The mean age of the participants was 27.2 years, and the mean number of prior pregnancies was 2.6. One hundred eighty-eight women (40%) reported receiving gynecologic health care in Mexico from 1 month to 20 years previously (mean of 6.1 years). One hundred four women reported having an IUD placed at some time, including 85 of those who had received care in Mexico (46%). Of these, 23 women reported having an IUD inserted without their knowledge or consent. All 23 reported that the IUD had been placed while receiving care in Mexico, representing 27 % of those who had received an IUD there. Twenty-one of the 23 women said the IUD had been placed immediately after they had given birth, and 2 while seeking family planning services. Sixteen reported that the IUD insertion occurred at a hospital, while 4 said it happened at a clinic; 3 did not respond to this question.

Five of the women realized within days that an IUD had been placed; 3 within weeks, 4 within months, and 9 did not find out for a year or more; 5 did not respond. Three felt a string, 3 said the IUD fell out, 7 reported that a second healthcare worker discovered it, 2 were informed later by the person who placed the IUD, and 4 had adverse symptoms (2 from infection). Of 19 responding, 2 said they had the IUD taken out the same day it was discovered, 4 about a month later, 5 about a year later, 5 one to five years later, 1 five to ten years later, and 2 still had the IUD in place. Sixteen provinces in Mexico were listed as the location where the IUD placement occurred.

A significant percentage of women of reproductive age from Mexico served by this border area family planning clinic reported that they had an IUD placed without their knowledge or consent in Mexico. It is possible that these women were actually informed about the procedure, and just did not fully understand it. We were not able to investigate if this practice occurs in other countries or if it affects women of all socioeconomic classes in Mexico.

It is not clear if the women who did not have the IUD removed chose to keep it as their contraceptive method of choice or because they lacked access to health care to have it removed. It is possible that many of them would have chosen an IUD had the option been presented to them.

Since our sample was one of convenience in a busy public clinic, the possibility for selection bias exists. Therefore these findings are very preliminary and need to be verified in larger, better controlled studies. However, all those who provide healthcare services to women immigrants from Mexico should add the possibility of complications from an unrecognized IUD to the differential diagnosis if patients present with pelvic or abdominal pain, pelvic infections, or infertility.

Corresponding author
 Doug Campos-Outcalt, MD, MPA, 4001 North 3rd Street, Phoenix, AZ 85012. E-mail: [email protected].

Like sensitivity and specificity, a likelihood ratio (LR) can be used to express the usefulness of diagnostic tests. A likelihood ratio is a ratio of 2 proportions: the subset of people with a particular test result among all those who have a specific disease, divided by the subset of people with the same test result among all those without the disease. The mathematical expression of this is:

What ratio results mean

Consider a study to assess the usefulness of a new blood test for colon cancer. Results of the blood test are reported as high probability of cancer, intermediate probability, or low probability.

All patients in the study undergo the blood test and colonoscopy, the gold standard for identifying colon cancer. The likelihood ratio of a “high probability” result is calculated thus:

In the example above, if L_R(high probability) is 10, this means that a high probability result is 10 times more likely to occur among people with the disease than among people without it.

Advantages of the likelihood ratio. Sensitivity and specificity can be used only with test results reported as positive or negative (dichotomous results). Likelihood ratios can be used with tests that have any number of outcomes. They can also be used in one form of Bayes’ theorem, as illustrated below, which has application to the Applied Evidence article on open-angle glaucoma in this issue.

Applying the likelihood ratio in this issue

On page 119 of this issue, Aref and Schmidt discuss the risk factors and diagnosis of open-angle glaucoma (OAG). Consider a 70-year-old African American woman who has difficulty seeing in the dark and has lost some peripheral vision in both eyes. Her sister has recently received a diagnosis of OAG. The patient’s risk factors and family history make a diagnosis of OAG likely.

How can a likelihood ratio help here? Direct ophthalmoscopy is warranted to determine if the patient has an elevated cup-disc ratio (>0.6). How useful would such an examination be in this case? In general, Bayes’ theorem tells us that new information should be interpreted in light of what is already known. The form of Bayes’ theorem applicable to diagnostic tests is the following:

Posttest odds of disease = Pretest odds of disease × likelihood ratio

First, calculate pretest odds. The relationship of odds to probability is fairly simple. Let’s assume, based on the patient’s history alone, we believe there is a 33% chance she has OAG. In other words, her pretest probability of OAG is 33%. We convert this probability to odds: Odds = probability/ 1 − probability = 0.33/ 1 − 0.33 = 1/2

Her odds of disease is therefore “1 in favor to 2 against.”

