Understanding and Addressing Provider Contribution to Disparities

1. Understanding and Addressing Provider Contribution to Disparities Michelle van Ryn, Ph.D., M.P.H. Director, Colorectal Cancer Quality Enhancement Research Initiative, VA Health Services Research, Minneapolis VAMC Associate Professor, University of Minnesota School of Public Health Michelle.vanRyn@med.va.gov

2. Unequal Treatment: Confronting racial and ethnic disparities in health care.IOM, National Academy of Sciences Press, 2002 • Disparities consistently found across a wide range of disease areas and clinical services. • Disparities are found even when clinical factors, such as stage of disease presentation, co-morbidities, age, and severity of disease are taken into account. • Disparities are found across a range of clinical settings, including public and private hospitals, teaching and non-teaching hospitals, etc. • Disparities in care are associated with higher mortality among minorities

3. Why? • Insurance a major contributor to receipt of guideline-consistent care (even in MD mention of a given treatment care). • Treatment site associated with residential segregation is close behind. • However, disparities persist independent of payer and within each payer category and have been documented independent of treatment site. • Disparities by class/SES probably as significant. • However, race/ethnicity disparities appear to persist independent of class.

4. Understanding the Provider Contribution to Disparities Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider Interpretation of Symptoms Explicit (conscious) Pt Race/ethnicity Treatment or Service Pt Received Provider Decision-making (Diagnosis,Treatment) Implicit (unconscious) Pt Behavior in Encounter (e.g., question-asking self-disclosure, assertiveness) Provider Behavior in Encounter (e.g., participatory style, warmth, content, information giving, question-asking) Class Encounter characteristics Pt Satisfaction Pt Behavior (e.g. self management, information-seeking, utilization) Pt Cognitive & Affective Factors (e.g., acceptance of advice, attitude, self-efficacy, intention, feelings of competence, attitude toward med care, trust) Culture van Ryn, Grantmakers in Health, 11- 6-03

5. Why? Patient preference hypothesis. • Early favorite. • Strong evidence of subgroup differences in health beliefs. • Failure to link such health belief differences to treatment differences. • Studies examining role of patient preferences find that disparities persist controlling for patient preferences. • (e.g. Ayanian et al, 1999, Conigliaro, 2002; Hannan & van Ryn et al, 1999, Kressin et al, 2002, Petersen, et al, 2002; van Ryn, et al, 2000; Whittle, et al, 1997). Pt Race/ethnicity Treatment or Service Pt Received Provider Decision-making (Diagnosis,Treatment) Class Pt Behavior (e.g. self management, information-seeking, utilization) Pt Cognitive & Affective Factors (e.g., acceptance of advice, attitude, self-efficacy, intention, feelings of competence, attitude toward med care, trust) Culture

6. Why? Provider decision-making hypothesis Setting Provider Characteristics Provider Decision-making (Diagnosis,Treatment) Treatment or Service Received Provider decision-making has been shown to be a factor in disparities in: Pt Race/ethnicity • Kidney transplant & placement on waiting lists, even among children. • Cardiac care (both preventive & aggressive tx for CAD). • Quality of in-hospital care. • Pain assessment, detection & treatment, even among children, in ED for fractures, nursing homes, for cancer pain, at end-of-life. • Mental health assessment, diagnosis and treatment • Treatment for peripheral artery disease: Amputation vs. lower-extremity revascularization (in VA). Class Culture

7. What causes provider-driven disparities? Several possible mechanisms: • Different, less evidence-based, providers see non-whites (site of care explanation). • Providers use different decision making processes with whites than non-whites. • Unmeasured variables that are (or are perceived to be) associated with patient race/ethnicity influence provider decision-making. • Overt bias (unlikely to explain magnitude of effect given factors against). “Evidence is only one of many factors that play a role in clinical decision-making.”(Rubenfeld, 2001)

