Susan A. Flocke, PhD

Over the past half century, changing medical technology, law, education, ethics, and research have influenced the current shape of physician-patient interactions.⁹ In 1956, the traditional model of Activity-Passivity (physician does something to the patient) was challenged with the revolutionary concept of active patient participation.¹⁰ The models of Guidance and Cooperation (physician tells patient what to do, patient cooperates) and Mutual Participation (physician enables patient to help him/herself, patient is a partner) were proposed¹⁰ and are reflected in modern theoretically-based interaction models. Numerous models have been proposed as variants of the Guidance/Cooperation model (eg, paternalistic model,¹¹ priestly model,¹² contractual model¹³) and the Mutual Participation model (eg, ethnographic model,¹⁴ consumerist model,^11,15 family systems model¹⁶). Few of these models, though, have been empirically evaluated. The best-developed and most-studied mutual participation model is the patient-centered method.^5,17-20

When data have been collected using quantitative or qualitative approaches, significant strides have been made in understanding physician-patient interaction^{3, 21-23} and the effect of such interactions on patient outcomes,^5,24,25 primarily patient satisfaction.^1,26-29 However, many studies have been limited by their focus on a narrow aspect of physician-patient communication, studying a small number of physicians or patients, and using medical students, residents, and hospital faculty as study subjects.

The purpose of this study was not to develop a new model of physician-patient interaction. Rather, variables characterizing physician style grounded by the direct observation of thousands of encounters for 138 community practicing family physicians were used to empirically cluster physicians into groups that represent distinct interaction styles. Because interaction style may be manifested in all phases of a patient encounter, we used as a guiding framework the 3 primary functions of an interview:^30,31gathering information, enhancing a healing relationship, and making and implementing decisions. The importance of each of these functions varies depending on the nature of the encounter, but our overall approach provides a practical way of conceptualizing physician-patient interaction style. The association of the empirically derived and theoretically-based physician styles are tested with 3 outcomes: 1) patient report of delivery of attributes of primary care measured using the Components of Primary Care Instrument (CPCI), 2) patient satisfaction with the visit, and 3) the duration of the visit.

Methods

This study was part of the larger Direct Observation of Primary Care (DOPC) study, a cross sectional observational study that examined the content of 4454 outpatient visits to family physicians in northeast Ohio. Details of the methods of the DOPC study have been described extensively elsewhere.^32,34 Briefly, 4 teams of 2 research nurses directly observed consecutive patient visits to 138 participating physicians in 84 practices between October 1994 and August 1995. The research nurses collected data on the content and context of consecutive office visits using the following methods: direct observation of the patient visit, patient exit questionnaire, medical record review, and collection of ethnographic field notes.^33,34

Measures

Patients’ perception of the delivery 5 attributes of primary care was measured by the Components of Primary Care Instrument (CPCI). Interpersonal communication was an evaluation of the ease of exchange of information between patient and physician. The physician’s accumulated knowledge about the patient refers to the physician’s understanding of the patient’s medical history, health care needs, and values. Coordination of care refers to the information received from referrals to specialists and previous health care visits, and its incorporation into the current and future care of the patient. Preference to see usual physician refers to the degree to which patients believed and valued that they could go to their regular physician for almost all problems. Scale scores demonstrate good internal consistency reliability (Cronbach’s alpha: .68–.79).³⁵ Continuity of care is measured by the Usual Provider Continuity index (UPC), which is the proportion of visits to the patient’s regular doctor in the past year out of the total number of physician visits in the past year.

Patient satisfaction was measured using the 4 physician-specific items from the MOS 9 Item Visit Rating Form³⁶ (Cronbach’s alpha = .89).³³ Also included on the patient survey was a single item assessing the degree to which patients’ expectations with the visit were met. Duration of the visit was the total face-to-face time the physician spent with the patient and was measured by direct observation.

Each physician’s interaction style was determined through a 2-step process. In the first step, ethnographic field notes were used to gather information that helps define core features of physician style. The field notes from 4 days of observation of 138 family physicians in 84 practices were transcribed and imported into FolioVIEWS37 for data management and coding. Analysis was conducted with an immersion-crystallization approach³⁸ involving repetitive reading and summarization of the text data. Case summaries were constructed from a sample of practices selected to maximize variation among practice characteristics such as size, physician sex, and practice location. The case summaries were independently reviewed, and important features were identified. These features were cross-checked against the original data. This process, and the resulting 30 features, are described in detail elsewhere.³²

Six of the features that emerged from the qualitative analyses pertain to physician style and are listed in Table 1. Each of the 3 primary interview functions³⁰ is represented by at least 1 feature, ensuring good coverage of the core aspects of the interaction. Gathering information is shaped by physician orientation and the clinical information allowed or elicited in the visits. Enhancing healing relationships is realized in part through affective connection with patients. The final function, making and implementing decisions, is influenced by the level of control or shared power with patients, the physician’s openness to patients’ agendas, and the physician’s willingness to negotiate options with patients.

The second step involved a cluster analysis of the 6 variables. First a hierarchical approach was used to estimate the number of clusters. Then a non-hierarchical clustering approach was used to determine physician classification among the clusters and the features that distinguish the clusters.³⁹ Analysis of variance was used to confirm that variables included in the cluster analysis significantly differed between at least 2 of the identified clusters, and thus were contributing to defining interaction style.

TABLE 1

Physician style variables

Analyses

The association of physician and patient characteristics with interaction style was assessed by chi square for categorical variables and by analysis of variance for continuous variables. The association of physician style with each of the 5 attributes of primary care measured by the CPCI, the indicators of patient satisfaction, and duration of the visit were tested using multilevel modeling,⁴⁰ to account for the hierarchical nature of data (ie, patients nested within physicians).

Results

Of the 4994 patients presenting for care by their family physicians, 4454 (89%) agreed to participate in the DOPC study. Physicians participating in the DOPC study were similar in age to national samples of family physicians, but over-represented female and residency-trained physicians.³⁴ Patient age, sex, and race were similar to the population of patients seeing family physicians and general practitioners nationally as reported in the National Ambulatory Medical Care Survey.³⁴ Patient questionnaires were returned by 3283 (74%) of the patients. Of those respondents, 2881 satisfactorily completed the CPCI, representing 88% of those returning a patient questionnaire and 65% of the total sample. The patients who completed the CPCI were more likely to be white, have private health care insurance, and be somewhat older than patients who did not complete the CPCI.³⁵

The cluster analysis identified 4 distinct groups of physicians. Each of the 138 physicians was classified into 1 group. Each of the 6 variables in the analysis contributed to defining the 4 groups by significantly (P

Forty-nine percent of physicians were classified as person focused. These physicians were more focused on the person than the disease, were perceived as personable and friendly, were open to the patient’s agenda, and frequently negotiated options with the patient. Physicians classified as biopsychosocial (16%) were more focused on the patient’s disease, but elicited psychosocial clinical information. Physicians classified as biomedical (20%) were also more focused on the patient’s disease and were unlikely to elicit psychosocial information. These physicians also demonstrated a low level of friendliness and were unlikely to negotiate options with the patient. The high physician control group’s major characteristics were domination of the encounter and disregard of the patient’s agenda (14%).

Association of physician characteristics with the interaction styles is presented in Table 2. The percent of male and female physicians differed greatly among the 4 style groups. The proportion of female physicians in the person-focused group was almost 4 times that of the biopsychosocial group and the high physician control group (P

As reported in Table 3, physician style is significantly associated with 3 of the 5 patient reports of the attributes of primary care. Physicians classified as having a person-focused approach have the highest mean score of communication; the other 3 styles score lower, with the high-physician-control style scoring the lowest. Person-focused and biopsychosocial physicians scored highest on patient reports of accumulated knowledge; those in the biomedical group scored the lowest. Coordination of care was highest among the person-focused group and lowest among the high-control group Across the different types of physician style, there was no difference in patient report of preference for his or her regular physician or the measure of continuity of care.

The associations of physician style with 2 indicators of patient satisfaction are displayed in Table 4. The highest group mean of patient satisfaction is for the person-focused style, and the lowest is for the high-physician-control group. The indicator of the degree to which patient expectations were met also follows this pattern. Also displayed in Table 4, the person-focused style demonstrated the longest average duration of visit, at 11.5 minutes; the high-physician-control group visits were the shortest in duration, at about 9.5 minutes.

TABLE 2

Physician and patient characteristics associated with interaction style

Association of physician style with attributes of primary care¹

Association of physician interaction style with patient satisfaction and duration of visit¹

Discussion

These data indicate that a person-focused approach is actively used in community practice, and is the style most congruent with patient-reported quality of primary care and satisfaction with care. Our data, in concert with data reported by others,^5,24 indicate strong support for the feasibility and value of the person-focused model. We found that, of the 4 distinct interaction styles, physicians with the person-focused style scored highest across all measures of the attributes of primary care and on the indicators of patient satisfaction, with the exception of continuity of care. In contrast, physicians with the high-control style were generally lowest on the primary care and satisfaction indicators.

It is important to emphasize that, even though the vast majority of patients in this sample are likely to have self-selected their primary care physician, patient rating of some attributes of primary care differed across the 4 physician styles. Patients of physicians with different styles equally valued seeing their regular physician, as reported by the preference-for-their-regular-doctor score; they exhibited similar proportions of continuity visits in the past year; and their satisfaction scores were all generally high. Patients appear to want to see their regular physician, regardless of interaction approach, even though some approaches—particularly the high-physician-control style—were rated poorer for communication, coordination of care, and accumulated knowledge.

There may be several explanations as to why a particular physician style is associated with specific patient reports of communication, accumulated knowledge, and coordination of care. Openness to the patient’s agenda and willingness to negotiate options—as was characteristic of the person-focused physicians—may facilitate good communication and convey an understanding of patient preferences and values regarding health. It is interesting to note that different groups scored lowest on some of the attributes of primary care. The high-physician-control group was the lowest on interpersonal communication and coordination of care. High-control physicians were more likely to dominate the agenda and the verbal exchanges. Patients may have felt they could not ask questions or that the physician did not listen to what they tried to say. The biomedical group of physicians were given the lowest scores by patients on accumulated knowledge, suggesting that patients thought these physicians were less likely to know their preferences and values regarding health care, know less about them as persons, and know less about their family and medical histories.

As others have proposed, we concur that interaction style is not a dichotomy or even a continuum of patient versus physician control, but is multidimensional, cutting across the main functions of the patient encounter (ie, information gathering, relationship building, and making and implementing decisions). These data provide some confirmation for the original scheme proposed by Szasz and Hollander,¹⁰ with the Mutual Participation model most represented by the person-focused approach and the Activity-Passivity model most represented by the high-physician-control group. The biopsychosocial and biomedical approaches represent different versions of the Guidance and Cooperative model.

The 4 types of physician style empirically derived from our data are similar to communication pattern types found by Roter et al,²⁷ in a study with similar aims but different methods. Of the 5 types reported, narrowly biomedical and expanded biomedical accounted for 65% of visits, and biopsychosocial accounted for 20%. Psychosocial and consumerist (distinguished by a high degree of patient questions) accounted for only 8% each. It is interesting that in our data, we found the person-focused style was by far the most common approach (49%) among this group of family physicians. These differences in use of particular interaction styles may have several explanations. First, these data were collected more recently.²⁷ Thus our data may reflect trends in a movement away from a paternalistic style and toward an increased patient participatory style. Second, our sample consisted entirely of family physicians practicing in the community, where the model of person-focused care may have a longer history of support and endorsement or be of greater importance to community family physicians, whose emphasis is on a breadth of care based on patient needs.^6,7,18

Physicians with a person-focused style granted the longest visits, while high-control-physicians granted the shortest—a difference of more than 2 minutes per visit on average. The associations were not explained away by accounting for patient or physician characteristics, suggesting that a person-focused style may require more time. However, others have found that physicians engaging in a patient participatory style had office visits that were of similar duration as found with other approaches,^{23, 27} although the average duration of visit for both of these studies were considerably longer than the office visits among our sample.

This study has several strengths. The use of community practicing physicians in real world conditions for whom visits were similar in content to the visits reported by NAMCS³⁴ adds to the generalizability of the findings. We have used an integration of qualitative and quantitative approaches to empirically derive categories of physician interaction style. Our data are based on nurse observation of an average of 32 encounters per physician and documented in rich and comprehensive qualitative fieldnotes. And finally, by using multilevel modeling, we have reported an honest estimate of the association of physician style and patient report of primary care by appropriately modeling the nested data structure.

The findings must be interpreted in light of potential study limitations. First, the patients who did not complete the patient questionnaire are somewhat different demographically than those patients who did complete it. However, non-completion of the questionnaire was not associated with physician style; therefore, it is unlikely that the associations would change, had these individuals been included. Second, because the study was cross-sectional we cannot control for patient self-selection of physicians. Nonetheless, since patients dissatisfied with the quality of care are likely to seek another physician, we would expect patient self-selection of physicians to bias the study toward the null, thus making our results even more remarkable.

These findings, in combination with the literature on the person-focused,²⁴ patient-centered^{5,17,19,20,41} and relationship-centered approaches,⁴² provide strong evidence to support the widespread implementation of this physician-patient interaction approach. Further investigation in community practice may lead to identification of ways to support and encourage person-focused care and the time needed to provide such care.

· Acknowledgments ·

The authors are indebted to the physicians, office staff members, and patients without whose participation this study would not have been possible. This paper was improved by helpful suggestions on an earlier draft by Kurt C. Stange, MD, PhD. This study was supported by a grant from the National Cancer Institute (1R01 CA60862) and in part by the Center for Research in Family Practice and Primary Care and the American Academy of Family Practice.

Over the past half century, changing medical technology, law, education, ethics, and research have influenced the current shape of physician-patient interactions.⁹ In 1956, the traditional model of Activity-Passivity (physician does something to the patient) was challenged with the revolutionary concept of active patient participation.¹⁰ The models of Guidance and Cooperation (physician tells patient what to do, patient cooperates) and Mutual Participation (physician enables patient to help him/herself, patient is a partner) were proposed¹⁰ and are reflected in modern theoretically-based interaction models. Numerous models have been proposed as variants of the Guidance/Cooperation model (eg, paternalistic model,¹¹ priestly model,¹² contractual model¹³) and the Mutual Participation model (eg, ethnographic model,¹⁴ consumerist model,^11,15 family systems model¹⁶). Few of these models, though, have been empirically evaluated. The best-developed and most-studied mutual participation model is the patient-centered method.^5,17-20

When data have been collected using quantitative or qualitative approaches, significant strides have been made in understanding physician-patient interaction^{3, 21-23} and the effect of such interactions on patient outcomes,^5,24,25 primarily patient satisfaction.^1,26-29 However, many studies have been limited by their focus on a narrow aspect of physician-patient communication, studying a small number of physicians or patients, and using medical students, residents, and hospital faculty as study subjects.

The purpose of this study was not to develop a new model of physician-patient interaction. Rather, variables characterizing physician style grounded by the direct observation of thousands of encounters for 138 community practicing family physicians were used to empirically cluster physicians into groups that represent distinct interaction styles. Because interaction style may be manifested in all phases of a patient encounter, we used as a guiding framework the 3 primary functions of an interview:^30,31gathering information, enhancing a healing relationship, and making and implementing decisions. The importance of each of these functions varies depending on the nature of the encounter, but our overall approach provides a practical way of conceptualizing physician-patient interaction style. The association of the empirically derived and theoretically-based physician styles are tested with 3 outcomes: 1) patient report of delivery of attributes of primary care measured using the Components of Primary Care Instrument (CPCI), 2) patient satisfaction with the visit, and 3) the duration of the visit.