Next, find posttest odds. Now let’s assume that ophthalmoscopy reveals a cup-disc ratio of 0.8. According to Aref and Schmidt’s article, a cup-disc ratio of >0.6 (ie, a “positive” ophthalmoscopic examination) has an LR+ of 16; a cup-disc ratio of <0.6 has an LR− of 0.375 (or 3/8). Since the patient has a “positive” test result, we obtain:

Posttest odds of OAG = 1/2 × 16 = 16/2 = 8/1

Translating back to probability. To make things easier, we can convert this posttest odds of 8/1 to a probability:

Probability = odds in favor/odds in favor + odds against = 8/8+1 = 0.89 (or 89%).

After combining our ophthalmoscopic examination with the history, we can conclude that the patient has an 89% chance of having OAG.

If the result of the ophthalmoscopic examination was negative,

Posttest odds of OAG = 1/2 × 3/8 = 3/16.

The corresponding probability is roughly 0.16 (or 16%).

When likelihood ratios are most useful

In general, diagnostic tests of any kind are most useful for patients like the one described— those who have an intermediate pre-test probability of disease (usually 20%–60%). Very high or very low pretest probabilities of disease are less likely to influence post-test probability of disease.

You may be concerned that the value of pretest probability we chose is subjective. Bear in mind that much of our thinking in medicine is subjective, but based upon clinical experience and knowledge. Evidence-based medicine is a complement to, not a substitute for, clinical experience. Combining the objectivity of likelihood ratios with subjective pretest probabilities using Bayes’ theorem is consistent with the principles of evidence-based medicine.

Corresponding author
Goutham Rao, MD. E-mail: [email protected].

Like sensitivity and specificity, a likelihood ratio (LR) can be used to express the usefulness of diagnostic tests. A likelihood ratio is a ratio of 2 proportions: the subset of people with a particular test result among all those who have a specific disease, divided by the subset of people with the same test result among all those without the disease. The mathematical expression of this is:

What ratio results mean

Consider a study to assess the usefulness of a new blood test for colon cancer. Results of the blood test are reported as high probability of cancer, intermediate probability, or low probability.

All patients in the study undergo the blood test and colonoscopy, the gold standard for identifying colon cancer. The likelihood ratio of a “high probability” result is calculated thus:

In the example above, if L_R(high probability) is 10, this means that a high probability result is 10 times more likely to occur among people with the disease than among people without it.

Advantages of the likelihood ratio. Sensitivity and specificity can be used only with test results reported as positive or negative (dichotomous results). Likelihood ratios can be used with tests that have any number of outcomes. They can also be used in one form of Bayes’ theorem, as illustrated below, which has application to the Applied Evidence article on open-angle glaucoma in this issue.

Applying the likelihood ratio in this issue

On page 119 of this issue, Aref and Schmidt discuss the risk factors and diagnosis of open-angle glaucoma (OAG). Consider a 70-year-old African American woman who has difficulty seeing in the dark and has lost some peripheral vision in both eyes. Her sister has recently received a diagnosis of OAG. The patient’s risk factors and family history make a diagnosis of OAG likely.

How can a likelihood ratio help here? Direct ophthalmoscopy is warranted to determine if the patient has an elevated cup-disc ratio (>0.6). How useful would such an examination be in this case? In general, Bayes’ theorem tells us that new information should be interpreted in light of what is already known. The form of Bayes’ theorem applicable to diagnostic tests is the following:

Posttest odds of disease = Pretest odds of disease × likelihood ratio

First, calculate pretest odds. The relationship of odds to probability is fairly simple. Let’s assume, based on the patient’s history alone, we believe there is a 33% chance she has OAG. In other words, her pretest probability of OAG is 33%. We convert this probability to odds: Odds = probability/ 1 − probability = 0.33/ 1 − 0.33 = 1/2

Her odds of disease is therefore “1 in favor to 2 against.”

Next, find posttest odds. Now let’s assume that ophthalmoscopy reveals a cup-disc ratio of 0.8. According to Aref and Schmidt’s article, a cup-disc ratio of >0.6 (ie, a “positive” ophthalmoscopic examination) has an LR+ of 16; a cup-disc ratio of <0.6 has an LR− of 0.375 (or 3/8). Since the patient has a “positive” test result, we obtain:

Posttest odds of OAG = 1/2 × 16 = 16/2 = 8/1

Translating back to probability. To make things easier, we can convert this posttest odds of 8/1 to a probability:

Probability = odds in favor/odds in favor + odds against = 8/8+1 = 0.89 (or 89%).

After combining our ophthalmoscopic examination with the history, we can conclude that the patient has an 89% chance of having OAG.

If the result of the ophthalmoscopic examination was negative,

Posttest odds of OAG = 1/2 × 3/8 = 3/16.

