Matching Patients and Practitioners Based on Beliefs About Care: Results of a Randomized Controlled Trial

November 1, 2004
Edward Krupat, PhD
Edward Krupat, PhD

,
John Hsu, MD, MBA, MSCE
John Hsu, MD, MBA, MSCE

,
Julie Irish, PhD
Julie Irish, PhD

,
Julie A. Schmittdiel, MA
Julie A. Schmittdiel, MA

,
Joe Selby, MD, MPH
Joe Selby, MD, MPH

Volume 10, Issue 11 Pt 1

Objective: To determine whether an intervention designed toinform and guide patients in choosing a primary care provider(PCP) could increase satisfaction and trust.

Design: A randomized controlled trial conducted at the SantaClara Medical Center of the Kaiser Permanente Medical CareProgram in Northern California.

Patients and Methods: Patients needing a PCP were randomlyassigned to Informed Choice (provider-level information offered),Guided Choice (PCP names provided based on the similarity ofpatients' and practitioners' patient-centered beliefs, and patientsthen chose), or Usual Care. One year later, mailed questionnairesasked about patients' trust and satisfaction, perceptions of thechoice process, and trust and satisfaction with Kaiser Permanente.More than 5000 adult patients were linked with a PCP as part ofthe project, and completed surveys were returned by 2437patients.

Results: Neither intervention arm generated better matches onpatient-practitioner beliefs than Usual Care; however, trust and satisfactionwere higher among Guided Choice patients. Across studyarms, patient-practitioner belief discrepancy showed a consistentpositive association with trust, satisfaction, and attitudes towardKaiser Permanente.

Conclusions: The discrepancy between patient and practitionerbeliefs about care is an important determinant of trust and satisfaction,and involving patients in the selection of their PCP can havean independent positive effect. It should be possible to build on theexperience of this project to develop programs that better matchpatients and practitioners within managed care plans.

(Am J Manag Care. 2004;10:814-822)

People in almost 50 million US households selectedor changed to a new physician in a recent 2-year period.1 In finding a new physician, patientsprefer to select rather than be assigned, and those whochoose are more satisfied, trusting, and likely to adhereto recommendations.2-6 Most people rely on the recommendationsof friends, family, or coworkers. However,those who desire information from their health planstypically indicate that they "pick blindly" from listsprovided to them, and many patients attribute the findingof a clinician with whom they were comfortable togood luck.1

Patients vary in the kinds of practitioner informationto which they would like access,7-9 with 4 major categoriesemerging2,10: descriptors and characteristics (eg,sex, medical school, and years in practice), convenienceand logistics (eg, location, hours, and wait forappointments), expertise and qualifications (eg, training,number of procedures performed, and malpracticesuits), and relationship and bedside manner (eg, compassion,communication skills, and practice style).Although expertise and qualifications were the characteristicsthat patients mentioned most often as important,finding a physician who shared a similar ideologyand had good communication skills was deemed bymany as at least as important.11

A review of 40 commercial (eg, WebMD and bestdoctors.com) and organization-specific (eg, CIGNA, Pacificare,and Boston Medical Center) Web sites that containedprovider-specific information indicated that the availableinformation varied considerably.10 For instance,physician location was contained in 92.5% of Web sites,specialty in 87.5%, medical school in 62.5%, and boardcertification in 55.0%. Information on physicians' philosophyof providing care was available in only 3 (7.5%)of the sites reviewed.

Although few sources provide information on practitionerbeliefs about care, several studies12-14 havedemonstrated that the extent to which patients andpractitioners hold similar or comparable beliefs is associatedwith satisfaction and trust. In these studies,patients and their physicians completed identical versionsof the Patient-Practitioner Orientation Scale(PPOS), a validated instrument that measures thedegree to which each holds patient-centered beliefsabout the sharing of power and information. Building onthese findings, this article presents the results of aprospective randomized trial in which patients wereguided in their choice of a practitioner based on theirbeliefs about care.