8. There is ample empirical evidence that patient sex, age, diagnosis, marital status, sexual orientation, sickness and race/ethnicity can influence providers’ beliefs about and expectations of patients independent of other factors. Example: Cardiac patients’ race/ethnicity and SES were found to independently and negatively influence physicians’ post-encounter ratings of patients’ personality, education, intelligence, career demands, and likely adherence (independent of physicians’ characteristics and patients’ personality, clinical, and socio-demographic characteristics) e.g., see Gerbert B., 1984; Hall et al., 1993; Kearney et al, 2000; Krupat et al., 1999; lewis et al., 1990; Revenson, 1989; Shortt, 2001; Ross et al., 1991; Stern et al., 2001; Stern and Arenson, 1989; Kelly et al., 1987; Schulman et al., 1999; van Ryn and Burke, 2000

9. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting Beliefs regarding “base-rate” in populations Provider cognition and affect regarding help-seeker Effects moderated by clinical uncertainty? Explicit (conscious) Pt Race/ethnicity Implicit (unconscious) • There is evidence that providers’ over-apply population estimates to individual patients. • Epidemiologic data on subgroup likelihood may be incorporated into physicians’ general belief systems such that group data is inaccurately applied to individuals (statistical discrimination). • (McKinley et al, 1986; van Ryn & Burke, 2000; Balsa & McGuire, 2001) Class Culture

10. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider perception of likelihood of adherence Provider Decision-making (Diagnosis,Treatment) Explicit (conscious) Pt Race/ethnicity Implicit (unconscious) • Bogart and colleagues found that physicians were more likely to provide highly active antiretroviral therapy (HAART) to HIV/AIDS patients when they perceived them to be likely to be adherent. • Randomly assigned physicians to review patient vignettes that varied only on patient gender, disease severity, ethnicity, and risk group. Physicians were significantly more likely to rate the African American simulated patients as non-adherent. Class Culture Bogart et al., 2000; Bogart et al, 2001

11. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider perception of adherence, likelihood of adequate SS Provider Decision-making (Diagnosis,Treatment) Explicit (conscious) Pt Race/ethnicity Implicit (unconscious) • Cardiologist ratings of patients' likelihood of having adequate social support and/or participating in cardiac rehabilitation was found to predict physicians' recommendations for revascularization (independent of clinical appropriateness for revascularization and other demographic characteristics). • Same group of physicians were more likely to rate African American patients as lacking in social support and unlikely to participate in cardiac rehabilitation than white patients, indep of pt MOS social support score. Class Culture van Ryn and Burke, 2000; van Ryn et al., 2001

12. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting Provider cognition and affect regarding help-seeker Explicit (conscious) Pt Race/ethnicity Provider Decision-making (Diagnosis,Treatment) Implicit (unconscious) • Cardiologist ratings of patients' likelihood of having adequate social support and/or participating in cardiac rehabilitation as found to predict physicians' recommendations for revascularization (independent of clinical appropriateness for revascularization and other demographic characteristics). • Same group of physicians were more likely to rate African American patients as lacking in social support and unlikely to participate in cardiac rehabilitation than white patients, indep of pt MOS social support score. Class Culture van Ryn and Burke, 2000; van Ryn et al., 2001

13. There is ample evidence that social and behavioral factors influences provider clinical decision-making, both explicitly (consciously) and implicitly (unconsciously). • When is this OK? • How are these “indications” assessed? • Clinicians often have very little training or guidance in valid and reliable ways to assess the likelihood that a given patient has a social or behavioral characteristic. • Considerable support for hypothesis that clinicians rely on “base-rates”, e.g. over-apply population statistics, to individual patients. 1. Shye et al 1998; Feldman et al, 1997; McKinlay et al,1996; 1997; Christakis, 1993; Wertz, 1993; Bradley, 1992; Grabet, 2000; Halm et al., 2000; Lockey and harden, 2001)

14. We assign different meaning to the same behavior depending on the characteristics of the actor. Effect is exacerbated for ambiguous behavior1 • Examples: • Providers underestimated the amount of pain minority cancer patients were experiencing more often than white counterparts. (Sollner et al., 2001) • Women’s chest pain more likely to be attributed to psychogenic causes even in presence of abnormal cardiac test results (Tobin et al, 1987) 1. Kunda, 1999; Kunda & Sherman-Williams 1993; Darley & Gross 2000; Darley & Gross 2000; Locksley, et al.1982; Sagar & Schofield 1980; Lepore & Brown 1997;Darley and Gross, 1983; Dunning and Sherman, 1997)