The corresponding probability is roughly 0.16 (or 16%).

When likelihood ratios are most useful

In general, diagnostic tests of any kind are most useful for patients like the one described— those who have an intermediate pre-test probability of disease (usually 20%–60%). Very high or very low pretest probabilities of disease are less likely to influence post-test probability of disease.

You may be concerned that the value of pretest probability we chose is subjective. Bear in mind that much of our thinking in medicine is subjective, but based upon clinical experience and knowledge. Evidence-based medicine is a complement to, not a substitute for, clinical experience. Combining the objectivity of likelihood ratios with subjective pretest probabilities using Bayes’ theorem is consistent with the principles of evidence-based medicine.

Corresponding author
Goutham Rao, MD. E-mail: [email protected].

Like sensitivity and specificity, a likelihood ratio (LR) can be used to express the usefulness of diagnostic tests. A likelihood ratio is a ratio of 2 proportions: the subset of people with a particular test result among all those who have a specific disease, divided by the subset of people with the same test result among all those without the disease. The mathematical expression of this is:

What ratio results mean

Consider a study to assess the usefulness of a new blood test for colon cancer. Results of the blood test are reported as high probability of cancer, intermediate probability, or low probability.

All patients in the study undergo the blood test and colonoscopy, the gold standard for identifying colon cancer. The likelihood ratio of a “high probability” result is calculated thus:

In the example above, if L_R(high probability) is 10, this means that a high probability result is 10 times more likely to occur among people with the disease than among people without it.

Advantages of the likelihood ratio. Sensitivity and specificity can be used only with test results reported as positive or negative (dichotomous results). Likelihood ratios can be used with tests that have any number of outcomes. They can also be used in one form of Bayes’ theorem, as illustrated below, which has application to the Applied Evidence article on open-angle glaucoma in this issue.

Applying the likelihood ratio in this issue

On page 119 of this issue, Aref and Schmidt discuss the risk factors and diagnosis of open-angle glaucoma (OAG). Consider a 70-year-old African American woman who has difficulty seeing in the dark and has lost some peripheral vision in both eyes. Her sister has recently received a diagnosis of OAG. The patient’s risk factors and family history make a diagnosis of OAG likely.

How can a likelihood ratio help here? Direct ophthalmoscopy is warranted to determine if the patient has an elevated cup-disc ratio (>0.6). How useful would such an examination be in this case? In general, Bayes’ theorem tells us that new information should be interpreted in light of what is already known. The form of Bayes’ theorem applicable to diagnostic tests is the following:

Posttest odds of disease = Pretest odds of disease × likelihood ratio

First, calculate pretest odds. The relationship of odds to probability is fairly simple. Let’s assume, based on the patient’s history alone, we believe there is a 33% chance she has OAG. In other words, her pretest probability of OAG is 33%. We convert this probability to odds: Odds = probability/ 1 − probability = 0.33/ 1 − 0.33 = 1/2

Her odds of disease is therefore “1 in favor to 2 against.”

Next, find posttest odds. Now let’s assume that ophthalmoscopy reveals a cup-disc ratio of 0.8. According to Aref and Schmidt’s article, a cup-disc ratio of >0.6 (ie, a “positive” ophthalmoscopic examination) has an LR+ of 16; a cup-disc ratio of <0.6 has an LR− of 0.375 (or 3/8). Since the patient has a “positive” test result, we obtain:

Posttest odds of OAG = 1/2 × 16 = 16/2 = 8/1

Translating back to probability. To make things easier, we can convert this posttest odds of 8/1 to a probability:

Probability = odds in favor/odds in favor + odds against = 8/8+1 = 0.89 (or 89%).

After combining our ophthalmoscopic examination with the history, we can conclude that the patient has an 89% chance of having OAG.

If the result of the ophthalmoscopic examination was negative,

Posttest odds of OAG = 1/2 × 3/8 = 3/16.

The corresponding probability is roughly 0.16 (or 16%).

When likelihood ratios are most useful

In general, diagnostic tests of any kind are most useful for patients like the one described— those who have an intermediate pre-test probability of disease (usually 20%–60%). Very high or very low pretest probabilities of disease are less likely to influence post-test probability of disease.

You may be concerned that the value of pretest probability we chose is subjective. Bear in mind that much of our thinking in medicine is subjective, but based upon clinical experience and knowledge. Evidence-based medicine is a complement to, not a substitute for, clinical experience. Combining the objectivity of likelihood ratios with subjective pretest probabilities using Bayes’ theorem is consistent with the principles of evidence-based medicine.

Corresponding author
Goutham Rao, MD. E-mail: [email protected].