It was hypothesized that providing patients with abasis for choosing a primary care provider (PCP) basedon beliefs about care in the Guided Choice arm wouldresult in physician-patient pairs who were in greateragreement than those who had not received such information.Moreover, it was expected that the greaterbelief similarity found in Guided Choice would result ingreater patient trust and satisfaction.

METHODS

The funding agencies for this project played no rolein any aspect of its design or in the collection, analysis,or interpretation of the data, and they made no requirementsof approval of the finished manuscript. TheKaiser Foundation Research Institute InstitutionalReview Board approved the study.

The site for the study was the Department ofMedicine at the Santa Clara Medical Center of theKaiser Permanente (KP) Medical Care Program inNorthern California, a large group-model health maintenanceorganization. At the time the study began,there were approximately 162 000 members and 80adult medicine PCPs on staff at the Santa ClaraMedical Center. Of these PCPs, 52 completed thePPOS and were linked with at least 1 patient throughthe intervention.

Any adult member 30 or older who was not linked toa PCP in the automated health plan database was eligiblefor the study, although only one family membercould participate. Patients were selected from recentlyenrolled members who did not have a PCP, ongoingmembers who did not have a PCP listed, and patients of2 PCPs with large panels who had retired shortly beforethe initiation of the study. Within each of these sources,patients were randomly assigned to 1 of 2 intervention(choice) arms or to the Usual Care arm. Further detailsof the selection and randomization processes have beenpreviously published.15

Intervention Arms

All patients in the 2 intervention arms were contactedby mail at the outset of the study and were encouragedto contact the center to obtain a new PCP viatelephone or a Web site set up especially for this project.The Web site had a set of simple instructions forselecting a PCP, and patients who used the telephonereached a research assistant who led them through thesame procedure.

Informed Choice.

Patients in this arm were offeredprovider-specific information on all available PCPs,which included sex, age, practitioner type (physician ornurse practitioner), medical office location, race or ethnicity,languages spoken, medical school attended, areasof clinical interest, areas of personal interest or hobbies,and openness to alternative medicine. A recent photographwas also available for those who used the Internet.

Guided Choice.

Patients in this arm had available tothem all the information of Informed Choice patients,but their selection process involved an additional element:they were given a list of PCPs whose beliefs aboutthe sharing of power and information were closest totheirs. To accomplish this, before the project began, allavailable PCPs were asked to complete a 9-item sub-scaleof the PPOS. All Guided Choice patients responded toidentical items as part of the selection process, and theiranswers were computer scored immediately (scorerange, 9-54, with a higher score indicating greaterpatient-centeredness). Patients were then offered a listof 2 to 5 PCPs whose scores were most similar (theywere told that these practitioners had a "communicationstyle" that most closely matched theirs).

The algorithm for recommending practitioners providedthe names of at least the 2 best matches, definedas the lowest absolute difference between the patient'sand the practitioner's scores. Up to 5 practitioners'names were offered if the absolute difference betweenthe patient's and the practitioner's PPOS scores waswithin 5 points. Guided Choice patients were notrequired or pressured to select one of the indicated practitionersand were encouraged to explore additionalinformation (the same information available to InformedChoice patients) about any PCP with an open panel.

Usual Care Arm

Patients in the Usual Care arm, which served as abaseline or control group, consisted of new and ongoingmembers who were not identified as part of the panel ofa specific PCP. They were contacted by mail and wereencouraged to call the medical center's patient servicesdepartment to obtain a PCP. Patient services staff didnot have access to any systematic information aboutpractitioners and were instructed not to offer unsolicitedinformation, although they were courteous inanswering practitioner-relevant questions. In theabsence of a stated preference, patients were linked toan available PCP by whatever nonsystematic mechanismswere typically used at that center.

Usual Care patients whose PCP had just retiredreceived a letter from the health plan informing them oftheir PCP's retirement and assigning them to a newPCP. Although this method of assignment was morerestrictive than that used for the other Usual Carepatients, the 2 methods of pairing patients and practitioners(which were combined for purposes of analysis)were low choice and low information and were standardprocedure (ie, usual care) at that center for connectingpatients with PCPs. All patients were surveyed by mailto assess outcomes approximately 1 year after theirlinkage with a PCP.