Methods

This study was part of the larger Direct Observation of Primary Care (DOPC) study, a cross sectional observational study that examined the content of 4454 outpatient visits to family physicians in northeast Ohio. Details of the methods of the DOPC study have been described extensively elsewhere.^32,34 Briefly, 4 teams of 2 research nurses directly observed consecutive patient visits to 138 participating physicians in 84 practices between October 1994 and August 1995. The research nurses collected data on the content and context of consecutive office visits using the following methods: direct observation of the patient visit, patient exit questionnaire, medical record review, and collection of ethnographic field notes.^33,34

Measures

Patients’ perception of the delivery 5 attributes of primary care was measured by the Components of Primary Care Instrument (CPCI). Interpersonal communication was an evaluation of the ease of exchange of information between patient and physician. The physician’s accumulated knowledge about the patient refers to the physician’s understanding of the patient’s medical history, health care needs, and values. Coordination of care refers to the information received from referrals to specialists and previous health care visits, and its incorporation into the current and future care of the patient. Preference to see usual physician refers to the degree to which patients believed and valued that they could go to their regular physician for almost all problems. Scale scores demonstrate good internal consistency reliability (Cronbach’s alpha: .68–.79).³⁵ Continuity of care is measured by the Usual Provider Continuity index (UPC), which is the proportion of visits to the patient’s regular doctor in the past year out of the total number of physician visits in the past year.

Patient satisfaction was measured using the 4 physician-specific items from the MOS 9 Item Visit Rating Form³⁶ (Cronbach’s alpha = .89).³³ Also included on the patient survey was a single item assessing the degree to which patients’ expectations with the visit were met. Duration of the visit was the total face-to-face time the physician spent with the patient and was measured by direct observation.

Each physician’s interaction style was determined through a 2-step process. In the first step, ethnographic field notes were used to gather information that helps define core features of physician style. The field notes from 4 days of observation of 138 family physicians in 84 practices were transcribed and imported into FolioVIEWS37 for data management and coding. Analysis was conducted with an immersion-crystallization approach³⁸ involving repetitive reading and summarization of the text data. Case summaries were constructed from a sample of practices selected to maximize variation among practice characteristics such as size, physician sex, and practice location. The case summaries were independently reviewed, and important features were identified. These features were cross-checked against the original data. This process, and the resulting 30 features, are described in detail elsewhere.³²

Six of the features that emerged from the qualitative analyses pertain to physician style and are listed in Table 1. Each of the 3 primary interview functions³⁰ is represented by at least 1 feature, ensuring good coverage of the core aspects of the interaction. Gathering information is shaped by physician orientation and the clinical information allowed or elicited in the visits. Enhancing healing relationships is realized in part through affective connection with patients. The final function, making and implementing decisions, is influenced by the level of control or shared power with patients, the physician’s openness to patients’ agendas, and the physician’s willingness to negotiate options with patients.

The second step involved a cluster analysis of the 6 variables. First a hierarchical approach was used to estimate the number of clusters. Then a non-hierarchical clustering approach was used to determine physician classification among the clusters and the features that distinguish the clusters.³⁹ Analysis of variance was used to confirm that variables included in the cluster analysis significantly differed between at least 2 of the identified clusters, and thus were contributing to defining interaction style.

TABLE 1

Physician style variables

Analyses

The association of physician and patient characteristics with interaction style was assessed by chi square for categorical variables and by analysis of variance for continuous variables. The association of physician style with each of the 5 attributes of primary care measured by the CPCI, the indicators of patient satisfaction, and duration of the visit were tested using multilevel modeling,⁴⁰ to account for the hierarchical nature of data (ie, patients nested within physicians).

Results

Of the 4994 patients presenting for care by their family physicians, 4454 (89%) agreed to participate in the DOPC study. Physicians participating in the DOPC study were similar in age to national samples of family physicians, but over-represented female and residency-trained physicians.³⁴ Patient age, sex, and race were similar to the population of patients seeing family physicians and general practitioners nationally as reported in the National Ambulatory Medical Care Survey.³⁴ Patient questionnaires were returned by 3283 (74%) of the patients. Of those respondents, 2881 satisfactorily completed the CPCI, representing 88% of those returning a patient questionnaire and 65% of the total sample. The patients who completed the CPCI were more likely to be white, have private health care insurance, and be somewhat older than patients who did not complete the CPCI.³⁵

The cluster analysis identified 4 distinct groups of physicians. Each of the 138 physicians was classified into 1 group. Each of the 6 variables in the analysis contributed to defining the 4 groups by significantly (P

Forty-nine percent of physicians were classified as person focused. These physicians were more focused on the person than the disease, were perceived as personable and friendly, were open to the patient’s agenda, and frequently negotiated options with the patient. Physicians classified as biopsychosocial (16%) were more focused on the patient’s disease, but elicited psychosocial clinical information. Physicians classified as biomedical (20%) were also more focused on the patient’s disease and were unlikely to elicit psychosocial information. These physicians also demonstrated a low level of friendliness and were unlikely to negotiate options with the patient. The high physician control group’s major characteristics were domination of the encounter and disregard of the patient’s agenda (14%).

Association of physician characteristics with the interaction styles is presented in Table 2. The percent of male and female physicians differed greatly among the 4 style groups. The proportion of female physicians in the person-focused group was almost 4 times that of the biopsychosocial group and the high physician control group (P

As reported in Table 3, physician style is significantly associated with 3 of the 5 patient reports of the attributes of primary care. Physicians classified as having a person-focused approach have the highest mean score of communication; the other 3 styles score lower, with the high-physician-control style scoring the lowest. Person-focused and biopsychosocial physicians scored highest on patient reports of accumulated knowledge; those in the biomedical group scored the lowest. Coordination of care was highest among the person-focused group and lowest among the high-control group Across the different types of physician style, there was no difference in patient report of preference for his or her regular physician or the measure of continuity of care.

The associations of physician style with 2 indicators of patient satisfaction are displayed in Table 4. The highest group mean of patient satisfaction is for the person-focused style, and the lowest is for the high-physician-control group. The indicator of the degree to which patient expectations were met also follows this pattern. Also displayed in Table 4, the person-focused style demonstrated the longest average duration of visit, at 11.5 minutes; the high-physician-control group visits were the shortest in duration, at about 9.5 minutes.

TABLE 2

Physician and patient characteristics associated with interaction style

Association of physician style with attributes of primary care¹

Association of physician interaction style with patient satisfaction and duration of visit¹

Discussion

These data indicate that a person-focused approach is actively used in community practice, and is the style most congruent with patient-reported quality of primary care and satisfaction with care. Our data, in concert with data reported by others,^5,24 indicate strong support for the feasibility and value of the person-focused model. We found that, of the 4 distinct interaction styles, physicians with the person-focused style scored highest across all measures of the attributes of primary care and on the indicators of patient satisfaction, with the exception of continuity of care. In contrast, physicians with the high-control style were generally lowest on the primary care and satisfaction indicators.

It is important to emphasize that, even though the vast majority of patients in this sample are likely to have self-selected their primary care physician, patient rating of some attributes of primary care differed across the 4 physician styles. Patients of physicians with different styles equally valued seeing their regular physician, as reported by the preference-for-their-regular-doctor score; they exhibited similar proportions of continuity visits in the past year; and their satisfaction scores were all generally high. Patients appear to want to see their regular physician, regardless of interaction approach, even though some approaches—particularly the high-physician-control style—were rated poorer for communication, coordination of care, and accumulated knowledge.

There may be several explanations as to why a particular physician style is associated with specific patient reports of communication, accumulated knowledge, and coordination of care. Openness to the patient’s agenda and willingness to negotiate options—as was characteristic of the person-focused physicians—may facilitate good communication and convey an understanding of patient preferences and values regarding health. It is interesting to note that different groups scored lowest on some of the attributes of primary care. The high-physician-control group was the lowest on interpersonal communication and coordination of care. High-control physicians were more likely to dominate the agenda and the verbal exchanges. Patients may have felt they could not ask questions or that the physician did not listen to what they tried to say. The biomedical group of physicians were given the lowest scores by patients on accumulated knowledge, suggesting that patients thought these physicians were less likely to know their preferences and values regarding health care, know less about them as persons, and know less about their family and medical histories.

As others have proposed, we concur that interaction style is not a dichotomy or even a continuum of patient versus physician control, but is multidimensional, cutting across the main functions of the patient encounter (ie, information gathering, relationship building, and making and implementing decisions). These data provide some confirmation for the original scheme proposed by Szasz and Hollander,¹⁰ with the Mutual Participation model most represented by the person-focused approach and the Activity-Passivity model most represented by the high-physician-control group. The biopsychosocial and biomedical approaches represent different versions of the Guidance and Cooperative model.

The 4 types of physician style empirically derived from our data are similar to communication pattern types found by Roter et al,²⁷ in a study with similar aims but different methods. Of the 5 types reported, narrowly biomedical and expanded biomedical accounted for 65% of visits, and biopsychosocial accounted for 20%. Psychosocial and consumerist (distinguished by a high degree of patient questions) accounted for only 8% each. It is interesting that in our data, we found the person-focused style was by far the most common approach (49%) among this group of family physicians. These differences in use of particular interaction styles may have several explanations. First, these data were collected more recently.²⁷ Thus our data may reflect trends in a movement away from a paternalistic style and toward an increased patient participatory style. Second, our sample consisted entirely of family physicians practicing in the community, where the model of person-focused care may have a longer history of support and endorsement or be of greater importance to community family physicians, whose emphasis is on a breadth of care based on patient needs.^6,7,18

Physicians with a person-focused style granted the longest visits, while high-control-physicians granted the shortest—a difference of more than 2 minutes per visit on average. The associations were not explained away by accounting for patient or physician characteristics, suggesting that a person-focused style may require more time. However, others have found that physicians engaging in a patient participatory style had office visits that were of similar duration as found with other approaches,^{23, 27} although the average duration of visit for both of these studies were considerably longer than the office visits among our sample.

This study has several strengths. The use of community practicing physicians in real world conditions for whom visits were similar in content to the visits reported by NAMCS³⁴ adds to the generalizability of the findings. We have used an integration of qualitative and quantitative approaches to empirically derive categories of physician interaction style. Our data are based on nurse observation of an average of 32 encounters per physician and documented in rich and comprehensive qualitative fieldnotes. And finally, by using multilevel modeling, we have reported an honest estimate of the association of physician style and patient report of primary care by appropriately modeling the nested data structure.

The findings must be interpreted in light of potential study limitations. First, the patients who did not complete the patient questionnaire are somewhat different demographically than those patients who did complete it. However, non-completion of the questionnaire was not associated with physician style; therefore, it is unlikely that the associations would change, had these individuals been included. Second, because the study was cross-sectional we cannot control for patient self-selection of physicians. Nonetheless, since patients dissatisfied with the quality of care are likely to seek another physician, we would expect patient self-selection of physicians to bias the study toward the null, thus making our results even more remarkable.

These findings, in combination with the literature on the person-focused,²⁴ patient-centered^{5,17,19,20,41} and relationship-centered approaches,⁴² provide strong evidence to support the widespread implementation of this physician-patient interaction approach. Further investigation in community practice may lead to identification of ways to support and encourage person-focused care and the time needed to provide such care.

· Acknowledgments ·

The authors are indebted to the physicians, office staff members, and patients without whose participation this study would not have been possible. This paper was improved by helpful suggestions on an earlier draft by Kurt C. Stange, MD, PhD. This study was supported by a grant from the National Cancer Institute (1R01 CA60862) and in part by the Center for Research in Family Practice and Primary Care and the American Academy of Family Practice.

Over the past half century, changing medical technology, law, education, ethics, and research have influenced the current shape of physician-patient interactions.⁹ In 1956, the traditional model of Activity-Passivity (physician does something to the patient) was challenged with the revolutionary concept of active patient participation.¹⁰ The models of Guidance and Cooperation (physician tells patient what to do, patient cooperates) and Mutual Participation (physician enables patient to help him/herself, patient is a partner) were proposed¹⁰ and are reflected in modern theoretically-based interaction models. Numerous models have been proposed as variants of the Guidance/Cooperation model (eg, paternalistic model,¹¹ priestly model,¹² contractual model¹³) and the Mutual Participation model (eg, ethnographic model,¹⁴ consumerist model,^11,15 family systems model¹⁶). Few of these models, though, have been empirically evaluated. The best-developed and most-studied mutual participation model is the patient-centered method.^5,17-20

When data have been collected using quantitative or qualitative approaches, significant strides have been made in understanding physician-patient interaction^{3, 21-23} and the effect of such interactions on patient outcomes,^5,24,25 primarily patient satisfaction.^1,26-29 However, many studies have been limited by their focus on a narrow aspect of physician-patient communication, studying a small number of physicians or patients, and using medical students, residents, and hospital faculty as study subjects.

The purpose of this study was not to develop a new model of physician-patient interaction. Rather, variables characterizing physician style grounded by the direct observation of thousands of encounters for 138 community practicing family physicians were used to empirically cluster physicians into groups that represent distinct interaction styles. Because interaction style may be manifested in all phases of a patient encounter, we used as a guiding framework the 3 primary functions of an interview:^30,31gathering information, enhancing a healing relationship, and making and implementing decisions. The importance of each of these functions varies depending on the nature of the encounter, but our overall approach provides a practical way of conceptualizing physician-patient interaction style. The association of the empirically derived and theoretically-based physician styles are tested with 3 outcomes: 1) patient report of delivery of attributes of primary care measured using the Components of Primary Care Instrument (CPCI), 2) patient satisfaction with the visit, and 3) the duration of the visit.

Methods

This study was part of the larger Direct Observation of Primary Care (DOPC) study, a cross sectional observational study that examined the content of 4454 outpatient visits to family physicians in northeast Ohio. Details of the methods of the DOPC study have been described extensively elsewhere.^32,34 Briefly, 4 teams of 2 research nurses directly observed consecutive patient visits to 138 participating physicians in 84 practices between October 1994 and August 1995. The research nurses collected data on the content and context of consecutive office visits using the following methods: direct observation of the patient visit, patient exit questionnaire, medical record review, and collection of ethnographic field notes.^33,34

Measures

Patients’ perception of the delivery 5 attributes of primary care was measured by the Components of Primary Care Instrument (CPCI). Interpersonal communication was an evaluation of the ease of exchange of information between patient and physician. The physician’s accumulated knowledge about the patient refers to the physician’s understanding of the patient’s medical history, health care needs, and values. Coordination of care refers to the information received from referrals to specialists and previous health care visits, and its incorporation into the current and future care of the patient. Preference to see usual physician refers to the degree to which patients believed and valued that they could go to their regular physician for almost all problems. Scale scores demonstrate good internal consistency reliability (Cronbach’s alpha: .68–.79).³⁵ Continuity of care is measured by the Usual Provider Continuity index (UPC), which is the proportion of visits to the patient’s regular doctor in the past year out of the total number of physician visits in the past year.

Patient satisfaction was measured using the 4 physician-specific items from the MOS 9 Item Visit Rating Form³⁶ (Cronbach’s alpha = .89).³³ Also included on the patient survey was a single item assessing the degree to which patients’ expectations with the visit were met. Duration of the visit was the total face-to-face time the physician spent with the patient and was measured by direct observation.

Each physician’s interaction style was determined through a 2-step process. In the first step, ethnographic field notes were used to gather information that helps define core features of physician style. The field notes from 4 days of observation of 138 family physicians in 84 practices were transcribed and imported into FolioVIEWS37 for data management and coding. Analysis was conducted with an immersion-crystallization approach³⁸ involving repetitive reading and summarization of the text data. Case summaries were constructed from a sample of practices selected to maximize variation among practice characteristics such as size, physician sex, and practice location. The case summaries were independently reviewed, and important features were identified. These features were cross-checked against the original data. This process, and the resulting 30 features, are described in detail elsewhere.³²

Six of the features that emerged from the qualitative analyses pertain to physician style and are listed in Table 1. Each of the 3 primary interview functions³⁰ is represented by at least 1 feature, ensuring good coverage of the core aspects of the interaction. Gathering information is shaped by physician orientation and the clinical information allowed or elicited in the visits. Enhancing healing relationships is realized in part through affective connection with patients. The final function, making and implementing decisions, is influenced by the level of control or shared power with patients, the physician’s openness to patients’ agendas, and the physician’s willingness to negotiate options with patients.

The second step involved a cluster analysis of the 6 variables. First a hierarchical approach was used to estimate the number of clusters. Then a non-hierarchical clustering approach was used to determine physician classification among the clusters and the features that distinguish the clusters.³⁹ Analysis of variance was used to confirm that variables included in the cluster analysis significantly differed between at least 2 of the identified clusters, and thus were contributing to defining interaction style.