15. Provider Uncertainty Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider Interpretation of Symptoms Treatment or Service Pt Received Explicit (conscious) Pt Race/ethnicity Provider Decision-making (Diagnosis,Treatment) Implicit (unconscious) • Oncologists’ were less able to accurately identify high levels of distress among their low income patients (& less likely to provide an appropriate therapeutic response) Cleeland et al., 1997. • Providers reported more uncertainty in their management of breast cancer for women they thought were low SES (McKinlay,et al, 1998) Class Culture

16. Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider Interpretation of Symptoms Treatment or Service Pt Received Explicit (conscious) Pt Race/ethnicity Provider Decision-making (Diagnosis,Treatment) Implicit (unconscious) Pt Behavior in Encounter (e.g., question-asking self-disclosure, assertiveness) Both medical students’ & MDs’ gave of normal children more negative assessments when they were told the child had been born prematurely. (Stern et al,2000;2001) Class Culture

17. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting • The Self-fulfilling Prophecy • Unconsciously activated stereotypes affect our behavior Provider cognition and affect regarding help-seeker Explicit (conscious) Pt Race/ethnicity Implicit (unconscious) Pt Behavior in Encounter (e.g., question-asking self-disclosure, assertiveness) Provider Behavior in Encounter (e.g., participatory style, warmth, content, information giving, question-asking) Class Encounter characteristics Our behavior toward others influences their behavior in turn Culture

18. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting • Extensively studied in educational & job interviewing. • Interviewers' behavior influenced by race of applicant • Interviewer behavior influences applicant behavior. Provider cognition and affect regarding help-seeker Explicit (conscious) Pt Race/ethnicity Implicit (unconscious) Pt Behavior in Encounter (e.g., question-asking self-disclosure, assertiveness) Provider Behavior in Encounter (e.g., participatory style, warmth, content, information giving, question-asking) content & affect Class Encounter characteristics Implicated as factor in studies of disparities in dx of AD/HD, autism Culture

19. Hypothesized Mechanisms Through Which Providers Contribute to Disparities Provider Characteristics Setting Provider cognition and affect regarding help-seeker Provider Interpretation of Symptoms Explicit (conscious) Pt Race/ethnicity Treatment or Service Pt Received Provider Decision-making (Diagnosis,Treatment) Implicit (unconscious) Pt Behavior in Encounter (e.g., question-asking self-disclosure, assertiveness) Provider Behavior in Encounter (e.g., participatory style, warmth, content, information giving, question-asking) Class Encounter characteristics • Differences in provider intake behaviors may have resulted in • Differences information obtained in the ED, and thus created • Differences in diagnoses Culture

20. Summary • Considerable evidence that providers do incorporate beliefs regarding patient social factors into their decisions. • Evidence of over-application of stereotypes and population estimates to non-whites, low SES. • Evidence that provider perceptions of patients’ social/behavioral factors mediates at least some of the observed disparities (e.g. Bogart et al, van Ryn et al) • Class effects likely to be as large or larger • Lots of noise about need to take patient social context, beliefs and attitudes into account • Relative silence on when, why, how to assess and apply.

21. What role do social and behavioral factors play? What role should they play? • Guidelines and performance measures are notably sketchy on the role of social or behavioral factors. • Some specify behavior risk factors that either are, or may be perceived to be, unevenly distributed by race/ethnicity (e.g. alcohol abuse). • Others contain implicit social and behavioral expectations – leaving it up to the provider to determine whether they are realistic for the patient. Unclear what provider should do if s/he judges patient will not adhere (e.g. AD/HD - adherence to repeat visits).

22. Lack of clarity on role of social factors • Lack of valid methods, encounter time, and provider training in assessing patient status on such factors combined with • Prevalent stereotypes reinforced by pop. statistics • fertile ground for disparities in treatment • Need for explicit attention to the development of consensus on role of social factors in clinical decision-making. • What factors matter when? • How do we assess patient status on these factors? • How do we incorporate such assessment into current clinical practice?