Patients Studied

There were 10 944 patients deemed eligible for thestudy. Of these, 5059 (46.2%) were linked with a PCPduring the study, 44.0% in the combined choice armsand 51.0% in the control arm. For the survey, we sampled3884 patients, which excluded half of the UsualCare arm patients of retiring physicians who had beenassigned a new PCP by mail (done to save survey costs)and more than 500 patients who had died, left thehealth plan, or for whom accurate contact informationwas unavailable. After 3 waves of mailing and 1 telephonereminder, we received completed surveys from2437 patients, a response rate of 62.7%. This wasreduced by 269 respondents who had selected a PCPwho practiced at a different site, 144 patients who indicatedthat the listed practitioner had never been theirPCP, and 15 people who had left KP. All baseline analysesof patients are based on these 2009 patients whohad been randomized to 1 of the 3 arms. However,because we did not have PPOS scores for both membersof some patient-practitioner pairs, analyses that investigatedpatient-practitioner belief similarity are based on1766 patients (Figure 1).

Data Collected

The following information was obtained via questionnaire:

Personal and Demographic Characteristics.

Thisincluded information such as sex, age, education,ethnicity, marital status, and current state ofhealth.

Patient Orientations.

Because patients in the controland Informed Choice arms had not completed thePPOS as part of the initial selection process,patients in all 3 arms were asked to respond tothe 9-item PPOS.

Practitioner Selection Process.

Five questions(based on a 6-point Likert scale) covered areassuch as whether patients had sufficient information,had enough practitioners to choose from, orbelieved that they were rushed in selecting apractitioner.

Satisfaction.

Eleven questions, adapted from theMedical Outcomes Study,16 measured satisfactionwith the practitioner on a 5-point scale(score range, excellent to poor). Two itemsreferred to technical skills (eg, use of latest medicaltechnologies), 1 was an overall satisfactionquestion, and the remaining 8 referred to interpersonalissues (eg, quality of explanations givenand how carefully the practitioner listens).

Trust.

Eight questions adapted from the trustscale by Thom et al5,17 were included. Theseincluded the degree to which the practitioner istrusted to manage medical problems, to tell thepatient if a mistake was made, and to put thepatient's medical needs above all other considerations,including costs.

Attitudes Toward Kaiser Permanente.

Two questions(based on a 6-point Likert scale) asked aboutthe patient's trust in KP (to put the patient's medicalneeds above other considerations, includingcosts, and to connect the patient with the bestpractitioner for his or her healthcare), and a singlequestion asked about overall satisfaction withKP (based on a 5-point scale; score range, excellentto poor).

Statistical Analysis

The analyses were conducted in several phases.First, 1-way analysis of variance (ANOVA) was used todetermine whether patients differed demographicallyaccording to study arm and whether these characteristicswere associated with PPOS scores. Second, toassess whether the 3 study arms generated differencesin patient-practitioner belief similarity, we created anew variable, namely, discrepancy score, which was thesigned difference between the patient's PPOS score andthat of his or her PCP. Analysis of variance was used tocompare mean discrepancy scores across study armsand to determine whether the intervention had an effecton the outcome measures. Paired comparisons testedthe differences among each of the study arms, andregression analyses allowed for a determination of themagnitude of effects after controlling for relevantcovariates.

To test whether patients and practitioners who heldsimilar beliefs had more positive outcomes (acrossstudy arms) in a manner parallel to that using study armas the predictor variable, 3 discrepancy score groupswere created. Cutoff points were based on the algorithmused in pairing patients and practitioners in GuidedChoice: small difference between patients and practitioners(absolute difference between patient's and practitioner'sscore is within 5 points, n = 579), patients morepatient-centered (patient's score is > 5 points higher thanpractitioner's, n = 528), and patients less patient-centered(patient's score is > 5 points lower than practitioner's, n= 659). Using discrepancy score as the independentvariable, an identical set of analyses was performed todetermine the relationship between belief similarity andthe outcome measures. To identify the independentand joint effects of physicians' and patients' PPOSscores on the outcomes, additional regression analysesand 2-way ANOVAs were performed. To account forpotential clustering effects because many patientsshared the same PCP, we used the cluster function inStata 6.0 (StataCorp LP, College Station, Tex) in allanalyses comparing experimental arms and discrepancyscore groups.