TABLE 1

Physician style variables

Analyses

The association of physician and patient characteristics with interaction style was assessed by chi square for categorical variables and by analysis of variance for continuous variables. The association of physician style with each of the 5 attributes of primary care measured by the CPCI, the indicators of patient satisfaction, and duration of the visit were tested using multilevel modeling,⁴⁰ to account for the hierarchical nature of data (ie, patients nested within physicians).

Results

Of the 4994 patients presenting for care by their family physicians, 4454 (89%) agreed to participate in the DOPC study. Physicians participating in the DOPC study were similar in age to national samples of family physicians, but over-represented female and residency-trained physicians.³⁴ Patient age, sex, and race were similar to the population of patients seeing family physicians and general practitioners nationally as reported in the National Ambulatory Medical Care Survey.³⁴ Patient questionnaires were returned by 3283 (74%) of the patients. Of those respondents, 2881 satisfactorily completed the CPCI, representing 88% of those returning a patient questionnaire and 65% of the total sample. The patients who completed the CPCI were more likely to be white, have private health care insurance, and be somewhat older than patients who did not complete the CPCI.³⁵

The cluster analysis identified 4 distinct groups of physicians. Each of the 138 physicians was classified into 1 group. Each of the 6 variables in the analysis contributed to defining the 4 groups by significantly (P

Forty-nine percent of physicians were classified as person focused. These physicians were more focused on the person than the disease, were perceived as personable and friendly, were open to the patient’s agenda, and frequently negotiated options with the patient. Physicians classified as biopsychosocial (16%) were more focused on the patient’s disease, but elicited psychosocial clinical information. Physicians classified as biomedical (20%) were also more focused on the patient’s disease and were unlikely to elicit psychosocial information. These physicians also demonstrated a low level of friendliness and were unlikely to negotiate options with the patient. The high physician control group’s major characteristics were domination of the encounter and disregard of the patient’s agenda (14%).

Association of physician characteristics with the interaction styles is presented in Table 2. The percent of male and female physicians differed greatly among the 4 style groups. The proportion of female physicians in the person-focused group was almost 4 times that of the biopsychosocial group and the high physician control group (P

As reported in Table 3, physician style is significantly associated with 3 of the 5 patient reports of the attributes of primary care. Physicians classified as having a person-focused approach have the highest mean score of communication; the other 3 styles score lower, with the high-physician-control style scoring the lowest. Person-focused and biopsychosocial physicians scored highest on patient reports of accumulated knowledge; those in the biomedical group scored the lowest. Coordination of care was highest among the person-focused group and lowest among the high-control group Across the different types of physician style, there was no difference in patient report of preference for his or her regular physician or the measure of continuity of care.

The associations of physician style with 2 indicators of patient satisfaction are displayed in Table 4. The highest group mean of patient satisfaction is for the person-focused style, and the lowest is for the high-physician-control group. The indicator of the degree to which patient expectations were met also follows this pattern. Also displayed in Table 4, the person-focused style demonstrated the longest average duration of visit, at 11.5 minutes; the high-physician-control group visits were the shortest in duration, at about 9.5 minutes.

TABLE 2

Physician and patient characteristics associated with interaction style

Association of physician style with attributes of primary care¹

Association of physician interaction style with patient satisfaction and duration of visit¹

Discussion

These data indicate that a person-focused approach is actively used in community practice, and is the style most congruent with patient-reported quality of primary care and satisfaction with care. Our data, in concert with data reported by others,^5,24 indicate strong support for the feasibility and value of the person-focused model. We found that, of the 4 distinct interaction styles, physicians with the person-focused style scored highest across all measures of the attributes of primary care and on the indicators of patient satisfaction, with the exception of continuity of care. In contrast, physicians with the high-control style were generally lowest on the primary care and satisfaction indicators.

It is important to emphasize that, even though the vast majority of patients in this sample are likely to have self-selected their primary care physician, patient rating of some attributes of primary care differed across the 4 physician styles. Patients of physicians with different styles equally valued seeing their regular physician, as reported by the preference-for-their-regular-doctor score; they exhibited similar proportions of continuity visits in the past year; and their satisfaction scores were all generally high. Patients appear to want to see their regular physician, regardless of interaction approach, even though some approaches—particularly the high-physician-control style—were rated poorer for communication, coordination of care, and accumulated knowledge.

There may be several explanations as to why a particular physician style is associated with specific patient reports of communication, accumulated knowledge, and coordination of care. Openness to the patient’s agenda and willingness to negotiate options—as was characteristic of the person-focused physicians—may facilitate good communication and convey an understanding of patient preferences and values regarding health. It is interesting to note that different groups scored lowest on some of the attributes of primary care. The high-physician-control group was the lowest on interpersonal communication and coordination of care. High-control physicians were more likely to dominate the agenda and the verbal exchanges. Patients may have felt they could not ask questions or that the physician did not listen to what they tried to say. The biomedical group of physicians were given the lowest scores by patients on accumulated knowledge, suggesting that patients thought these physicians were less likely to know their preferences and values regarding health care, know less about them as persons, and know less about their family and medical histories.

As others have proposed, we concur that interaction style is not a dichotomy or even a continuum of patient versus physician control, but is multidimensional, cutting across the main functions of the patient encounter (ie, information gathering, relationship building, and making and implementing decisions). These data provide some confirmation for the original scheme proposed by Szasz and Hollander,¹⁰ with the Mutual Participation model most represented by the person-focused approach and the Activity-Passivity model most represented by the high-physician-control group. The biopsychosocial and biomedical approaches represent different versions of the Guidance and Cooperative model.

The 4 types of physician style empirically derived from our data are similar to communication pattern types found by Roter et al,²⁷ in a study with similar aims but different methods. Of the 5 types reported, narrowly biomedical and expanded biomedical accounted for 65% of visits, and biopsychosocial accounted for 20%. Psychosocial and consumerist (distinguished by a high degree of patient questions) accounted for only 8% each. It is interesting that in our data, we found the person-focused style was by far the most common approach (49%) among this group of family physicians. These differences in use of particular interaction styles may have several explanations. First, these data were collected more recently.²⁷ Thus our data may reflect trends in a movement away from a paternalistic style and toward an increased patient participatory style. Second, our sample consisted entirely of family physicians practicing in the community, where the model of person-focused care may have a longer history of support and endorsement or be of greater importance to community family physicians, whose emphasis is on a breadth of care based on patient needs.^6,7,18

Physicians with a person-focused style granted the longest visits, while high-control-physicians granted the shortest—a difference of more than 2 minutes per visit on average. The associations were not explained away by accounting for patient or physician characteristics, suggesting that a person-focused style may require more time. However, others have found that physicians engaging in a patient participatory style had office visits that were of similar duration as found with other approaches,^{23, 27} although the average duration of visit for both of these studies were considerably longer than the office visits among our sample.

This study has several strengths. The use of community practicing physicians in real world conditions for whom visits were similar in content to the visits reported by NAMCS³⁴ adds to the generalizability of the findings. We have used an integration of qualitative and quantitative approaches to empirically derive categories of physician interaction style. Our data are based on nurse observation of an average of 32 encounters per physician and documented in rich and comprehensive qualitative fieldnotes. And finally, by using multilevel modeling, we have reported an honest estimate of the association of physician style and patient report of primary care by appropriately modeling the nested data structure.

The findings must be interpreted in light of potential study limitations. First, the patients who did not complete the patient questionnaire are somewhat different demographically than those patients who did complete it. However, non-completion of the questionnaire was not associated with physician style; therefore, it is unlikely that the associations would change, had these individuals been included. Second, because the study was cross-sectional we cannot control for patient self-selection of physicians. Nonetheless, since patients dissatisfied with the quality of care are likely to seek another physician, we would expect patient self-selection of physicians to bias the study toward the null, thus making our results even more remarkable.

These findings, in combination with the literature on the person-focused,²⁴ patient-centered^{5,17,19,20,41} and relationship-centered approaches,⁴² provide strong evidence to support the widespread implementation of this physician-patient interaction approach. Further investigation in community practice may lead to identification of ways to support and encourage person-focused care and the time needed to provide such care.

· Acknowledgments ·

The authors are indebted to the physicians, office staff members, and patients without whose participation this study would not have been possible. This paper was improved by helpful suggestions on an earlier draft by Kurt C. Stange, MD, PhD. This study was supported by a grant from the National Cancer Institute (1R01 CA60862) and in part by the Center for Research in Family Practice and Primary Care and the American Academy of Family Practice.

Primary care disciplines continue to have a central role in the health care of Americans. They provide breadth of care within an ongoing relationship, bridging the boundaries between health and illness and guiding access to more narrowly focused care when needed.¹ The ability to orchestrate a broad health agenda during a visit is central to primary care, but this ability is challenged by competing demands for time.²

Attempts to influence provision of care and treatment decisions by primary care physicians, such as financial incentives, administrative restrictions, and the implementation of evidence-based clinical guidelines add to the demands on physicians’ time and may affect how time is allocated during the day and with each patient. Within this context a primary care physician must prioritize the agenda for each patient visit. This may include providing services beyond the patient’s primary reason for the visit as time permits, such as including preventive services,³ follow-up of acute or chronic illnesses,¹ mental health⁴ or family issues,^5-7 or investigating “by the way” patient comments that may indicate serious medical issues.

The competing demands for time are compounded by patient requests during the visit. Based on an audiotape of 139 patient encounters, Kravitz and colleagues⁸ reported that on average a patient makes 5 requests for physician action or information per visit, and the number of unfulfilled requests was negatively associated with patient satisfaction. Such findings may fuel a sense of pressure to address patient requests. Also, another recent report indicates that the majority of patients do not have the opportunity to express all of their concerns before the physician redirects the interview; once redirected, additional patient concerns are rarely elicited.⁹ Fitting both the physician’s and patient’s agenda into the time allotted for an outpatient visit has important implications for the duration of the visit, physician productivity, and possibly patient outcomes.

Data on the number of problems raised and addressed have been limited by the lack of appropriate collection methods. Primarily audio and video technology have been used for the study of physician-patient communication.^10-12 Direct observation of patient encounters^12,13 and incorporation of ethnographic approaches have more recently been employed to fill a large void in the understanding of the content, context, and complexity of primary care.^13-15 Findings from the Direct Observation of Primary Care study, which employed such methods, indicate that among 4454 patient visits care was provided to a secondary patient during 18% of the visits and preventive services were addressed during 32% of the illness visits.³ Data from that study provide a glimpse into some types of problems addressed in addition to the main reason for the visit; however, data about the number of problems addressed during patient encounters were not specifically collected by the nurse observer.

When additional issues are raised during a patient encounter, little is known about the nature of these problems, how additional problems affect the duration of the visit, and how well additional problems are reflected in the billing record. This led us to conduct an observational study to ask: How many problems are addressed during family practice outpatient visits, and who is raising additional problems? How much work and time is associated with addressing problems raised beyond the initial problem? How well does the billing list represent the number of problems addressed during the outpatient visit? Our study was designed to directly observe and record how many problems were raised and addressed during outpatient visits to family physicians.

Methods

Seven first-year medical students observed patient care provided by their summer fellowship family physician preceptor and other physicians in the preceptor’s practice from June through August 1999. Six of the sites were located in Northeast Ohio, and one was in Tulsa, Oklahoma.

Each student collected data on one randomly selected adult patient encounter for each half day of precepting. At the beginning of each half-day of patient care the student rolled a die to generate a random number to select a patient from the patient schedule. To ensure random selection of encounters within each half-day session, on alternating days the random number was counted from the beginning or the end of the half-day schedule. If the selected patient was aged younger than 18 years, the patient or physician preferred the encounter not be observed, or the patient did not show up for the scheduled appointment, the next scheduled appointment was selected as a replacement. Patient age and sex were collected for those who were no-shows or chose not to be observed, so they could be compared with those patients who were observed. Each student was to collect data on approximately 50 patient encounters during the 6-week summer fellowship. The physicians were blinded to the study purpose and were not told which patient encounter would be included in the study.

A problem was operationalized as an issue requiring physician action in the form of a decision, diagnosis, treatment, or monitoring. Each item was listed as it was raised, and the type of problem, who raised it, and what physician actions were involved to address it were coded. Each problem was coded as 1 of 14 categories: acute, acute follow-up, chronic, chronic follow-up, prevention, prevention follow-up, psychosocial, psychosocial follow-up, work-related administrative, health care system-related administrative, other family member’s problem, pregnancy, emergent, and other. The person who raised the problem was coded as 1 of 3 options: the physician, the patient or another person in the room. Multiple physician actions could be coded for how the problem was addressed. The 19 physician action categories included: question, reassurance, examination, procedure, referral, return visit, advice, review tests, order laboratory testing, prescription, provide written material, imaging, admits uncertainty, counseling, return to work/time off work letter, defer, complementary/alternative medicine, ignored or lost, and other.

Patient characteristics, the duration of the visit, and the billing diagnoses for each visit were also recorded on the data collection form. Videotaped encounters were used to pilot test the data collection form, to allow the observers to practice using the form in real time, and to calibrate the observers before data collection in the field.

We used descriptive statistics to address most research questions. Student t tests and chi-square tests were used to compare age and sex differences between participants and nonparticipants. We tested the association of the number of problems with the duration of the visit with analysis of variance and a test for linear trend. A difference score of the number of problems observed and the number of problems recorded on the billing sheet for the encounter was computed and summarized graphically.

Results

We collected usable data on 266 encounters representing 37 physicians. Patient and visit characteristics are displayed in Table 1. The patients had an average age of 48 years, and 69% were women. They were predominately white. A large proportion was observed visiting their regular primary care physician (83%), and 85% were established patients of the practice. Most of the observed patients had some kind of commercial health care insurance, 19% had Medicare, and a small proportion had Medicaid or no insurance. The visit duration ranged from 2 to 65 minutes; the median was 15 minutes with a mean of 19.3 (standard deviation [SD]=12.7). The first problem raised was most commonly an acute problem (49%); prevention and chronic illness were the first problem raised during 21% and 19% of encounters, respectively. Patients who were randomly selected but were not observed (n=52, primarily no-shows) were similar in sex (67% women, c2 =0.119, P=.73 ) but were younger than those patients who were observed (mean age=32.1 years, t=3.79, P=.001).

On average, 2.7 problems were raised during an encounter Table 2. Forty-four percent of all problems were classified as acute, 30% chronic, 14% prevention, 4% administrative, 2% psychosocial, and 6% were classified as other. Of the observed encounters, 73% had more than one problem addressed. The physician raised 36% of these additional problems, and patients raised 58%. The problems raised by physicians were most frequently pertaining to chronic illness, prevention, and follow-up issues. The problems raised by patients were most likely to be acute illness problems. Additional problems were least likely to arise when the first problem addressed was an acute problem (61%) compared with visits during which the first problem addressed was chronic or prevention focused, where 88% and 87%, respectively, included additional problems during the visit (c2=21.2, P <.001).

On average, 8 (SD=4.5) physician actions were observed per encounter Table 2. Physicians performed an average of 3.3 (SD=1.2) actions per problem. The most common physician actions were questioning (77%), physical examination (49%), prescription writing (32%), providing advice (31%), and reassurance (25%). Of the 452 additional problems raised, only 3% of problems were ignored, and 6% were deferred to another visit.

The association of the number of problems addressed with the duration of the visit was assessed by analysis of variance and a test for linear trend. As shown in Figure 1, the duration of the visit increased approximately 2.5 minutes for each additional problem addressed (P <.001 for linear trend). The visit duration within each of the number of problem groups varied greatly as indicated by the large range for each group; however, the SD for each of the groups as indicated by the shaded bars are a similar size for each of the groups (Levene’s test of equality of error variance=1.48, P=.195).

The concordance between the number of problems observed and the number of problems on the billing sheet was modest, with a trend toward billing for fewer problems than were observed. As shown in Figure 2, 29% of encounters represented a match between the number of problems observed and the number of problems on the billing sheet. Fifty-eight percent of the encounters had more problems observed than recorded on the billing sheet. A much smaller proportion of encounters recorded more problems on the billing sheet than were observed during the encounter.

