Methods

Evidence Search. We developed strategies to search the existing literature addressing each of the four lines of evidence discussed above: service patterns, concordance, trust in health care, and professional advocacy. We searched the MEDLINE, HealthSTAR, and CINAHL databases using search terms available in each database. For the concordance hypothesis, we also searched the PsycINFO database, because we knew that many of the studies related to patient-practitioner concordance were conducted in the context of mental health counseling and published in journals not included in the other three databases. We supplemented these database searches in four ways. First, we conducted a “gray” literature search, for studies that may not have been published as journal articles but rather as monographs or book chapters. Second, we manually searched the reference lists of included studies and relevant review articles. Third, we searched selected Web sites for relevant references. Finally, we presented our initial results to several audiences including experts in health professions diversity and solicited their input on relevant evidence not yet included in our review. We retrieved articles or documents identified by these supplementary approaches and reviewed them for relevance.

We reviewed titles and abstracts from our database searches and retrieved full articles for those that met inclusion criteria specific to each of our four lines of evidence. When it was not possible to determine from the title or abstract whether an article should be included, we reviewed the full article. We limited our review to articles that included original, empirical data generated within the United States. We believe that, due to the highly variable social significance and meaning of race, ethnicity, and social class, data from other countries would not be sufficiently generalizable to the U.S. In the same vein, we limited our review to studies published in or after 1985, since the social significance and meaning of race in particular has changed, and continues to change, over time. Studies published before 1985 often included data from a period when minority representation in the health professions was substantially lower and when racial attitudes were closer to those of the pre-Civil Rights era than they are today.

Evidence Abstraction and Synthesis. We reviewed all retrieved articles and included those that met our inclusion criteria. We critically reviewed the included articles and abstracted relevant information about the health professional groups and patient populations examined, the principal results of the study, and important study features or limitations. For lines of evidence with more than a small number of studies available (service patterns and concordance), we tabulated the abstracted information in evidence tables, to facilitate comparison, discussion, and qualitative synthesis of the evidence (Appendices A, B, C). For minority professional service patterns and racial, ethnic, and language concordance, we also created tables in which we compared the number of studies that supported each hypothesis with the number that either did not support or refuted the hypothesis. We stratified these tables by race/ethnicity. For example, we counted the number of studies that supported the hypothesis that patient-practitioner race concordance improves quality of care for African-Americans, as well as the number that did not support this hypothesis or refuted it. The purpose of these tables was to provide perspective on the balance of the evidence for each hypothesis. In creating these tables, we found that some studies assessed multiple outcomes related to a single hypothesis (e.g., self-reported delay in seeking care and emergency department visits as measures of access or utilization). When studies like this found evidence supporting the hypothesis in question for one outcome measure, we counted the study as supporting the hypothesis if, for the second outcome measure, the hypothesis was not refuted. If for the second outcome measure the hypothesis was refuted, we counted the study as one in which the overall hypothesis was not supported.

In this paper we present the overall results of our review and highlight findings from representative studies. The highlighted studies illustrate our qualitative synthesis of the evidence and make points that we considered important to understanding the state of the current evidence base—including its implications and limitations—on the rationale for health professions diversity.