RESULTS

Characteristics of the Sample

P

As indicated in Table 1, the mean age of the samplewas 57.0 years, with somewhat more women (55.1%)than men (44.9%). The patients were well educated(40.2% with a college degree or higher), and most werewhite (71.6%) and married or in a marriage-like relationship(63.6%). Only 44.9% reported that their healthwas very good or excellent, and 86.5% had had 1 ormore office visits with their PCP during the study period,with a mean of 3.0 visits per patient. The patients'mean &#177; SD PPOS score, with higher numbers indicatinggreater patient-centeredness, was 38.97 &#177; 8.91 (outof a possible 54) and was similar to previous administrationsof the scale. Patients did not vary demographicallyaccording to intervention arm, except thatpatients in the Usual Care arm were significantlyyounger and more often male. In addition, tests comparingall patient respondents (n = 2009) with thosewhose data were included in the analyses of patient-practitionerbelief similarity (n = 1766) indicated nostatistically significant differences on any of these variables.Female patients and those who were younger,better educated, and healthier held more patient-centeredbeliefs ( < .001 for all).

The physicians' mean age was 40.3 years; 53.8%were male, with an identical percentage who werewhite. They had been practicing for a mean of 12.7years. Their mean &#177; SD PPOS score (41.67 &#177; 6.75) wasslightly more patient-centered than that of the patients,which is consistent with previous administrations of thescale. None of the demographic characteristics were statisticallyrelated to practitioner PPOS scores.

Effect of the Intervention

The mean discrepancy score in the Guided Choicecondition (&#8722;1.35) was not significantly lower than thediscrepancy scores in the Informed Choice and UsualCare arms (&#8722;2.25 and &#8722;1.62, respectively). Comparisonsusing the absolute mean discrepancies betweenpatients' and physicians' scores also showed no significantdifferences across study arms. The findings for satisfaction,trust, perception of the choice process, andattitudes toward KP are summarized in Table 2.

Satisfaction.

P

Of the 11 satisfaction items, 7 were significantlydifferent ( < .04) and 2 marginally significantaccording to study arm. Among the 7satisfaction items showing significant differences,Guided Choice patients were significantly moresatisfied on all 7 items compared with Usual Carepatients and on 3 items compared with InformedChoice patients. After controlling for age, sex, currentstate of health, education, race or ethnicity,and marital status, only one satisfaction itemremained significant, with 3 others at marginalsignificance.

Trust.

P

Six of the 8 trust items were significantly different( < .05) according to study arm, withanother at marginal significance. Guided Choicepatients made significantly more positive ratingson 5 of the 6 trust items compared with InformedChoice and Usual Care patients. However,only 2 of the individual trust items remained atmarginal significance once the other variableswere controlled for.

Perception of the Process.

P

Patients' ratings of thechoice process showed a similar pattern, with 4 ofthe 5 items showing significant differences accordingto study arm ( < .04). Guided Choice patientsrated the process significantly higher on all 4 itemscompared with Informed Choice patients and on 2items compared with Usual Care patients. Aftercontrolling for other covariates, 3 of the 4 itemsthat were originally significant remained so, and 1was marginally significant.

Attitudes Toward Kaiser Permanente.

Patients' ratingsof their attitudes toward KP showed no significantdifferences on any of the items measuringtrust and satisfaction with the organization.

Effects of Belief Discrepancy

To test the extent to which the discrepancy betweenpatient and practitioner beliefs made a difference independentof the intervention, patient-PCP pairs werecompared across study arms. In these analyses, the 3discrepancy groups consisted of pairs in which the relationshipof the physician's to the patient's score wasclosely matched, the patient was more patient-centered,or the patient was less patient-centered. The findingsfor each of the outcome variables are summarizedin Table 3.