23. Recommended Funding Priorities • Development of consensus on role of social factors in clinical decision-making… • What factors matter when? • How do we assess patient status on these factors? • How do we incorporate such assessment into current clinical practice? • System change to create conditions that support providers in accurate assessment and quality communication. • Reimbursement methodologies • Time allotted for encounters • Increase non-white workforce • Provider training incorporating best of cultural competence and stereotype reduction strategies

24. Current Initiatives Increase non-white workforce • Intrinsically good idea: improving opportunity structure, diverse organizational members create beneficial effects. • I support it. • I wonder how far will it take us in reducing disparities? • Class & professional socialization effects likely to persist • Minority groups subject to same, internalized stereotypes and majority group members. • I support it.

25. Current Initiatives Increase non-white workforce Positive: • African Americans who saw African American provider reported higher satisfaction levels than counterparts (LaVeist and Carroll, 2002) • Race concordance increased ratings of providers’ participatory style. So did seeing a woman provider for both sexes and races (Cooper-Patrick et al, 1999) Less Promising • Race concordance did not improve dx of depression (Cooper, 2002) • SES concordance, but not race concordance, positively predicted time in encounter (Malat, 2001)

26. Current Initiatives Increase non-white workforce The Contact Hypothesis: prejudice is a result of ignorance. • “Getting to know one another” will solve problem. • Refinement of hypothesis: under certain conditions contact can reduce prejudice. (Allport) • equal status (no power or resource differential) • common goals & interdependence • no competition • authority sanctions contact • (success at joint tasks) • Recent meta-analysis of 203 studies (73% US) • Contact between individuals under stated conditions fairly consistently reduced individual prejudice. • More so for majority than minority group members.

27. Current Initiatives Increase non-white workforce • In some instances where conditions are not met, bias increases. • Are the conditions descriptive of medical student and residents’ joint experiences? • What if jobs are perceived to be scarce? Also: • There is evidence that when faced with disconfirming instances, we subtype. • More likely to subtype when disconfirming instances hang together in some way.

28. Current Initiatives Increase non-white workforce Unresolved issues: • What are implications of allowing (or encouraging) people to choose a provider by race/ethnicity? • What about social class discordance? • Will working with more non-whites as peers in organization influence white provider implicit bias? • Maybe, could make an argument based on current research either way.

29. Using EBM & QI to ameliorate disparities • Some evidence that intensive Evidence-Based Medicine Quality Improvement interventions differentially benefit the most vulnerable (who have furthest to go). Thus, they show promise for eliminating race/ethnicity disparities in care. Owen WF, Jr., Szczech LA, Frankenfield DL. Healthcare system interventions for inequality in quality: corrective action through evidence-based medicine. J Natl Med Assoc. Aug 2002;94(8 Suppl):83S-91S. Sehgal AR. Impact of quality improvement efforts on race and sex disparities in hemodialysis. Jama. Feb 26 2003;289(8):996-1000. Smith JH, Jr. Eliminating health disparities: our mission, our vision, our cause. Am Fam Physician. Oct 15 2001;64(8):1333-1334.

30. Using EBM to ameliorate disparities Caveats: • QI approaches limited to conditions with evidence base. • Must account for or address effect of race/ethnicity and class on assessment • Explicit attention to social and behavioral factors as contributors to treatment decision

31. Current Initiatives Cultural Competence Training • Training types that give information of “typical” characteristics and issues associated with a given subgroup likely to exacerbate problem. • Improving interpersonal and communication skills • eliciting self-disclosure • active listening • self-awareness … show promise for increasing likelihood providers will obtain individual information rather than rely on base-rates or implicit stereotypes.

32. Current Initiatives Cultural Competence Training • Possible unintended adverse effects: • Efforts at stereotype suppression can backfire (“stereotype suppression rebound”). • When experimental participants are asked to suppress stereotypes in arriving at judgements of an individual, they can do so. • However, in some conditions initial suppression of stereotypes leads to increased activation and use in other settings encountered shortly thereafter. Macrae et all, 1994. Out of mind but back in sight: stereotypes on the rebound. Journal of Personality and Social-Psychology, 67, 808-817

33. Current InitiativesCultural Competence Training • Do these programs affect explicit beliefs? If so, for how long? • (Some evidence of at least a short term effect for program lead by African American professor with diverse participants & affective component(Rudman et al, 2001). • If they do, increases in conscious egalitarian intention should increase motivation to overcome automatic deleterious beliefs. • Individuals with explicit egalitarian belief systems are less likely to activate negative implicit beliefs when exposure is neutral (but not when negative). Blair, I. (2001). Implicit Stereotypes and Prejudice. Cognitive Social Psychology: The Princeton Symposium on the Legacy and Future of Social Cognition. G. Moskowitz. NJ, Lawrence Erlbaum.

