When is risk stratification by race or ethnicity justified in medical care?

Marshall H. Chin*, Catherine A. Humikowski

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

30 Scopus citations


Issues of race and ethnicity have been controversial in both clinical care and medical education. In daily practice, many physicians struggle to be culturally competent and avoid racial stereotyping. One educational development that makes this goal more complex is the rise of clinical epidemiology and Bayesian thinking. These population-based, probabilistic approaches to medicine help guide the diagnostic and therapeutic pathways for patients, and are foundations of the evidence-based medicine movement. Can Bayesian thinking be applied effectively to issues of race and ethnicity in medical care, or are the dangers of prejudicial stereotyping too great? The authors draw upon lessons from recent cases of racial profiling, and develop a conceptual framework for thinking about ethnicity as a clinical tool. In their typology of ethnicity as a proxy, they argue that the costs of using ethnicity as a proxy for socioeconomic status and behavior are too high, but that ethnicity may appropriately be used as an initial proxy for history, language, culture, and health beliefs. They discuss their approach within the context of new curricula in cultural competence, and argue that viewing the patient within a wider cultural setting can help guide the initial clinical approach, but individualized care is mandatory. Also, physicians must remain sensitive to the changing nature of cultural norms; thus lifelong learning and flexibility are necessary.

Original languageEnglish (US)
Pages (from-to)202-208
Number of pages7
JournalAcademic Medicine
Issue number3
StatePublished - 2002
Externally publishedYes

ASJC Scopus subject areas

  • Education

Fingerprint Dive into the research topics of 'When is risk stratification by race or ethnicity justified in medical care?'. Together they form a unique fingerprint.

Cite this