Satisfaction.

P

Nine of the 11 satisfaction items weresignificantly different ( < .03). Of these, the groupin which practitioners were more patient-centeredthan their patients was more satisfied than thosepairs whose beliefs were similar on all 9 items, andwas more satisfied on 8 of 9 items compared withpairs in which the practitioner was less patient-centered.The groups in which patients had similarbeliefs or in which the practitioner was lesspatient-centered were not significantly differenton any of these items. Once the other covariatesused in the prior analyses were accounted for, 5items remained significant and 4 were marginal.

Trust.

P

All 8 trust items were highly significant ( < .001for 7 of 8) according to discrepancy score group.Comparing these pairwise, the group in which thepractitioner was more patient-centered generatedsignificantly higher ratings than either of the other2 discrepancy score groups on every trust item,and the latter 2 groups did not differ significantlyon any of the items. After controlling for othercovariates, 7 of these items remained significantand 1 was marginal.

Perception of the Process.

Only 2 of the 5 items weresignificantly different by discrepancy score group,and for these the pairwise comparisons and regressionanalyses indicated a pattern identical to thoseof satisfaction and trust. The group in which thepractitioner was more patient-centered had higherratings than either of the others, with the effectson both items remaining significant after thecovariates were accounted for.

Attitudes Toward Kaiser Permanente.

P

Discrepancyscore group was significantly related to each of the3 questions in this domain ( < .001 for all). Consistentwith the other outcomes, the practitionerswho were more patient-centered had higher ratingsthan the other 2 groups. The differences on all3 variables remained significant after controllingfor the covariates.

Tests of Independent and Joint Effects

To clarify the extent to which patient PPOS score,provider PPOS score, and discrepancy score independentlycontributed to these findings, additional analyseswere performed. Using regression analysis, signeddiscrepancy scores were entered along with thecovariates of age, sex, education, ethnicity, currentstate of health, and marital status. In a separate set ofanalyses, physician and patient PPOS scores wereentered as independent predictors, along with thesame covariates.

The results of each regression analysis showed aconsistent pattern across outcomes. Discrepancy scorewas a strong independent predictor of satisfaction (significanton 5 items and marginally significant on 4 others),trust (significant on 7 items and marginallysignificant on the remaining other), attitudes towardKP (significant on all 3 items), and evaluation of thechoice process (significant on 2 of 5items). In the second set of regressionanalyses, physician PPOS was not a significantpredictor of any of the outcomemeasures. Patient PPOS, however, was astrong predictor, with patient-centeredpatients giving lower scores on 8 of the 11physician satisfaction items, all 8 trustitems, all 3 items measuring attitudestoward KP, and on 4 of the 5 items ratingthe choice process.

P

P

P

Using ANOVA to determine the extentto which a joint effect existed in the formof an interaction between physician andpatient PPOS score, we divided the physicianand patient groups into thirdsaccording to their PPOS scores (to simplifythe analysis, we included in the analysisonly physicians in the upper and lowerthirds). The results of these 2 × 3 ANOVAs were noteworthyin the area of trust, with consistent main effectsindicating that patients with high PPOS scores were lesstrusting ( < .004 for all 8 items), but with no significantmain effects for physician PPOS scores. For 6 of the 8trust items, there was a consistent strong interaction(range, = .04 to = .08) of physician and patient PPOSscores.

P

To summarize the overall relationship between trustand physician and patient PPOS scores, the 8 trust itemswere summed into a single index and an additional 2 × 3ANOVA was performed. The interaction between patientand physician scores ( < .04) is illustrated in Figure 2.It shows that patient-centered patients were less trustingregardless of their practitioners' orientation, yet theywere more trusting of those physicians whose beliefsabout care were more similar to their own.

COMMENT

To our knowledge, the project reported is the first ofits kind to involve patients in selecting their practitionersbased on beliefs about care and to systematicallytest its effect. By allowing patients this opportunity, itwas expected that better matches would be created andthat satisfaction and trust would be high as a result.