Discussion

Our exploratory study suggests that it is common for multiple problems to be addressed during visits to a family physician regardless of the initial reason for the visit. Additional problems are raised by both physicians and patients and are rarely deferred or ignored by the physician. Although the phenomenon of integrating a broad health agenda and addressing multiple problems during a single outpatient visit may be well known by practicing community-based family physicians, it may not be recognized by policymakers or health services researchers whose window into the process of outpatient care is provided by the medical record and billing data.

Addressing the majority of a patient’s health care needs and providing comprehensive care is a core feature of quality primary care.^16-20 Previous work has documented the wide range of diagnoses and clusters of diagnoses that family physicians commonly address during outpatient care.^13,21 However, truly comprehensive care goes beyond providing a broad array of services; it also involves the integration of care in a physician-patient relationship context. Prioritizing, providing, and orchestrating care for acute and undifferentiated illness, chronic disease, preventive services, and mental health care represents a key feature of primary care practice such that the care is greater than the sum of its individual commodities.¹ These data suggest that single visits often address a broad agenda of health care.

Overall, as the number of problems increase so does the length of the visit. Others have found that ordering or performing more tests, providing preventive services, and conducting ambulatory surgical procedures increase the length of the visit.²² It is not surprising that doing more is associated with a longer visit. However, the findings from our study suggest that longer visits and more physician actions are associated with addressing multiple unrelated problems during the patient encounter, which provides a different perspective on the intensity of the physician’s work.^23-26

Factors that affect the duration of the visit are of interest to those who use physician productivity as a measure for making policy and management decisions. Primary care physician productivity is commonly defined as the number of patients seen per hour.^27,28 Such indicators of productivity would rate a physician who saw many patients in a short time productive, while a physician who provided care to fewer patients but addressed multiple problems would be viewed as less productive. This viewpoint overlooks the cost savings that may result from the reduced number of future visits the patient may require to address these problems, the enhanced quality of care that may be attributable to follow-up of previously identified health concerns, and the enhanced patient satisfaction that may result from the physician’s expanded approach. The current measures of productivity are crude and possibly misleading indicators of the work involved with providing comprehensive primary care to patients. Perhaps health service researchers and policymakers should reconsider the definition of productivity in light of the number of problems addressed or the number of physician actions necessary to address the problems during a patient visit.

Our findings also have implications for evaluating the quality of care provided by family physicians. The current narrowly diseased-focused assessments of quality care are limited because they neglect to take into account the wide range of competing multiple illnesses, prevention, and psychosocial and family context issues confronting family physicians. Quality indicators for primary care should also assess the degree to which family physicians are making the right choices about how to prioritize among the multiple problems that could be addressed during an outpatient visit.

In combination with other reports,²⁹ these data should caution the use of billing records as an indicator of the content of the visit. These data indicate that the billing record generally underrepresents the number of problems addressed during the visit. The lack of concordance between what was observed and what was billed may have several explanations. Underrecording on the billing sheet may be due to the lack of an adequate way to code some problems addressed. Some physicians may approach the completion of the billing sheet by documenting just enough to justify the time spent. Also, the mode of recording the billing (forms or computer programs) may limit the number of problems that can be recorded per visit. Nonconcordance may have also occurred if the physician made decisions about management of ongoing illnesses that were not overtly apparent to the observer.

Limitations

The generalizability of our findings is limited by the modest-sized convenience sample of physicians observed. The higher no-show rate by younger patients may have increased the number of problems seen per visit, since older patients tend to have more problems. However, the patient visits included in our study were randomly selected from all adult patient visits during the 6-week data collection period and were similar in sex to the few patients who were not observed and are likely to be reflective of the patients presenting for care. Although not assessed directly, inter-rater reliability among the 7 students was maximized through the use of videotaped patient encounters for practicing completing the data collection form and for calibrating the observers before data collection in the field.

Conclusions

Prioritizing and delivering a diverse array of services within a relationship context is a hallmark of family practice. Our data suggest that addressing multiple problems during a single outpatient visit is one important mechanism family physicians use to provide comprehensive care. The value of addressing multiple problems per visit in terms of patient satisfaction, cost, and quality of care deserves further investigation.

Acknowledgments

We are grateful to Catharine Symmonds, Catherine Bettcher, Elizabeth Welsh, Tracy Lemonovich, Robin Baines, and Sarah Younkin who contributed to the study design and data collection phase and without whose participation our study would not have been possible. William R. Phillips, MD, MPH, and Kurt C. Stange, MD, PhD, provided valuable suggestions on an earlier draft of this paper.

Related Resources

• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/CRFPPC.
• American Academy of Family Practice policy studies in family practice and primary care http://www.aafppolicy.org

Primary care disciplines continue to have a central role in the health care of Americans. They provide breadth of care within an ongoing relationship, bridging the boundaries between health and illness and guiding access to more narrowly focused care when needed.¹ The ability to orchestrate a broad health agenda during a visit is central to primary care, but this ability is challenged by competing demands for time.²

Attempts to influence provision of care and treatment decisions by primary care physicians, such as financial incentives, administrative restrictions, and the implementation of evidence-based clinical guidelines add to the demands on physicians’ time and may affect how time is allocated during the day and with each patient. Within this context a primary care physician must prioritize the agenda for each patient visit. This may include providing services beyond the patient’s primary reason for the visit as time permits, such as including preventive services,³ follow-up of acute or chronic illnesses,¹ mental health⁴ or family issues,^5-7 or investigating “by the way” patient comments that may indicate serious medical issues.

The competing demands for time are compounded by patient requests during the visit. Based on an audiotape of 139 patient encounters, Kravitz and colleagues⁸ reported that on average a patient makes 5 requests for physician action or information per visit, and the number of unfulfilled requests was negatively associated with patient satisfaction. Such findings may fuel a sense of pressure to address patient requests. Also, another recent report indicates that the majority of patients do not have the opportunity to express all of their concerns before the physician redirects the interview; once redirected, additional patient concerns are rarely elicited.⁹ Fitting both the physician’s and patient’s agenda into the time allotted for an outpatient visit has important implications for the duration of the visit, physician productivity, and possibly patient outcomes.

Data on the number of problems raised and addressed have been limited by the lack of appropriate collection methods. Primarily audio and video technology have been used for the study of physician-patient communication.^10-12 Direct observation of patient encounters^12,13 and incorporation of ethnographic approaches have more recently been employed to fill a large void in the understanding of the content, context, and complexity of primary care.^13-15 Findings from the Direct Observation of Primary Care study, which employed such methods, indicate that among 4454 patient visits care was provided to a secondary patient during 18% of the visits and preventive services were addressed during 32% of the illness visits.³ Data from that study provide a glimpse into some types of problems addressed in addition to the main reason for the visit; however, data about the number of problems addressed during patient encounters were not specifically collected by the nurse observer.

When additional issues are raised during a patient encounter, little is known about the nature of these problems, how additional problems affect the duration of the visit, and how well additional problems are reflected in the billing record. This led us to conduct an observational study to ask: How many problems are addressed during family practice outpatient visits, and who is raising additional problems? How much work and time is associated with addressing problems raised beyond the initial problem? How well does the billing list represent the number of problems addressed during the outpatient visit? Our study was designed to directly observe and record how many problems were raised and addressed during outpatient visits to family physicians.

Methods

Seven first-year medical students observed patient care provided by their summer fellowship family physician preceptor and other physicians in the preceptor’s practice from June through August 1999. Six of the sites were located in Northeast Ohio, and one was in Tulsa, Oklahoma.

Each student collected data on one randomly selected adult patient encounter for each half day of precepting. At the beginning of each half-day of patient care the student rolled a die to generate a random number to select a patient from the patient schedule. To ensure random selection of encounters within each half-day session, on alternating days the random number was counted from the beginning or the end of the half-day schedule. If the selected patient was aged younger than 18 years, the patient or physician preferred the encounter not be observed, or the patient did not show up for the scheduled appointment, the next scheduled appointment was selected as a replacement. Patient age and sex were collected for those who were no-shows or chose not to be observed, so they could be compared with those patients who were observed. Each student was to collect data on approximately 50 patient encounters during the 6-week summer fellowship. The physicians were blinded to the study purpose and were not told which patient encounter would be included in the study.

A problem was operationalized as an issue requiring physician action in the form of a decision, diagnosis, treatment, or monitoring. Each item was listed as it was raised, and the type of problem, who raised it, and what physician actions were involved to address it were coded. Each problem was coded as 1 of 14 categories: acute, acute follow-up, chronic, chronic follow-up, prevention, prevention follow-up, psychosocial, psychosocial follow-up, work-related administrative, health care system-related administrative, other family member’s problem, pregnancy, emergent, and other. The person who raised the problem was coded as 1 of 3 options: the physician, the patient or another person in the room. Multiple physician actions could be coded for how the problem was addressed. The 19 physician action categories included: question, reassurance, examination, procedure, referral, return visit, advice, review tests, order laboratory testing, prescription, provide written material, imaging, admits uncertainty, counseling, return to work/time off work letter, defer, complementary/alternative medicine, ignored or lost, and other.

Patient characteristics, the duration of the visit, and the billing diagnoses for each visit were also recorded on the data collection form. Videotaped encounters were used to pilot test the data collection form, to allow the observers to practice using the form in real time, and to calibrate the observers before data collection in the field.

We used descriptive statistics to address most research questions. Student t tests and chi-square tests were used to compare age and sex differences between participants and nonparticipants. We tested the association of the number of problems with the duration of the visit with analysis of variance and a test for linear trend. A difference score of the number of problems observed and the number of problems recorded on the billing sheet for the encounter was computed and summarized graphically.

Results

We collected usable data on 266 encounters representing 37 physicians. Patient and visit characteristics are displayed in Table 1. The patients had an average age of 48 years, and 69% were women. They were predominately white. A large proportion was observed visiting their regular primary care physician (83%), and 85% were established patients of the practice. Most of the observed patients had some kind of commercial health care insurance, 19% had Medicare, and a small proportion had Medicaid or no insurance. The visit duration ranged from 2 to 65 minutes; the median was 15 minutes with a mean of 19.3 (standard deviation [SD]=12.7). The first problem raised was most commonly an acute problem (49%); prevention and chronic illness were the first problem raised during 21% and 19% of encounters, respectively. Patients who were randomly selected but were not observed (n=52, primarily no-shows) were similar in sex (67% women, c2 =0.119, P=.73 ) but were younger than those patients who were observed (mean age=32.1 years, t=3.79, P=.001).

On average, 2.7 problems were raised during an encounter Table 2. Forty-four percent of all problems were classified as acute, 30% chronic, 14% prevention, 4% administrative, 2% psychosocial, and 6% were classified as other. Of the observed encounters, 73% had more than one problem addressed. The physician raised 36% of these additional problems, and patients raised 58%. The problems raised by physicians were most frequently pertaining to chronic illness, prevention, and follow-up issues. The problems raised by patients were most likely to be acute illness problems. Additional problems were least likely to arise when the first problem addressed was an acute problem (61%) compared with visits during which the first problem addressed was chronic or prevention focused, where 88% and 87%, respectively, included additional problems during the visit (c2=21.2, P <.001).

On average, 8 (SD=4.5) physician actions were observed per encounter Table 2. Physicians performed an average of 3.3 (SD=1.2) actions per problem. The most common physician actions were questioning (77%), physical examination (49%), prescription writing (32%), providing advice (31%), and reassurance (25%). Of the 452 additional problems raised, only 3% of problems were ignored, and 6% were deferred to another visit.

The association of the number of problems addressed with the duration of the visit was assessed by analysis of variance and a test for linear trend. As shown in Figure 1, the duration of the visit increased approximately 2.5 minutes for each additional problem addressed (P <.001 for linear trend). The visit duration within each of the number of problem groups varied greatly as indicated by the large range for each group; however, the SD for each of the groups as indicated by the shaded bars are a similar size for each of the groups (Levene’s test of equality of error variance=1.48, P=.195).

The concordance between the number of problems observed and the number of problems on the billing sheet was modest, with a trend toward billing for fewer problems than were observed. As shown in Figure 2, 29% of encounters represented a match between the number of problems observed and the number of problems on the billing sheet. Fifty-eight percent of the encounters had more problems observed than recorded on the billing sheet. A much smaller proportion of encounters recorded more problems on the billing sheet than were observed during the encounter.

Discussion

Our exploratory study suggests that it is common for multiple problems to be addressed during visits to a family physician regardless of the initial reason for the visit. Additional problems are raised by both physicians and patients and are rarely deferred or ignored by the physician. Although the phenomenon of integrating a broad health agenda and addressing multiple problems during a single outpatient visit may be well known by practicing community-based family physicians, it may not be recognized by policymakers or health services researchers whose window into the process of outpatient care is provided by the medical record and billing data.

Addressing the majority of a patient’s health care needs and providing comprehensive care is a core feature of quality primary care.^16-20 Previous work has documented the wide range of diagnoses and clusters of diagnoses that family physicians commonly address during outpatient care.^13,21 However, truly comprehensive care goes beyond providing a broad array of services; it also involves the integration of care in a physician-patient relationship context. Prioritizing, providing, and orchestrating care for acute and undifferentiated illness, chronic disease, preventive services, and mental health care represents a key feature of primary care practice such that the care is greater than the sum of its individual commodities.¹ These data suggest that single visits often address a broad agenda of health care.

Overall, as the number of problems increase so does the length of the visit. Others have found that ordering or performing more tests, providing preventive services, and conducting ambulatory surgical procedures increase the length of the visit.²² It is not surprising that doing more is associated with a longer visit. However, the findings from our study suggest that longer visits and more physician actions are associated with addressing multiple unrelated problems during the patient encounter, which provides a different perspective on the intensity of the physician’s work.^23-26

Factors that affect the duration of the visit are of interest to those who use physician productivity as a measure for making policy and management decisions. Primary care physician productivity is commonly defined as the number of patients seen per hour.^27,28 Such indicators of productivity would rate a physician who saw many patients in a short time productive, while a physician who provided care to fewer patients but addressed multiple problems would be viewed as less productive. This viewpoint overlooks the cost savings that may result from the reduced number of future visits the patient may require to address these problems, the enhanced quality of care that may be attributable to follow-up of previously identified health concerns, and the enhanced patient satisfaction that may result from the physician’s expanded approach. The current measures of productivity are crude and possibly misleading indicators of the work involved with providing comprehensive primary care to patients. Perhaps health service researchers and policymakers should reconsider the definition of productivity in light of the number of problems addressed or the number of physician actions necessary to address the problems during a patient visit.

Our findings also have implications for evaluating the quality of care provided by family physicians. The current narrowly diseased-focused assessments of quality care are limited because they neglect to take into account the wide range of competing multiple illnesses, prevention, and psychosocial and family context issues confronting family physicians. Quality indicators for primary care should also assess the degree to which family physicians are making the right choices about how to prioritize among the multiple problems that could be addressed during an outpatient visit.

In combination with other reports,²⁹ these data should caution the use of billing records as an indicator of the content of the visit. These data indicate that the billing record generally underrepresents the number of problems addressed during the visit. The lack of concordance between what was observed and what was billed may have several explanations. Underrecording on the billing sheet may be due to the lack of an adequate way to code some problems addressed. Some physicians may approach the completion of the billing sheet by documenting just enough to justify the time spent. Also, the mode of recording the billing (forms or computer programs) may limit the number of problems that can be recorded per visit. Nonconcordance may have also occurred if the physician made decisions about management of ongoing illnesses that were not overtly apparent to the observer.

Limitations

The generalizability of our findings is limited by the modest-sized convenience sample of physicians observed. The higher no-show rate by younger patients may have increased the number of problems seen per visit, since older patients tend to have more problems. However, the patient visits included in our study were randomly selected from all adult patient visits during the 6-week data collection period and were similar in sex to the few patients who were not observed and are likely to be reflective of the patients presenting for care. Although not assessed directly, inter-rater reliability among the 7 students was maximized through the use of videotaped patient encounters for practicing completing the data collection form and for calibrating the observers before data collection in the field.