34. Current InitiativesCultural Competence Training • Under certain conditions individuals can consciously replace automatically activated stereotype with egalitarian response: • Must be aware of potential for judgment, emotions and behaviors to be biased. • Have awareness of stereotype activation. • Have high level of motivation. • Have sufficient cognitive resources (time and cognitive capacity). Blair, I. (2001). Implicit Stereotypes and Prejudice. Cognitive Social Psychology: The Princeton Symposium on the Legacy and Future of Social Cognition. G. Moskowitz. NJ, Lawrence Erlbaum.

35. Current InitiativesCultural Competence TrainingPotentially promising findings • Perspective-taking has been successfully used to reduce stereotypes and prejudice. (Compared to a no-instruction control group and a “stereotype suppression group” that was instructed to actively try to avoid thinking about the person in a stereotypic manner.) • For example, Whites who wrote about a day in the life of an elderly or Black person, showed less explicit and implicit stereotyping • "imagine a day in the life of this individual as if you were that person, looking at the world through his eyes and walking through the world in his shoes.“ Galinsky, A. D. & Moskowitz, G. B. (2000). Decreasing Stereotype Expression, Stereotype Accessibility, and In-Group Favoritism. Journal of Personality and Social Psychology.

36. Current Initiatives Cultural Competence Training • Current cultural competence programs target explicit beliefs. • Test of addition of strategies to overcome automatic activation of negative beliefs needed. • Relationship building, listening, and communication skills predictive of a wide range of improved outcomes. • Unlikely to completely ameliorate disparities as sole approach. • EVALUATION OF PROGRAMS NEEDED

37. QI monitoring and feedback strategies Disparities-specific QI? Some recent studies show that when low-prejudice (in explicit beliefs) people get feedback regarding their behavior that are discrepant from their conscious beliefs &… • Experience guilt or negative self-evaluation as a result & • Have freed up cognitive resources… …they are able to exert control over implicit negative beliefs and reactions in a subsequent situation.

38. Evidence-based & rigorous methods for incorporating social factors into clinical decision-making • Lots of noise about need to take patient social context into account. • Relative silence on when, why, how to assess and apply. • Considerable evidence that providers do incorporate beliefs regarding patient social factors into their decisions. • Evidence of over-application of stereotypes and population estimates to non-whites, low SES. • Evidence this mediates at least some of the observed disparities. (e.g. Bogart et al, van Ryn et al) • Class effects likely to be as large or larger.

39. Need for system change:Situational factors influence the likelihood of stereotype activation and application Considerable evidence that stereotypes are more likely to be activated and applied (vs. individual information) when individuals are: • tired • distracted • pressed for time • anxious • cognitively busy Probably because these conditions leave little cognitive resources for processing individuating information and/or suppressing stereotypes.

40. Conclusion • Considerably more research on provider contribution to disparities needed. • Addressing the provider contribution to disparities likely to require a multi-modal and multilevel approach. • There are few ideas for reducing disparities that have not already been suggested in other contexts for other or related reasons. • Strategies to reduce disparities are likely to improve care for all. • Significant devotion of resources and will required.

41. Recommended Funding Priorities • Development of consensus on role of social factors in clinical decision-making… • What factors matter when? • How do we assess patient status on these factors? • How do we incorporate such assessment into current clinical practice? • System change to create conditions that support providers in accurate assessment and quality communication. • Reimbursement methodologies • Time allotted for encounters • Increase non-white workforce • Provider training incorporating best of cultural competence and stereotype reduction strategies

42. Despite massive evidence to the contrary… … evidence indicates that people tend to believe that they can see through stereotypes and are unprejudiced in their own perceptions and expectancies.