Surprisingly, the discrepancy between patients' andpractitioners' belief scores was not smaller in GuidedChoice than either of the other 2 arms. This findingmay be explained several ways. First, because the PPOSwas not completed by all patients at the time theyselected a PCP, the patients' PPOS scores used in thisanalysis were collected a year later. Although beliefstoward care are thought to be fixed and the PPOS hasbeen shown to have acceptable levels of test-retest reliability,some shifting of scores may have introduced adegree of random error.

Second, given the small number of practitionersinvolved, the distribution of scores may not have beenlarge enough to provide all patients with an ideal arrayof possible practitioners from which to choose. Even ina large care center such as the one used, there may havebeen a limited number of practitioners with open panelswhose belief scores were within 5 points of anygiven patient.

Third, some Guided Choice patients may have chosennot to select one of the practitioners whose nameswere offered. No formal records were kept as to whichpatients or what percentage of patients selected one ofthe recommended practitioners, although study staffrecall that this was not common. However, the actualextent of this cannot be determined. Finally, patientsmay have their own informal ways of finding physicianswith compatible beliefs that were not measured ordetected in this study.

Because there were no meaningful differences amongthe 3 arms in their relative discrepancy scores, whichwas the mechanism assumed to affect satisfaction andtrust, it was surprising that the intervention nonethelesshad a small positive effect. We suggest that the salutaryeffect of the intervention may have resulted from a differentmechanism, namely, that soliciting patients' preferencesmay be an independent factor that leadspatients to have positive associations. This suggests thatcombining the act of choice with a comprehensive systemfor assisting or guiding choice should be effective.

Although the Guided Choice intervention did notcreate closer belief matches, we were able to explore theprinciple that served as the basis for the interventionby looking at the effect of belief discrepancy acrossarms. We found that discrepancy score had a small butconsistent relationship with trust and satisfaction withone's practitioner and was associated with perceptionsof trust and satisfaction with the larger organization.

An unanticipated finding was that the group inwhich the patients' scores were less patient-centeredthan their practitioners' had more positive associationsthan that in which belief discrepancy was small. Toexplain this, we note that the practitioners' mean PPOSscores were higher and the standard deviation slightlylower than those of the patients. This suggests that thebest pairings of patients and practitioners may be thosewith equivalent or compatible scores rather than scoresthat are identical or matching.

The conclusions of this research are limited by thefact that few practitioners (n = 65) were involved, thatthe setting was a single group-model healthcare center,and that a large percentage of the patients were whiteand middle class. In addition, like most studies dealingwith the patient-practitioner relationship, this studywas cross-sectional, capturing satisfaction and trust atone point in time rather than following the manner inwhich the beliefs of both parties and levels of satisfactionand trust evolve over time. Nonetheless, given thefindings that patients who were involved in the choiceprocess were more positive on several of the outcomemeasures, and the association of patient-practitionerbelief discrepancy with positive outcomes (nowdemonstrated in several studies),12-14 the study reinforcesthe idea that interventions such as the oneattempted can have significant value.

Subsequent interventions to match patients andpractitioners can benefit from the lessons of this one tocreate a better system of matching. If the PPOS wereused as a tool to create such a system, it would beimportant to determine an appropriate algorithm thatserves to define what levels of relative patient-practitionerscores constitute a best match. More generally, itwould be necessary to determine which beliefs aremost relevant and important to patients and to find themost appropriate means of providing this informationin an accessible format. By determining an individualpatient's rank ordering of choice criteria, it might bepossible to develop a system in which choice is guidedvia a series of steps (eg, providing the names of onlythose male or female practitioners with compatiblebeliefs if it were determined that practitioner sex was acritical issue for a given patient). Given the number ofinstances in which patients select or switch to a newpractitioner, the desire for patients to choose in aninformed manner, the growing technological capabilityof providing relevant information, and the suggestiveresults of this project, it would seem that healthcareorganizations would do well to devote growing attentionto address issues such as these to determine the bestways of matching patients with practitioners.