Conclusions

Prioritizing and delivering a diverse array of services within a relationship context is a hallmark of family practice. Our data suggest that addressing multiple problems during a single outpatient visit is one important mechanism family physicians use to provide comprehensive care. The value of addressing multiple problems per visit in terms of patient satisfaction, cost, and quality of care deserves further investigation.

Acknowledgments

We are grateful to Catharine Symmonds, Catherine Bettcher, Elizabeth Welsh, Tracy Lemonovich, Robin Baines, and Sarah Younkin who contributed to the study design and data collection phase and without whose participation our study would not have been possible. William R. Phillips, MD, MPH, and Kurt C. Stange, MD, PhD, provided valuable suggestions on an earlier draft of this paper.

Related Resources

• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/CRFPPC.
• American Academy of Family Practice policy studies in family practice and primary care http://www.aafppolicy.org

Primary care disciplines continue to have a central role in the health care of Americans. They provide breadth of care within an ongoing relationship, bridging the boundaries between health and illness and guiding access to more narrowly focused care when needed.¹ The ability to orchestrate a broad health agenda during a visit is central to primary care, but this ability is challenged by competing demands for time.²

Attempts to influence provision of care and treatment decisions by primary care physicians, such as financial incentives, administrative restrictions, and the implementation of evidence-based clinical guidelines add to the demands on physicians’ time and may affect how time is allocated during the day and with each patient. Within this context a primary care physician must prioritize the agenda for each patient visit. This may include providing services beyond the patient’s primary reason for the visit as time permits, such as including preventive services,³ follow-up of acute or chronic illnesses,¹ mental health⁴ or family issues,^5-7 or investigating “by the way” patient comments that may indicate serious medical issues.

The competing demands for time are compounded by patient requests during the visit. Based on an audiotape of 139 patient encounters, Kravitz and colleagues⁸ reported that on average a patient makes 5 requests for physician action or information per visit, and the number of unfulfilled requests was negatively associated with patient satisfaction. Such findings may fuel a sense of pressure to address patient requests. Also, another recent report indicates that the majority of patients do not have the opportunity to express all of their concerns before the physician redirects the interview; once redirected, additional patient concerns are rarely elicited.⁹ Fitting both the physician’s and patient’s agenda into the time allotted for an outpatient visit has important implications for the duration of the visit, physician productivity, and possibly patient outcomes.

Data on the number of problems raised and addressed have been limited by the lack of appropriate collection methods. Primarily audio and video technology have been used for the study of physician-patient communication.^10-12 Direct observation of patient encounters^12,13 and incorporation of ethnographic approaches have more recently been employed to fill a large void in the understanding of the content, context, and complexity of primary care.^13-15 Findings from the Direct Observation of Primary Care study, which employed such methods, indicate that among 4454 patient visits care was provided to a secondary patient during 18% of the visits and preventive services were addressed during 32% of the illness visits.³ Data from that study provide a glimpse into some types of problems addressed in addition to the main reason for the visit; however, data about the number of problems addressed during patient encounters were not specifically collected by the nurse observer.

When additional issues are raised during a patient encounter, little is known about the nature of these problems, how additional problems affect the duration of the visit, and how well additional problems are reflected in the billing record. This led us to conduct an observational study to ask: How many problems are addressed during family practice outpatient visits, and who is raising additional problems? How much work and time is associated with addressing problems raised beyond the initial problem? How well does the billing list represent the number of problems addressed during the outpatient visit? Our study was designed to directly observe and record how many problems were raised and addressed during outpatient visits to family physicians.

Methods

Seven first-year medical students observed patient care provided by their summer fellowship family physician preceptor and other physicians in the preceptor’s practice from June through August 1999. Six of the sites were located in Northeast Ohio, and one was in Tulsa, Oklahoma.

Each student collected data on one randomly selected adult patient encounter for each half day of precepting. At the beginning of each half-day of patient care the student rolled a die to generate a random number to select a patient from the patient schedule. To ensure random selection of encounters within each half-day session, on alternating days the random number was counted from the beginning or the end of the half-day schedule. If the selected patient was aged younger than 18 years, the patient or physician preferred the encounter not be observed, or the patient did not show up for the scheduled appointment, the next scheduled appointment was selected as a replacement. Patient age and sex were collected for those who were no-shows or chose not to be observed, so they could be compared with those patients who were observed. Each student was to collect data on approximately 50 patient encounters during the 6-week summer fellowship. The physicians were blinded to the study purpose and were not told which patient encounter would be included in the study.

A problem was operationalized as an issue requiring physician action in the form of a decision, diagnosis, treatment, or monitoring. Each item was listed as it was raised, and the type of problem, who raised it, and what physician actions were involved to address it were coded. Each problem was coded as 1 of 14 categories: acute, acute follow-up, chronic, chronic follow-up, prevention, prevention follow-up, psychosocial, psychosocial follow-up, work-related administrative, health care system-related administrative, other family member’s problem, pregnancy, emergent, and other. The person who raised the problem was coded as 1 of 3 options: the physician, the patient or another person in the room. Multiple physician actions could be coded for how the problem was addressed. The 19 physician action categories included: question, reassurance, examination, procedure, referral, return visit, advice, review tests, order laboratory testing, prescription, provide written material, imaging, admits uncertainty, counseling, return to work/time off work letter, defer, complementary/alternative medicine, ignored or lost, and other.

Patient characteristics, the duration of the visit, and the billing diagnoses for each visit were also recorded on the data collection form. Videotaped encounters were used to pilot test the data collection form, to allow the observers to practice using the form in real time, and to calibrate the observers before data collection in the field.

We used descriptive statistics to address most research questions. Student t tests and chi-square tests were used to compare age and sex differences between participants and nonparticipants. We tested the association of the number of problems with the duration of the visit with analysis of variance and a test for linear trend. A difference score of the number of problems observed and the number of problems recorded on the billing sheet for the encounter was computed and summarized graphically.

Results

We collected usable data on 266 encounters representing 37 physicians. Patient and visit characteristics are displayed in Table 1. The patients had an average age of 48 years, and 69% were women. They were predominately white. A large proportion was observed visiting their regular primary care physician (83%), and 85% were established patients of the practice. Most of the observed patients had some kind of commercial health care insurance, 19% had Medicare, and a small proportion had Medicaid or no insurance. The visit duration ranged from 2 to 65 minutes; the median was 15 minutes with a mean of 19.3 (standard deviation [SD]=12.7). The first problem raised was most commonly an acute problem (49%); prevention and chronic illness were the first problem raised during 21% and 19% of encounters, respectively. Patients who were randomly selected but were not observed (n=52, primarily no-shows) were similar in sex (67% women, c2 =0.119, P=.73 ) but were younger than those patients who were observed (mean age=32.1 years, t=3.79, P=.001).

On average, 2.7 problems were raised during an encounter Table 2. Forty-four percent of all problems were classified as acute, 30% chronic, 14% prevention, 4% administrative, 2% psychosocial, and 6% were classified as other. Of the observed encounters, 73% had more than one problem addressed. The physician raised 36% of these additional problems, and patients raised 58%. The problems raised by physicians were most frequently pertaining to chronic illness, prevention, and follow-up issues. The problems raised by patients were most likely to be acute illness problems. Additional problems were least likely to arise when the first problem addressed was an acute problem (61%) compared with visits during which the first problem addressed was chronic or prevention focused, where 88% and 87%, respectively, included additional problems during the visit (c2=21.2, P <.001).

On average, 8 (SD=4.5) physician actions were observed per encounter Table 2. Physicians performed an average of 3.3 (SD=1.2) actions per problem. The most common physician actions were questioning (77%), physical examination (49%), prescription writing (32%), providing advice (31%), and reassurance (25%). Of the 452 additional problems raised, only 3% of problems were ignored, and 6% were deferred to another visit.

The association of the number of problems addressed with the duration of the visit was assessed by analysis of variance and a test for linear trend. As shown in Figure 1, the duration of the visit increased approximately 2.5 minutes for each additional problem addressed (P <.001 for linear trend). The visit duration within each of the number of problem groups varied greatly as indicated by the large range for each group; however, the SD for each of the groups as indicated by the shaded bars are a similar size for each of the groups (Levene’s test of equality of error variance=1.48, P=.195).

The concordance between the number of problems observed and the number of problems on the billing sheet was modest, with a trend toward billing for fewer problems than were observed. As shown in Figure 2, 29% of encounters represented a match between the number of problems observed and the number of problems on the billing sheet. Fifty-eight percent of the encounters had more problems observed than recorded on the billing sheet. A much smaller proportion of encounters recorded more problems on the billing sheet than were observed during the encounter.

Discussion

Our exploratory study suggests that it is common for multiple problems to be addressed during visits to a family physician regardless of the initial reason for the visit. Additional problems are raised by both physicians and patients and are rarely deferred or ignored by the physician. Although the phenomenon of integrating a broad health agenda and addressing multiple problems during a single outpatient visit may be well known by practicing community-based family physicians, it may not be recognized by policymakers or health services researchers whose window into the process of outpatient care is provided by the medical record and billing data.

Addressing the majority of a patient’s health care needs and providing comprehensive care is a core feature of quality primary care.^16-20 Previous work has documented the wide range of diagnoses and clusters of diagnoses that family physicians commonly address during outpatient care.^13,21 However, truly comprehensive care goes beyond providing a broad array of services; it also involves the integration of care in a physician-patient relationship context. Prioritizing, providing, and orchestrating care for acute and undifferentiated illness, chronic disease, preventive services, and mental health care represents a key feature of primary care practice such that the care is greater than the sum of its individual commodities.¹ These data suggest that single visits often address a broad agenda of health care.

Overall, as the number of problems increase so does the length of the visit. Others have found that ordering or performing more tests, providing preventive services, and conducting ambulatory surgical procedures increase the length of the visit.²² It is not surprising that doing more is associated with a longer visit. However, the findings from our study suggest that longer visits and more physician actions are associated with addressing multiple unrelated problems during the patient encounter, which provides a different perspective on the intensity of the physician’s work.^23-26

Factors that affect the duration of the visit are of interest to those who use physician productivity as a measure for making policy and management decisions. Primary care physician productivity is commonly defined as the number of patients seen per hour.^27,28 Such indicators of productivity would rate a physician who saw many patients in a short time productive, while a physician who provided care to fewer patients but addressed multiple problems would be viewed as less productive. This viewpoint overlooks the cost savings that may result from the reduced number of future visits the patient may require to address these problems, the enhanced quality of care that may be attributable to follow-up of previously identified health concerns, and the enhanced patient satisfaction that may result from the physician’s expanded approach. The current measures of productivity are crude and possibly misleading indicators of the work involved with providing comprehensive primary care to patients. Perhaps health service researchers and policymakers should reconsider the definition of productivity in light of the number of problems addressed or the number of physician actions necessary to address the problems during a patient visit.

Our findings also have implications for evaluating the quality of care provided by family physicians. The current narrowly diseased-focused assessments of quality care are limited because they neglect to take into account the wide range of competing multiple illnesses, prevention, and psychosocial and family context issues confronting family physicians. Quality indicators for primary care should also assess the degree to which family physicians are making the right choices about how to prioritize among the multiple problems that could be addressed during an outpatient visit.

In combination with other reports,²⁹ these data should caution the use of billing records as an indicator of the content of the visit. These data indicate that the billing record generally underrepresents the number of problems addressed during the visit. The lack of concordance between what was observed and what was billed may have several explanations. Underrecording on the billing sheet may be due to the lack of an adequate way to code some problems addressed. Some physicians may approach the completion of the billing sheet by documenting just enough to justify the time spent. Also, the mode of recording the billing (forms or computer programs) may limit the number of problems that can be recorded per visit. Nonconcordance may have also occurred if the physician made decisions about management of ongoing illnesses that were not overtly apparent to the observer.

Limitations

The generalizability of our findings is limited by the modest-sized convenience sample of physicians observed. The higher no-show rate by younger patients may have increased the number of problems seen per visit, since older patients tend to have more problems. However, the patient visits included in our study were randomly selected from all adult patient visits during the 6-week data collection period and were similar in sex to the few patients who were not observed and are likely to be reflective of the patients presenting for care. Although not assessed directly, inter-rater reliability among the 7 students was maximized through the use of videotaped patient encounters for practicing completing the data collection form and for calibrating the observers before data collection in the field.

Conclusions

Prioritizing and delivering a diverse array of services within a relationship context is a hallmark of family practice. Our data suggest that addressing multiple problems during a single outpatient visit is one important mechanism family physicians use to provide comprehensive care. The value of addressing multiple problems per visit in terms of patient satisfaction, cost, and quality of care deserves further investigation.

Acknowledgments

We are grateful to Catharine Symmonds, Catherine Bettcher, Elizabeth Welsh, Tracy Lemonovich, Robin Baines, and Sarah Younkin who contributed to the study design and data collection phase and without whose participation our study would not have been possible. William R. Phillips, MD, MPH, and Kurt C. Stange, MD, PhD, provided valuable suggestions on an earlier draft of this paper.

Related Resources

• Center for Research in Family Practice and Primary Care http://mediswww.cwru.edu/dept/CRFPPC.
• American Academy of Family Practice policy studies in family practice and primary care http://www.aafppolicy.org

Managed care has become the predominant approach to health care financing in the United States.1 This explosive growth has been accompanied by an increasingly complex array of types of managed care plans and a growth in the use of restrictions and financial incentives to influence physician practice behavior.2 Contributing to the diversity and complexity of managed care are new incentive systems, strategies to manage patterns of care,3,4 and a shift toward national investor-owned plans. In addition, many practicing physicians are participating in new business relationships, including physician hospital organizations, medical service organizations, and risk-sharing arrangements. The treatment of all managed care plans as a single entity for comparison with fee-for-service plans is no longer adequate to capture the effect of the health care context on the delivery of care or health outcomes.5 A typology of features that represents a plan’s organizational and incentive features would facilitate understanding of what specific aspects affect outcomes of care across the nation.6,7 The Managed Care Survey was developed for use in this study to build on previous work by measuring specific attributes of different managed care plans that may affect both physician and patient outcomes.

Managed care organizations traditionally position primary care clinicians as the cornerstones of their delivery system⁸; however, the effect of the restrictiveness of managed care plans on the patient-physician relationship and the delivery of important attributes of primary care (as described by the Institute of Medicine⁹) are poorly understood. A common assumption is that managed care fosters primary care because of its gatekeeper and first-contact functions.⁸ Since health systems organized around primary care have been shown to have better population-level outcomes,¹⁰ one might expect similar results from managed care systems.^11-13 Several aspects of the current competitive managed care marketplace, however, do not appear to be conducive to achieving the goals of primary care.^14,15 The restrictions on clinicians’ and patients’ choices have raised concerns about the potential detrimental effect of managed care on the patient-physician relationship.¹⁶ In addition, the practice of annual re-bidding of managed care contracts can cause a forced disruption in continuity of care¹⁷ with detrimental effects on patients.^15,18 For these reasons, the Institute of Medicine recommends monitoring the performance of health care systems to assess the adequacy of the delivery of attributes of primary care.⁹

The restrictions and incentives imposed by managed care organizations that are designed to modify physician practice behavior may inadvertently effect other valued aspects of patient care. The purpose of our study is to evaluate the association of managed care restrictiveness with specific attributes of primary care, visit-based patient satisfaction, and perceived inconvenience (or “hassle”) of using the plan.

Methods

Study Design, Sites, and Sample

A cross-sectional design was used to collect data from 15 member practices of the Ambulatory Sentinel Practice Network (ASPN). ASPN, composed of 752 community-based practicing clinicians, was established in 1982 to conduct practice-based research. Its 122 practices in 34 states have been shown to serve a nationally representative patient population and provide access to health care markets with a wide range of penetration and maturity of managed care.¹⁹ We solicited volunteer practice sites and chose 15 US ASPN sites to represent high-, medium-, and low-levels of managed care penetration in both urban and rural areas. Clinician and practice characteristics, including the clinicians’ estimate of the proportion of managed care in the practice, were obtained from the ASPN member database, which is updated annually. All 15 sites that were invited to participate in the study agreed to complete it.

Data Collection

The ASPN central office recruited and trained participating practice personnel and coordinated project implementation. Between April and August 1997, practices were sent explicit protocol instructions and copies of the Managed Care Survey and the patient survey and were instructed to choose a start date for administering surveys to 50 consecutive patients of each participating clinician. Patients were asked to complete the survey before leaving the office. A preassigned number corresponding to the patient’s insurance plan was written on the survey before it was given to the patient. The staff also kept track of the age, sex, and insurance plan of those patients who declined to participate, so that any nonparticipation bias could be evaluated.

In addition, for each practice a single Managed Care Survey was completed jointly by a physician and office manager to characterize each individual managed care health care plan with a minimum of 5% of all patients. Each insurance plan on the survey was identified by the same number that was used on the corresponding individual patients’ health care plans on the patient survey.

Measures

Managed care was conceptualized as a set of organizational restraints and financial incentives that are intended to focus and limit clinicians’ use of health care resources. The Managed Care Survey was designed to characterize managed care plans along several dimensions. The survey was developed by the ASPN Task Force on Managed Care, which consisted of 6 family physicians from the United States and 1 from Canada representing diversity in gender, geographic location, organization of practice, managed care market, and years in practice. Group consensus was used to identify and define the key managed care features that represent organizational restraints, financial incentives, and other aspects affecting the restrictiveness of plans.

The plan features measured by the managed care survey included the proportion of each practice’s patients in the plan, the plan’s financial restrictiveness and organizational restrictiveness, and the level of hassle associated with it. The financial restrictiveness portion of the survey included the type of reimbursement (capitation global risk, capitation professional risk, capitation primary care risk, discounted fee-for-service, or fee-for-service) and whether the plan carried a clinician-withhold fund or an incentive-bonus fund. The organizational restrictiveness part of the survey included plan characterization on the following features: mental health carve-out, laboratory services, formulary, preauthorization for diagnostic or treatment procedures, preauthorization for physician referrals, specialty network, and procedure (site of service) constraints. The managed care survey features and their definitions are listed in Appendix A.*

Some plan features were viewed as more important than others for describing a plan’s financial and organizational restrictiveness. Each member of the ASPN Managed Care Task Force assigned a value of importance for each feature (using a Likert scale where 1 = somewhat important; 10 = very important). The group mean assigned weight for each feature was used to calculate the 2 weighted summary scores representing the financial and organizational restrictiveness of each plan.

In addition to the managed care plan features, clinicians completing the survey were asked to rate the degree of hassle, defined as the degree of time-consuming interference with routine practice activities perceived to be associated with the plan, on a scale of 1 to 5. Additional items on the Managed Care Survey included questions about the type of practice (solo, multispecialty group, and so forth), political/business affiliations, recent mergers/buyouts, and type of clinician compensation.

The attributes of delivery of primary care were measured by the revised Components of Primary Care Instrument (CPCI)^20,21 which measures key attributes of the patient-provider relationship based on the recent Institute of Medicine definition of primary care.⁹ The CPCI assesses interpersonal communication, comprehensive care, continuity of care, coordination of care, provider’s accumulated knowledge about the patient, family orientation, community orientation, advocacy, and patient preference for their usual provider. Each attribute is measured from the patient’s perspective of the patient-provider relationship. Descriptive statistics, internal consistency reliabilities, and scale content are displayed in Appendix B.* Missing data on the CPCI scale scores were handled by setting a maximum number of missing values allowed per scale and computing a score using individual responses to the remaining scale items. Questionnaires missing more data than the maximum allowed were given no score for that scale. Therefore, the total number of patients with complete data per scale, and the sample size for analyses, varies by scale.

Patient satisfaction with the visit was measured using the Medical Outcomes Study 9-Item Visit Rating Form.²² Two scores were computed, patient satisfaction with the physician and with practice operation.²³ Another specific patient item (satisfaction with the amount of time spent with the physician) was also assessed separately. Eight items were written to assess the patient’s perceived hassle in obtaining health care, and a summary scale score was computed (internal consistency reliability = .80). Additional items on the patient survey included patient age, sex, 2 reports of health status, whether today’s physician is the patient’s regular physician, and if no, whether the patient’s regular physician is a member of the office. Standard demographic items were included on the abbreviated survey for new patients. Also included were questions regarding whether that visit was for well-care or serious illness, and whether they had been forced to change physicians in the past 2 years.

Analyses

Data from the Managed Care Survey and the patient survey were linked by a unique identifier on the basis of site and site-specific health care plan. Patient surveys that could not be linked (eg, because of a missing plan identifier or because the plan was not rated on the Managed Care Survey) were included in the descriptive statistics of the study sample, but are excluded from the analyses involving the managed care features. Descriptive statistics of the sites, clinicians, and plan features are calculated. We used chi-square tests to compare the data available from the nonresponders with the data from the responders to assess bias.

We used the Pearson correlation to test the association of the plans’ financial and organizational restrictiveness scores with the clinicians’ reports of hassle. The association of the managed care plan restrictiveness scores with each of the CPCI scale scores, patient satisfaction with the visit, and patient perceived hassle was tested with multilevel modeling techniques using hierarchical linear regression software.²⁴ Multilevel modeling is an analysis technique that accounts for the nested structural context of the data. Two potential confounding variables, patient age and health status, were included as covariates in these analyses.

Results

All 15 sites returned a completed Managed Care Survey. Practice characteristics are displayed in Table 1. One fourth of the practices had experienced a recent professional merger and one third had undergone a recent purchase or buyout. The average number of managed care plans in each practice ranged from 1 to 25, with an average of 13.3 plans. The average proportion of patients in a managed care plan per site was 51% (range = 21% to 100%). The 41 clinicians participating in the study are characterized in Table 2. The vast majority of clinicians had MD degrees, and 66% were men. On average, clinicians spent 90% of their time on patient care.

The patient response rate was also excellent. Of the 1922 patients approached, 1839 (96%) agreed to complete the patient survey. One hundred and six patients returned a blank survey and represent passive refusers. Of the 1733 patients returning a survey at least partially completed, 1503 were established patients, and 230 were new patients. Twenty-eight established patients did not see their regular physician, and that physician was not a member of the office they were visiting that day. These patients were excluded, bringing the final patient sample size to 1475.

Patient characteristics are reported in Table 3. The majority of patients were women, and health status, on average, was good. Most patients (84%) saw their regular physician, approximately half had a well-care visit within the past 2 years, and approximately one fourth were treated for a serious illness within the past 2 years. Fifty-nine percent of established patients had some type of managed care insurance. Standard Medicare and Medicaid insurance accounted for 21% of patients, and only 11% were categorized as having traditional commercial insurance. Established patients who declined to complete the patient survey (n = 41) were similar in average age and type of insurance but were more likely to be men than the patients who completed the survey.

Table 1, Table 4 displays the frequency of the different managed care features measured by the Managed Care Survey. Laboratory services, preauthorization, specialty networks, and site of service were features of more than 50% of the 199 managed care plans characterized. Physicians rated plans with a restrictive feature as generating greater hassle on average than plans without restrictive features. The 2 exceptions to this trend were plans with point-of-service and withhold features.

We investigated the association of managed care plan restrictiveness with each of the CPCI scale scores and patient satisfaction with the visit. Of the 870 patients with a type of managed care insurance, 786 patients had complete data for this analysis. For ease of interpretation, the managed care restrictiveness scores were divided into low (27%), medium (46%), and high (27%).²⁵ This categorization of the restrictiveness scores has 3 advantages: interpretation of 3 group means versus a b coefficient is easier; a nonlinear association is readily determined; and the distributions of the outcome measure for the high and low groups can be shown to be nonoverlapping. If no statistically significant difference is found between these 2 distinctly different extreme groups, this can be taken as evidence for not rejecting the null hypothesis of no association.

As indicated in Table 5, the mean of the different CPCI, hassle, and patient satisfaction scores were very similar across each level of managed care plan financial restrictiveness. Similarly, organizational restrictiveness was not significantly associated with any of the CPCI scale scores, patient report of hassle, or the satisfaction scores. These analyses were adjusted for patient age and health status and the nested effect of the data.

Discussion

We used innovative measures and a unique practice-based laboratory to assess the impact of specific aspects of managed care on the delivery of important attributes of primary care. The findings suggest that the restrictiveness of managed care plans does not affect patients’ perceptions of multiple attributes of primary care or their satisfaction with the visit. However, both financial and organizational restrictiveness were associated with greater clinician-reported hassle. These findings may not conform to the widespread belief by practicing clinicians that plan characteristics affect patients in a direct way,^8,26 and clinicians may be reassured to find that they are able to maintain good primary care relationships with patients amidst the challenges they experience.

Clinician hassle was rated in terms of the time required for insurance-mandated administrative activities generated by the plan (eg, the length and repetition of required forms and written or verbal requirements). Other studies have reported specific physician-reported hassles associated with particular plans.^8,27 Most of the hassles can be attributed to an added administrative burden, such as the need to make phone calls, write letters, and gather information from medical records in response to denial of payment, requests for patient information, or precertification of services.²⁷ Our findings that clinicians reported increased administrative burden with more restrictive plans reinforce the idea by Freberg²⁸ that it is difficult to know whether managed care plans are cost effective or merely add to the hidden cost of administrative overhead. Future studies should investigate the amount of effort required for additional administrative burdens relative to the cost savings of the plan.

The lack of association between patient-perceived hassle and plan restrictiveness indicates that the burdens of plan restrictiveness fall squarely on the shoulders of clinicians and staff. It is also likely that patients who are less concerned about plan restrictions may have self-selected a restrictive plan for cost savings or other perceived benefits. For these patients, the benefits (eg, lower deductibles, coverage of health maintenance visits) may outweigh the disadvantages (eg, restricted freedom of choice, increased personal cost incurred to opt for out-of-plan services).

Concerns have been raised about conflict of interest, the effect of financial incentives on physician behavior, the quality of the patient-physician relationship and decision making, time constraints, and the potential for underservice with managed care systems.^16,26,29-34 Grumbach and colleagues²⁶ found that 57% of physicians surveyed reported that they felt pressure from the managed care organization to limit referrals; 75% felt pressure to see more patients per day; and 17% and 24%, respectively, felt that limiting referrals and seeing more patients per day compromised patient care. In our study, patients in highly restrictive managed care plans did not perceive their physician to be any less of an advocate for their health care than patients in the low- or medium-restrictiveness groups. Thus, this sample of primary care clinicians continued to engage in trusting relationships with their patients despite the potential conflict of interest that could arise from managed care plans’ financial incentives to restrict care.

In our sample of patients, as well as in others,²¹ the CPCI assessed important aspects of primary care with good internal consistency. The instrument’s scale scores have been shown to be associated with patient satisfaction²¹ and delivery of preventive services,³⁵ and have been shown to detect differences in the delivery of primary care to patients who faced forced discontinuity of care and those who remained with their regular physician.¹⁵ The CPCI should be sensitive to many of the potential ill effects of managed care on the patient-physician relationship and delivery of primary care. The lack of association between plan restrictiveness and patient report of primary care is striking, and there is strong evidence that the clinicians and office staff who report being hassled by these restrictions are not allowing those hassles to interfere with their delivery of patient care.

Others²⁶ have evaluated physician satisfaction with specific plan features and physician-rated quality of specific health care plans.³⁶ We asked physicians to objectively report the presence or absence of specific organizational and financial features of each of the managed care plans in their practices. Using the Managed Care Survey to characterize specific organizational and financial aspects of plans is a major advance in being able to test the importance of these features on physician behavior and processes of care and patient outcomes.

Limitations

The main potential threat to the internal validity of the study is patient nonresponse. The nonrespondents were more likely to be men than patients who completed the survey. It is possible that these patients may have been less satisfied with care and may have reported lower scores on the CPCI. However, nonrespondents represent only 10% of those approached, and it is unlikely that the findings of the study would have changed if they had been included. We are also unable to comment on how long a patient had been with their current insurance plan. Length of exposure and actual experience with the features of a plan could potentially affect the association of plan restrictiveness with perceived delivery of primary care. However, consecutive patients were enrolled, which should reduce the likelihood of a selection bias of such a variable.

Replication of this study in a larger number of community-based practice sites and in a general community sample would add to the generalizability of the findings. In these times of increased business interest in medicine,³⁷ it is important to continue to monitor and evaluate the immediate, long-term, intended, and unintended outcomes of specific features of managed care.

Conclusions

Managed care plan restrictiveness does not appear to be affecting the delivery of primary care as measured from the perspective of the patient. However, the financial and organizational restrictiveness of managed care plans does lead to greater clinician hassle. We interpret these results to suggest that primary care clinicians are able to effectively buffer the effects of health plan structure on their patients. These findings raise questions about the effect of plan restrictiveness on efficient use of clinician time and the clinician’s ability to continue to deliver quality primary care amidst competing administrative demands.

Acknowledgments

We would like to thank each of the participating ASPN practices, their staffs, and patients, without whom the study would not have been possible. The practices participating in this study included the following: Batesville Family Practice Clinic, Batesville, Arkansas; Loma Linda University, Department of Family Medicine, Loma Linda, California; C. Frazer Shipman, Wheatridge, Colorado; The Family Medical Group, Bristol, Connecticut; St. John’s Mercy Family Medicine, St. Louis, Missouri; Manchester Family Health Center, Manchester, New Hampshire; Primary Care Center at Hillsborough, Belle Mead, New Jersey, and Family Medicine at Monument Square, New Brunswick, New Jersey; Central Square Health Services Center, Central Square, New York; Enid Family Medicine Clinic, Enid, Oklahoma; Good Samaritan Family Practice, ELCO, Lebanon, Pennsylvania; Michael Hartsell, MD, Greeneville, Tennessee; Annadale Family Medicine, PC, Annadale, Virginia, and Tappahannack Family Practice, Tappahannack, Virginia; and Cle Elum Family Medicine Center, Cle Elum, Washington.

Thanks to each of the ASPN Managed Care Task Force members who developed the Managed Care Survey and pilot tested the project instruments and protocol: A. John Orzano, MD; H. Andrew Selinger, MD; Robert James, MD; William Fosmire, MD; Linda French, MD; Frank Reed, MD; John Scott, MD; and Dennis de Leon, MD.

Managed care has become the predominant approach to health care financing in the United States.1 This explosive growth has been accompanied by an increasingly complex array of types of managed care plans and a growth in the use of restrictions and financial incentives to influence physician practice behavior.2 Contributing to the diversity and complexity of managed care are new incentive systems, strategies to manage patterns of care,3,4 and a shift toward national investor-owned plans. In addition, many practicing physicians are participating in new business relationships, including physician hospital organizations, medical service organizations, and risk-sharing arrangements. The treatment of all managed care plans as a single entity for comparison with fee-for-service plans is no longer adequate to capture the effect of the health care context on the delivery of care or health outcomes.5 A typology of features that represents a plan’s organizational and incentive features would facilitate understanding of what specific aspects affect outcomes of care across the nation.6,7 The Managed Care Survey was developed for use in this study to build on previous work by measuring specific attributes of different managed care plans that may affect both physician and patient outcomes.

Managed care organizations traditionally position primary care clinicians as the cornerstones of their delivery system⁸; however, the effect of the restrictiveness of managed care plans on the patient-physician relationship and the delivery of important attributes of primary care (as described by the Institute of Medicine⁹) are poorly understood. A common assumption is that managed care fosters primary care because of its gatekeeper and first-contact functions.⁸ Since health systems organized around primary care have been shown to have better population-level outcomes,¹⁰ one might expect similar results from managed care systems.^11-13 Several aspects of the current competitive managed care marketplace, however, do not appear to be conducive to achieving the goals of primary care.^14,15 The restrictions on clinicians’ and patients’ choices have raised concerns about the potential detrimental effect of managed care on the patient-physician relationship.¹⁶ In addition, the practice of annual re-bidding of managed care contracts can cause a forced disruption in continuity of care¹⁷ with detrimental effects on patients.^15,18 For these reasons, the Institute of Medicine recommends monitoring the performance of health care systems to assess the adequacy of the delivery of attributes of primary care.⁹

The restrictions and incentives imposed by managed care organizations that are designed to modify physician practice behavior may inadvertently effect other valued aspects of patient care. The purpose of our study is to evaluate the association of managed care restrictiveness with specific attributes of primary care, visit-based patient satisfaction, and perceived inconvenience (or “hassle”) of using the plan.