From the Office of Educational Development, Harvard Medical School (EK), Divisionof Research, Kaiser Permanente (JH, JS), and The Health Institute, New England MedicalCenter (JI), Boston, Mass; and School of Public Health, University of California at Berkeley(JAS).

This research was supported by a grant from The Robert Wood Johnson Foundation,Princeton, NJ. Additional support for the analysis of the data was provided by a grant fromThe Commonwealth Fund, New York, NY.

Address correspondence to: Edward Krupat, PhD, Office of Educational Development,Harvard Medical School, 383 MEC, 260 Longwood Avenue, Boston, MA 02115. E-mail:ed_krupat@hms.harvard.edu.

How We Choose Doctors:What Is and What Could Be.

1. Consumer Information and Education Committee. Chicago, Ill: Midwest Business Group on Health;2000.

California Consumers Talk About HealthCare Quality: Findings From Focus Group Discussions.

2. Peter D, Hart Research Associates. Oakland: CaliforniaHealthcare Foundation; 1999.

JAMA.

3. Schmittdiel JA, Selby JV, Grumbach K, Quesenberry CP. Choice of personalphysician and patient satisfaction in a health maintenance organization. 1997;278:1596-1599.

JAMA.

4. Grumbach K, Selby JV, Damberg C, Bindman AB, Quesenberry C Jr, UratsuC. Resolving the gatekeeper conundrum: what patients value in primary care andreferrals to specialists. 1999;282:261-266.

Med Care.

5. Thom DH, Ribisl KM, Stewart AL, Luke DA. Validation of a measure ofpatients' trust in their physician: the Trust in Physician Scale. 1999;37:510-517.

AmJ Manag Care.

6. Krupat E, Stein T, Selby JV, Yeager CM, Schmittdiel J. Choice of a primarycare physician and its relationship to adherence among patients with diabetes. 2002;8:777-784.

Health Aff (Millwood).

7. Tumlinson A, Bottigheimer H, Mahoney P, Stone EM, Hendricks A. Choosinga health plan: what information will consumers use? 1997;16(3):229-238.

Health Aff (Millwood).

8. Edgman-Levitan S, Cleary PD. What information do consumers want and need? 1996;15(4):42-56.

Med Care Rev.

9. Klinkman MS. The process of choice of health care plan and provider: developmentof an integrated analytic framework. 1991;48:295-330.

Accessing PhysicianInformation on the Internet.

10. Stone EM, Heinold JW, Ewing LM, Schoenbaum SC. New York, NY: Commonwealth Fund; 2002.

J Gen Intern Med.

11. Haidet P, Stone DA, Juran D, Delbanco TL. Choosing a primary physician ina managed care environment: the importance of practice style and power[abstract]. 1998;13(suppl):67.

Patient Educ Counseling.

12. Krupat E, Rosenkranz RL, Yeager CM, Barnard K, Putnam SM, Inui TS. Thepractice orientations of physicians and patients: the effect of doctor-patient congruenceon satisfaction. 2000;39:49-59.

J Fam Pract.

13. Krupat E, Bell RA, Kravitz RL, Thom D, Azari R. When physicians andpatients think alike: patient-centered beliefs and their impact on satisfaction andtrust. 2001;50:1057-1062.

Psychol Health.

14. Krupat E, Yeager CM, Putnam S. Patient role expectations, doctor-patient fit,and visit satisfaction. 2000;15:707-719.

J Gen Intern Med.

15. Hsu J, Schmittdiel J, Krupat E, et al. Patient choice: a randomized controlledtrial of provider selection. 2003;18:319-325.

JAMA.

16. Rubin RHR, Gandek B, Rogers WH, Kosinski M, McHorney CA, Ware JE.Patients' ratings of outpatient visits in different practice settings: results from theMedical Outcomes Study. 1993;270:835-840.

J FamPract.

17. Thom DH, Campbell B. Patient-clinician trust: an exploratory study. 1997;44:169-176.