Methods

Study Design, Sites, and Sample

A cross-sectional design was used to collect data from 15 member practices of the Ambulatory Sentinel Practice Network (ASPN). ASPN, composed of 752 community-based practicing clinicians, was established in 1982 to conduct practice-based research. Its 122 practices in 34 states have been shown to serve a nationally representative patient population and provide access to health care markets with a wide range of penetration and maturity of managed care.¹⁹ We solicited volunteer practice sites and chose 15 US ASPN sites to represent high-, medium-, and low-levels of managed care penetration in both urban and rural areas. Clinician and practice characteristics, including the clinicians’ estimate of the proportion of managed care in the practice, were obtained from the ASPN member database, which is updated annually. All 15 sites that were invited to participate in the study agreed to complete it.

Data Collection

The ASPN central office recruited and trained participating practice personnel and coordinated project implementation. Between April and August 1997, practices were sent explicit protocol instructions and copies of the Managed Care Survey and the patient survey and were instructed to choose a start date for administering surveys to 50 consecutive patients of each participating clinician. Patients were asked to complete the survey before leaving the office. A preassigned number corresponding to the patient’s insurance plan was written on the survey before it was given to the patient. The staff also kept track of the age, sex, and insurance plan of those patients who declined to participate, so that any nonparticipation bias could be evaluated.

In addition, for each practice a single Managed Care Survey was completed jointly by a physician and office manager to characterize each individual managed care health care plan with a minimum of 5% of all patients. Each insurance plan on the survey was identified by the same number that was used on the corresponding individual patients’ health care plans on the patient survey.

Measures

Managed care was conceptualized as a set of organizational restraints and financial incentives that are intended to focus and limit clinicians’ use of health care resources. The Managed Care Survey was designed to characterize managed care plans along several dimensions. The survey was developed by the ASPN Task Force on Managed Care, which consisted of 6 family physicians from the United States and 1 from Canada representing diversity in gender, geographic location, organization of practice, managed care market, and years in practice. Group consensus was used to identify and define the key managed care features that represent organizational restraints, financial incentives, and other aspects affecting the restrictiveness of plans.

The plan features measured by the managed care survey included the proportion of each practice’s patients in the plan, the plan’s financial restrictiveness and organizational restrictiveness, and the level of hassle associated with it. The financial restrictiveness portion of the survey included the type of reimbursement (capitation global risk, capitation professional risk, capitation primary care risk, discounted fee-for-service, or fee-for-service) and whether the plan carried a clinician-withhold fund or an incentive-bonus fund. The organizational restrictiveness part of the survey included plan characterization on the following features: mental health carve-out, laboratory services, formulary, preauthorization for diagnostic or treatment procedures, preauthorization for physician referrals, specialty network, and procedure (site of service) constraints. The managed care survey features and their definitions are listed in Appendix A.*

Some plan features were viewed as more important than others for describing a plan’s financial and organizational restrictiveness. Each member of the ASPN Managed Care Task Force assigned a value of importance for each feature (using a Likert scale where 1 = somewhat important; 10 = very important). The group mean assigned weight for each feature was used to calculate the 2 weighted summary scores representing the financial and organizational restrictiveness of each plan.

In addition to the managed care plan features, clinicians completing the survey were asked to rate the degree of hassle, defined as the degree of time-consuming interference with routine practice activities perceived to be associated with the plan, on a scale of 1 to 5. Additional items on the Managed Care Survey included questions about the type of practice (solo, multispecialty group, and so forth), political/business affiliations, recent mergers/buyouts, and type of clinician compensation.

The attributes of delivery of primary care were measured by the revised Components of Primary Care Instrument (CPCI)^20,21 which measures key attributes of the patient-provider relationship based on the recent Institute of Medicine definition of primary care.⁹ The CPCI assesses interpersonal communication, comprehensive care, continuity of care, coordination of care, provider’s accumulated knowledge about the patient, family orientation, community orientation, advocacy, and patient preference for their usual provider. Each attribute is measured from the patient’s perspective of the patient-provider relationship. Descriptive statistics, internal consistency reliabilities, and scale content are displayed in Appendix B.* Missing data on the CPCI scale scores were handled by setting a maximum number of missing values allowed per scale and computing a score using individual responses to the remaining scale items. Questionnaires missing more data than the maximum allowed were given no score for that scale. Therefore, the total number of patients with complete data per scale, and the sample size for analyses, varies by scale.

Patient satisfaction with the visit was measured using the Medical Outcomes Study 9-Item Visit Rating Form.²² Two scores were computed, patient satisfaction with the physician and with practice operation.²³ Another specific patient item (satisfaction with the amount of time spent with the physician) was also assessed separately. Eight items were written to assess the patient’s perceived hassle in obtaining health care, and a summary scale score was computed (internal consistency reliability = .80). Additional items on the patient survey included patient age, sex, 2 reports of health status, whether today’s physician is the patient’s regular physician, and if no, whether the patient’s regular physician is a member of the office. Standard demographic items were included on the abbreviated survey for new patients. Also included were questions regarding whether that visit was for well-care or serious illness, and whether they had been forced to change physicians in the past 2 years.

Analyses

Data from the Managed Care Survey and the patient survey were linked by a unique identifier on the basis of site and site-specific health care plan. Patient surveys that could not be linked (eg, because of a missing plan identifier or because the plan was not rated on the Managed Care Survey) were included in the descriptive statistics of the study sample, but are excluded from the analyses involving the managed care features. Descriptive statistics of the sites, clinicians, and plan features are calculated. We used chi-square tests to compare the data available from the nonresponders with the data from the responders to assess bias.

We used the Pearson correlation to test the association of the plans’ financial and organizational restrictiveness scores with the clinicians’ reports of hassle. The association of the managed care plan restrictiveness scores with each of the CPCI scale scores, patient satisfaction with the visit, and patient perceived hassle was tested with multilevel modeling techniques using hierarchical linear regression software.²⁴ Multilevel modeling is an analysis technique that accounts for the nested structural context of the data. Two potential confounding variables, patient age and health status, were included as covariates in these analyses.

Results

All 15 sites returned a completed Managed Care Survey. Practice characteristics are displayed in Table 1. One fourth of the practices had experienced a recent professional merger and one third had undergone a recent purchase or buyout. The average number of managed care plans in each practice ranged from 1 to 25, with an average of 13.3 plans. The average proportion of patients in a managed care plan per site was 51% (range = 21% to 100%). The 41 clinicians participating in the study are characterized in Table 2. The vast majority of clinicians had MD degrees, and 66% were men. On average, clinicians spent 90% of their time on patient care.

The patient response rate was also excellent. Of the 1922 patients approached, 1839 (96%) agreed to complete the patient survey. One hundred and six patients returned a blank survey and represent passive refusers. Of the 1733 patients returning a survey at least partially completed, 1503 were established patients, and 230 were new patients. Twenty-eight established patients did not see their regular physician, and that physician was not a member of the office they were visiting that day. These patients were excluded, bringing the final patient sample size to 1475.

Patient characteristics are reported in Table 3. The majority of patients were women, and health status, on average, was good. Most patients (84%) saw their regular physician, approximately half had a well-care visit within the past 2 years, and approximately one fourth were treated for a serious illness within the past 2 years. Fifty-nine percent of established patients had some type of managed care insurance. Standard Medicare and Medicaid insurance accounted for 21% of patients, and only 11% were categorized as having traditional commercial insurance. Established patients who declined to complete the patient survey (n = 41) were similar in average age and type of insurance but were more likely to be men than the patients who completed the survey.

Table 1, Table 4 displays the frequency of the different managed care features measured by the Managed Care Survey. Laboratory services, preauthorization, specialty networks, and site of service were features of more than 50% of the 199 managed care plans characterized. Physicians rated plans with a restrictive feature as generating greater hassle on average than plans without restrictive features. The 2 exceptions to this trend were plans with point-of-service and withhold features.

We investigated the association of managed care plan restrictiveness with each of the CPCI scale scores and patient satisfaction with the visit. Of the 870 patients with a type of managed care insurance, 786 patients had complete data for this analysis. For ease of interpretation, the managed care restrictiveness scores were divided into low (27%), medium (46%), and high (27%).²⁵ This categorization of the restrictiveness scores has 3 advantages: interpretation of 3 group means versus a b coefficient is easier; a nonlinear association is readily determined; and the distributions of the outcome measure for the high and low groups can be shown to be nonoverlapping. If no statistically significant difference is found between these 2 distinctly different extreme groups, this can be taken as evidence for not rejecting the null hypothesis of no association.

As indicated in Table 5, the mean of the different CPCI, hassle, and patient satisfaction scores were very similar across each level of managed care plan financial restrictiveness. Similarly, organizational restrictiveness was not significantly associated with any of the CPCI scale scores, patient report of hassle, or the satisfaction scores. These analyses were adjusted for patient age and health status and the nested effect of the data.

Discussion

We used innovative measures and a unique practice-based laboratory to assess the impact of specific aspects of managed care on the delivery of important attributes of primary care. The findings suggest that the restrictiveness of managed care plans does not affect patients’ perceptions of multiple attributes of primary care or their satisfaction with the visit. However, both financial and organizational restrictiveness were associated with greater clinician-reported hassle. These findings may not conform to the widespread belief by practicing clinicians that plan characteristics affect patients in a direct way,^8,26 and clinicians may be reassured to find that they are able to maintain good primary care relationships with patients amidst the challenges they experience.

Clinician hassle was rated in terms of the time required for insurance-mandated administrative activities generated by the plan (eg, the length and repetition of required forms and written or verbal requirements). Other studies have reported specific physician-reported hassles associated with particular plans.^8,27 Most of the hassles can be attributed to an added administrative burden, such as the need to make phone calls, write letters, and gather information from medical records in response to denial of payment, requests for patient information, or precertification of services.²⁷ Our findings that clinicians reported increased administrative burden with more restrictive plans reinforce the idea by Freberg²⁸ that it is difficult to know whether managed care plans are cost effective or merely add to the hidden cost of administrative overhead. Future studies should investigate the amount of effort required for additional administrative burdens relative to the cost savings of the plan.

The lack of association between patient-perceived hassle and plan restrictiveness indicates that the burdens of plan restrictiveness fall squarely on the shoulders of clinicians and staff. It is also likely that patients who are less concerned about plan restrictions may have self-selected a restrictive plan for cost savings or other perceived benefits. For these patients, the benefits (eg, lower deductibles, coverage of health maintenance visits) may outweigh the disadvantages (eg, restricted freedom of choice, increased personal cost incurred to opt for out-of-plan services).

Concerns have been raised about conflict of interest, the effect of financial incentives on physician behavior, the quality of the patient-physician relationship and decision making, time constraints, and the potential for underservice with managed care systems.^16,26,29-34 Grumbach and colleagues²⁶ found that 57% of physicians surveyed reported that they felt pressure from the managed care organization to limit referrals; 75% felt pressure to see more patients per day; and 17% and 24%, respectively, felt that limiting referrals and seeing more patients per day compromised patient care. In our study, patients in highly restrictive managed care plans did not perceive their physician to be any less of an advocate for their health care than patients in the low- or medium-restrictiveness groups. Thus, this sample of primary care clinicians continued to engage in trusting relationships with their patients despite the potential conflict of interest that could arise from managed care plans’ financial incentives to restrict care.

In our sample of patients, as well as in others,²¹ the CPCI assessed important aspects of primary care with good internal consistency. The instrument’s scale scores have been shown to be associated with patient satisfaction²¹ and delivery of preventive services,³⁵ and have been shown to detect differences in the delivery of primary care to patients who faced forced discontinuity of care and those who remained with their regular physician.¹⁵ The CPCI should be sensitive to many of the potential ill effects of managed care on the patient-physician relationship and delivery of primary care. The lack of association between plan restrictiveness and patient report of primary care is striking, and there is strong evidence that the clinicians and office staff who report being hassled by these restrictions are not allowing those hassles to interfere with their delivery of patient care.

Others²⁶ have evaluated physician satisfaction with specific plan features and physician-rated quality of specific health care plans.³⁶ We asked physicians to objectively report the presence or absence of specific organizational and financial features of each of the managed care plans in their practices. Using the Managed Care Survey to characterize specific organizational and financial aspects of plans is a major advance in being able to test the importance of these features on physician behavior and processes of care and patient outcomes.

Limitations

The main potential threat to the internal validity of the study is patient nonresponse. The nonrespondents were more likely to be men than patients who completed the survey. It is possible that these patients may have been less satisfied with care and may have reported lower scores on the CPCI. However, nonrespondents represent only 10% of those approached, and it is unlikely that the findings of the study would have changed if they had been included. We are also unable to comment on how long a patient had been with their current insurance plan. Length of exposure and actual experience with the features of a plan could potentially affect the association of plan restrictiveness with perceived delivery of primary care. However, consecutive patients were enrolled, which should reduce the likelihood of a selection bias of such a variable.

Replication of this study in a larger number of community-based practice sites and in a general community sample would add to the generalizability of the findings. In these times of increased business interest in medicine,³⁷ it is important to continue to monitor and evaluate the immediate, long-term, intended, and unintended outcomes of specific features of managed care.

Conclusions

Managed care plan restrictiveness does not appear to be affecting the delivery of primary care as measured from the perspective of the patient. However, the financial and organizational restrictiveness of managed care plans does lead to greater clinician hassle. We interpret these results to suggest that primary care clinicians are able to effectively buffer the effects of health plan structure on their patients. These findings raise questions about the effect of plan restrictiveness on efficient use of clinician time and the clinician’s ability to continue to deliver quality primary care amidst competing administrative demands.

Acknowledgments

We would like to thank each of the participating ASPN practices, their staffs, and patients, without whom the study would not have been possible. The practices participating in this study included the following: Batesville Family Practice Clinic, Batesville, Arkansas; Loma Linda University, Department of Family Medicine, Loma Linda, California; C. Frazer Shipman, Wheatridge, Colorado; The Family Medical Group, Bristol, Connecticut; St. John’s Mercy Family Medicine, St. Louis, Missouri; Manchester Family Health Center, Manchester, New Hampshire; Primary Care Center at Hillsborough, Belle Mead, New Jersey, and Family Medicine at Monument Square, New Brunswick, New Jersey; Central Square Health Services Center, Central Square, New York; Enid Family Medicine Clinic, Enid, Oklahoma; Good Samaritan Family Practice, ELCO, Lebanon, Pennsylvania; Michael Hartsell, MD, Greeneville, Tennessee; Annadale Family Medicine, PC, Annadale, Virginia, and Tappahannack Family Practice, Tappahannack, Virginia; and Cle Elum Family Medicine Center, Cle Elum, Washington.

Thanks to each of the ASPN Managed Care Task Force members who developed the Managed Care Survey and pilot tested the project instruments and protocol: A. John Orzano, MD; H. Andrew Selinger, MD; Robert James, MD; William Fosmire, MD; Linda French, MD; Frank Reed, MD; John Scott, MD; and Dennis de Leon, MD.

Managed care has become the predominant approach to health care financing in the United States.1 This explosive growth has been accompanied by an increasingly complex array of types of managed care plans and a growth in the use of restrictions and financial incentives to influence physician practice behavior.2 Contributing to the diversity and complexity of managed care are new incentive systems, strategies to manage patterns of care,3,4 and a shift toward national investor-owned plans. In addition, many practicing physicians are participating in new business relationships, including physician hospital organizations, medical service organizations, and risk-sharing arrangements. The treatment of all managed care plans as a single entity for comparison with fee-for-service plans is no longer adequate to capture the effect of the health care context on the delivery of care or health outcomes.5 A typology of features that represents a plan’s organizational and incentive features would facilitate understanding of what specific aspects affect outcomes of care across the nation.6,7 The Managed Care Survey was developed for use in this study to build on previous work by measuring specific attributes of different managed care plans that may affect both physician and patient outcomes.

Managed care organizations traditionally position primary care clinicians as the cornerstones of their delivery system⁸; however, the effect of the restrictiveness of managed care plans on the patient-physician relationship and the delivery of important attributes of primary care (as described by the Institute of Medicine⁹) are poorly understood. A common assumption is that managed care fosters primary care because of its gatekeeper and first-contact functions.⁸ Since health systems organized around primary care have been shown to have better population-level outcomes,¹⁰ one might expect similar results from managed care systems.^11-13 Several aspects of the current competitive managed care marketplace, however, do not appear to be conducive to achieving the goals of primary care.^14,15 The restrictions on clinicians’ and patients’ choices have raised concerns about the potential detrimental effect of managed care on the patient-physician relationship.¹⁶ In addition, the practice of annual re-bidding of managed care contracts can cause a forced disruption in continuity of care¹⁷ with detrimental effects on patients.^15,18 For these reasons, the Institute of Medicine recommends monitoring the performance of health care systems to assess the adequacy of the delivery of attributes of primary care.⁹

The restrictions and incentives imposed by managed care organizations that are designed to modify physician practice behavior may inadvertently effect other valued aspects of patient care. The purpose of our study is to evaluate the association of managed care restrictiveness with specific attributes of primary care, visit-based patient satisfaction, and perceived inconvenience (or “hassle”) of using the plan.

Methods

Study Design, Sites, and Sample

A cross-sectional design was used to collect data from 15 member practices of the Ambulatory Sentinel Practice Network (ASPN). ASPN, composed of 752 community-based practicing clinicians, was established in 1982 to conduct practice-based research. Its 122 practices in 34 states have been shown to serve a nationally representative patient population and provide access to health care markets with a wide range of penetration and maturity of managed care.¹⁹ We solicited volunteer practice sites and chose 15 US ASPN sites to represent high-, medium-, and low-levels of managed care penetration in both urban and rural areas. Clinician and practice characteristics, including the clinicians’ estimate of the proportion of managed care in the practice, were obtained from the ASPN member database, which is updated annually. All 15 sites that were invited to participate in the study agreed to complete it.

Data Collection

The ASPN central office recruited and trained participating practice personnel and coordinated project implementation. Between April and August 1997, practices were sent explicit protocol instructions and copies of the Managed Care Survey and the patient survey and were instructed to choose a start date for administering surveys to 50 consecutive patients of each participating clinician. Patients were asked to complete the survey before leaving the office. A preassigned number corresponding to the patient’s insurance plan was written on the survey before it was given to the patient. The staff also kept track of the age, sex, and insurance plan of those patients who declined to participate, so that any nonparticipation bias could be evaluated.

In addition, for each practice a single Managed Care Survey was completed jointly by a physician and office manager to characterize each individual managed care health care plan with a minimum of 5% of all patients. Each insurance plan on the survey was identified by the same number that was used on the corresponding individual patients’ health care plans on the patient survey.

Measures

Managed care was conceptualized as a set of organizational restraints and financial incentives that are intended to focus and limit clinicians’ use of health care resources. The Managed Care Survey was designed to characterize managed care plans along several dimensions. The survey was developed by the ASPN Task Force on Managed Care, which consisted of 6 family physicians from the United States and 1 from Canada representing diversity in gender, geographic location, organization of practice, managed care market, and years in practice. Group consensus was used to identify and define the key managed care features that represent organizational restraints, financial incentives, and other aspects affecting the restrictiveness of plans.

The plan features measured by the managed care survey included the proportion of each practice’s patients in the plan, the plan’s financial restrictiveness and organizational restrictiveness, and the level of hassle associated with it. The financial restrictiveness portion of the survey included the type of reimbursement (capitation global risk, capitation professional risk, capitation primary care risk, discounted fee-for-service, or fee-for-service) and whether the plan carried a clinician-withhold fund or an incentive-bonus fund. The organizational restrictiveness part of the survey included plan characterization on the following features: mental health carve-out, laboratory services, formulary, preauthorization for diagnostic or treatment procedures, preauthorization for physician referrals, specialty network, and procedure (site of service) constraints. The managed care survey features and their definitions are listed in Appendix A.*

Some plan features were viewed as more important than others for describing a plan’s financial and organizational restrictiveness. Each member of the ASPN Managed Care Task Force assigned a value of importance for each feature (using a Likert scale where 1 = somewhat important; 10 = very important). The group mean assigned weight for each feature was used to calculate the 2 weighted summary scores representing the financial and organizational restrictiveness of each plan.

In addition to the managed care plan features, clinicians completing the survey were asked to rate the degree of hassle, defined as the degree of time-consuming interference with routine practice activities perceived to be associated with the plan, on a scale of 1 to 5. Additional items on the Managed Care Survey included questions about the type of practice (solo, multispecialty group, and so forth), political/business affiliations, recent mergers/buyouts, and type of clinician compensation.

The attributes of delivery of primary care were measured by the revised Components of Primary Care Instrument (CPCI)^20,21 which measures key attributes of the patient-provider relationship based on the recent Institute of Medicine definition of primary care.⁹ The CPCI assesses interpersonal communication, comprehensive care, continuity of care, coordination of care, provider’s accumulated knowledge about the patient, family orientation, community orientation, advocacy, and patient preference for their usual provider. Each attribute is measured from the patient’s perspective of the patient-provider relationship. Descriptive statistics, internal consistency reliabilities, and scale content are displayed in Appendix B.* Missing data on the CPCI scale scores were handled by setting a maximum number of missing values allowed per scale and computing a score using individual responses to the remaining scale items. Questionnaires missing more data than the maximum allowed were given no score for that scale. Therefore, the total number of patients with complete data per scale, and the sample size for analyses, varies by scale.

Patient satisfaction with the visit was measured using the Medical Outcomes Study 9-Item Visit Rating Form.²² Two scores were computed, patient satisfaction with the physician and with practice operation.²³ Another specific patient item (satisfaction with the amount of time spent with the physician) was also assessed separately. Eight items were written to assess the patient’s perceived hassle in obtaining health care, and a summary scale score was computed (internal consistency reliability = .80). Additional items on the patient survey included patient age, sex, 2 reports of health status, whether today’s physician is the patient’s regular physician, and if no, whether the patient’s regular physician is a member of the office. Standard demographic items were included on the abbreviated survey for new patients. Also included were questions regarding whether that visit was for well-care or serious illness, and whether they had been forced to change physicians in the past 2 years.

Analyses

Data from the Managed Care Survey and the patient survey were linked by a unique identifier on the basis of site and site-specific health care plan. Patient surveys that could not be linked (eg, because of a missing plan identifier or because the plan was not rated on the Managed Care Survey) were included in the descriptive statistics of the study sample, but are excluded from the analyses involving the managed care features. Descriptive statistics of the sites, clinicians, and plan features are calculated. We used chi-square tests to compare the data available from the nonresponders with the data from the responders to assess bias.

We used the Pearson correlation to test the association of the plans’ financial and organizational restrictiveness scores with the clinicians’ reports of hassle. The association of the managed care plan restrictiveness scores with each of the CPCI scale scores, patient satisfaction with the visit, and patient perceived hassle was tested with multilevel modeling techniques using hierarchical linear regression software.²⁴ Multilevel modeling is an analysis technique that accounts for the nested structural context of the data. Two potential confounding variables, patient age and health status, were included as covariates in these analyses.

Results

All 15 sites returned a completed Managed Care Survey. Practice characteristics are displayed in Table 1. One fourth of the practices had experienced a recent professional merger and one third had undergone a recent purchase or buyout. The average number of managed care plans in each practice ranged from 1 to 25, with an average of 13.3 plans. The average proportion of patients in a managed care plan per site was 51% (range = 21% to 100%). The 41 clinicians participating in the study are characterized in Table 2. The vast majority of clinicians had MD degrees, and 66% were men. On average, clinicians spent 90% of their time on patient care.

The patient response rate was also excellent. Of the 1922 patients approached, 1839 (96%) agreed to complete the patient survey. One hundred and six patients returned a blank survey and represent passive refusers. Of the 1733 patients returning a survey at least partially completed, 1503 were established patients, and 230 were new patients. Twenty-eight established patients did not see their regular physician, and that physician was not a member of the office they were visiting that day. These patients were excluded, bringing the final patient sample size to 1475.

Patient characteristics are reported in Table 3. The majority of patients were women, and health status, on average, was good. Most patients (84%) saw their regular physician, approximately half had a well-care visit within the past 2 years, and approximately one fourth were treated for a serious illness within the past 2 years. Fifty-nine percent of established patients had some type of managed care insurance. Standard Medicare and Medicaid insurance accounted for 21% of patients, and only 11% were categorized as having traditional commercial insurance. Established patients who declined to complete the patient survey (n = 41) were similar in average age and type of insurance but were more likely to be men than the patients who completed the survey.

Table 1, Table 4 displays the frequency of the different managed care features measured by the Managed Care Survey. Laboratory services, preauthorization, specialty networks, and site of service were features of more than 50% of the 199 managed care plans characterized. Physicians rated plans with a restrictive feature as generating greater hassle on average than plans without restrictive features. The 2 exceptions to this trend were plans with point-of-service and withhold features.

We investigated the association of managed care plan restrictiveness with each of the CPCI scale scores and patient satisfaction with the visit. Of the 870 patients with a type of managed care insurance, 786 patients had complete data for this analysis. For ease of interpretation, the managed care restrictiveness scores were divided into low (27%), medium (46%), and high (27%).²⁵ This categorization of the restrictiveness scores has 3 advantages: interpretation of 3 group means versus a b coefficient is easier; a nonlinear association is readily determined; and the distributions of the outcome measure for the high and low groups can be shown to be nonoverlapping. If no statistically significant difference is found between these 2 distinctly different extreme groups, this can be taken as evidence for not rejecting the null hypothesis of no association.

As indicated in Table 5, the mean of the different CPCI, hassle, and patient satisfaction scores were very similar across each level of managed care plan financial restrictiveness. Similarly, organizational restrictiveness was not significantly associated with any of the CPCI scale scores, patient report of hassle, or the satisfaction scores. These analyses were adjusted for patient age and health status and the nested effect of the data.

Discussion

We used innovative measures and a unique practice-based laboratory to assess the impact of specific aspects of managed care on the delivery of important attributes of primary care. The findings suggest that the restrictiveness of managed care plans does not affect patients’ perceptions of multiple attributes of primary care or their satisfaction with the visit. However, both financial and organizational restrictiveness were associated with greater clinician-reported hassle. These findings may not conform to the widespread belief by practicing clinicians that plan characteristics affect patients in a direct way,^8,26 and clinicians may be reassured to find that they are able to maintain good primary care relationships with patients amidst the challenges they experience.

Clinician hassle was rated in terms of the time required for insurance-mandated administrative activities generated by the plan (eg, the length and repetition of required forms and written or verbal requirements). Other studies have reported specific physician-reported hassles associated with particular plans.^8,27 Most of the hassles can be attributed to an added administrative burden, such as the need to make phone calls, write letters, and gather information from medical records in response to denial of payment, requests for patient information, or precertification of services.²⁷ Our findings that clinicians reported increased administrative burden with more restrictive plans reinforce the idea by Freberg²⁸ that it is difficult to know whether managed care plans are cost effective or merely add to the hidden cost of administrative overhead. Future studies should investigate the amount of effort required for additional administrative burdens relative to the cost savings of the plan.

The lack of association between patient-perceived hassle and plan restrictiveness indicates that the burdens of plan restrictiveness fall squarely on the shoulders of clinicians and staff. It is also likely that patients who are less concerned about plan restrictions may have self-selected a restrictive plan for cost savings or other perceived benefits. For these patients, the benefits (eg, lower deductibles, coverage of health maintenance visits) may outweigh the disadvantages (eg, restricted freedom of choice, increased personal cost incurred to opt for out-of-plan services).

Concerns have been raised about conflict of interest, the effect of financial incentives on physician behavior, the quality of the patient-physician relationship and decision making, time constraints, and the potential for underservice with managed care systems.^16,26,29-34 Grumbach and colleagues²⁶ found that 57% of physicians surveyed reported that they felt pressure from the managed care organization to limit referrals; 75% felt pressure to see more patients per day; and 17% and 24%, respectively, felt that limiting referrals and seeing more patients per day compromised patient care. In our study, patients in highly restrictive managed care plans did not perceive their physician to be any less of an advocate for their health care than patients in the low- or medium-restrictiveness groups. Thus, this sample of primary care clinicians continued to engage in trusting relationships with their patients despite the potential conflict of interest that could arise from managed care plans’ financial incentives to restrict care.

In our sample of patients, as well as in others,²¹ the CPCI assessed important aspects of primary care with good internal consistency. The instrument’s scale scores have been shown to be associated with patient satisfaction²¹ and delivery of preventive services,³⁵ and have been shown to detect differences in the delivery of primary care to patients who faced forced discontinuity of care and those who remained with their regular physician.¹⁵ The CPCI should be sensitive to many of the potential ill effects of managed care on the patient-physician relationship and delivery of primary care. The lack of association between plan restrictiveness and patient report of primary care is striking, and there is strong evidence that the clinicians and office staff who report being hassled by these restrictions are not allowing those hassles to interfere with their delivery of patient care.

Others²⁶ have evaluated physician satisfaction with specific plan features and physician-rated quality of specific health care plans.³⁶ We asked physicians to objectively report the presence or absence of specific organizational and financial features of each of the managed care plans in their practices. Using the Managed Care Survey to characterize specific organizational and financial aspects of plans is a major advance in being able to test the importance of these features on physician behavior and processes of care and patient outcomes.

Limitations

The main potential threat to the internal validity of the study is patient nonresponse. The nonrespondents were more likely to be men than patients who completed the survey. It is possible that these patients may have been less satisfied with care and may have reported lower scores on the CPCI. However, nonrespondents represent only 10% of those approached, and it is unlikely that the findings of the study would have changed if they had been included. We are also unable to comment on how long a patient had been with their current insurance plan. Length of exposure and actual experience with the features of a plan could potentially affect the association of plan restrictiveness with perceived delivery of primary care. However, consecutive patients were enrolled, which should reduce the likelihood of a selection bias of such a variable.

Replication of this study in a larger number of community-based practice sites and in a general community sample would add to the generalizability of the findings. In these times of increased business interest in medicine,³⁷ it is important to continue to monitor and evaluate the immediate, long-term, intended, and unintended outcomes of specific features of managed care.

Conclusions

Managed care plan restrictiveness does not appear to be affecting the delivery of primary care as measured from the perspective of the patient. However, the financial and organizational restrictiveness of managed care plans does lead to greater clinician hassle. We interpret these results to suggest that primary care clinicians are able to effectively buffer the effects of health plan structure on their patients. These findings raise questions about the effect of plan restrictiveness on efficient use of clinician time and the clinician’s ability to continue to deliver quality primary care amidst competing administrative demands.

Acknowledgments

We would like to thank each of the participating ASPN practices, their staffs, and patients, without whom the study would not have been possible. The practices participating in this study included the following: Batesville Family Practice Clinic, Batesville, Arkansas; Loma Linda University, Department of Family Medicine, Loma Linda, California; C. Frazer Shipman, Wheatridge, Colorado; The Family Medical Group, Bristol, Connecticut; St. John’s Mercy Family Medicine, St. Louis, Missouri; Manchester Family Health Center, Manchester, New Hampshire; Primary Care Center at Hillsborough, Belle Mead, New Jersey, and Family Medicine at Monument Square, New Brunswick, New Jersey; Central Square Health Services Center, Central Square, New York; Enid Family Medicine Clinic, Enid, Oklahoma; Good Samaritan Family Practice, ELCO, Lebanon, Pennsylvania; Michael Hartsell, MD, Greeneville, Tennessee; Annadale Family Medicine, PC, Annadale, Virginia, and Tappahannack Family Practice, Tappahannack, Virginia; and Cle Elum Family Medicine Center, Cle Elum, Washington.

Thanks to each of the ASPN Managed Care Task Force members who developed the Managed Care Survey and pilot tested the project instruments and protocol: A. John Orzano, MD; H. Andrew Selinger, MD; Robert James, MD; William Fosmire, MD; Linda French, MD; Frank Reed, MD; John Scott, MD; and Dennis de Leon, MD.