Schulman KA, Berlin JA, Harless W, Kerner JF, Sistrunk S, Gersh BJ, Dube R, Taleghani CK, Burke JE, Williams S, Eisenberg JM, Escarce JJ.
The effect of race and sex on physicians’ recommendations for cardiac catheterization.
N Engl J Med 1999;340(8):618-26.
(Erratum in: N Engl J Med;340(14):1130.
Comment in: N Engl J Med. 1999;341(26):2021-2. N Engl J Med. 1999;341(26):2021; discussion 2022. N Engl J Med. 1999;341(4):279-83; discussion 286-7. N Engl J Med. 1999;341(4):285-7. N Engl J Med. 1999;341(4):285; discussion 286-7.)

The goal of this study was to assess, in a controlled experiment, physicians' treatment recommendations for patients presenting with various types of chest pain in order to determine the extent to which physicians are responsible for the differences in cardiac treatment with respect to race and sex.

The study presented video recorded interviews and tests presenting descriptions of patients with chest patient to clinicians. Patients in the videos were varied by race (black or white), sex, age (55 or 70 years), level or coronary risk (low or high), type of chest paint, and the results of a stress test (a total of 144 different scenarios). Clinicians were then asked via a computerized survey instrument to decide how to properly manage these patients; each clinician viewed one randomly-selected patient scenario. Clinicians participating in the study were physicians in full-time clinical practice who attended the 1997 annual meeting for the ACP or the 1996 annual meeting for the AAFP.

Seven hundred and twenty physicians participated in this study. Female physicians were more frequently assigned to black female patients. In univariate analyses, the race (OR for white versus black=0.60, 95% CI=0.40, 0.90) and sex (OR for male versus female=0.60, 95% CI=0.40, 0.90) of the patient were significantly associated with physicians’ decisions about whether to make referrals for cardiac catheterization. In multivariate analyses that included all covariates (above), both race and sex were independent predictors of referrals. Additionally, there was a significant interaction effect that showed that black females were less likely than white males to be referred (but black males and white females did not differ from white males).

There were substantial differences among the race/sex groups with regard to physician perceptions of the characteristics of the patients (including hostility, dependent, sad/happy, affect, worry, and SES) and the likelihood of specific behaviors (including overreporting symptoms, suing, and compliance).

The authors conclude that the race and sex of the patient influence the recommendations of physicians for the management of chest pain and suggest that there is bias on the part of the clinicians. They note, however, that the form of bias could not be assessed. "Bias may represent overt prejudice on the part of the physicians or, more likely, could be the result of subconscious perceptions rather than deliberate actions or thoughts."

This article was widely cited, but there were also critiques of the way in which the authors interpreted the results. The first criticism was that the authors reported odds ratios, which in this type of study design are larger in magnitude than the corresponding relative risk would be. Given that the main analyses were logistic regressions, the reporting of odds ratios is both correct and conventional. However, the journal editors felt it would have been prudent to clarify the relationship between the OR and the RR for readers.

The second criticism was that the authors over-emphasized the independent effects of race and sex, when the data showed that the physician bias effect was limited to black women. Further, the authors subsequently showed that these effects were strongest among black women over the age of 70 years. However, the three-way interaction of race-sex-age was not part of the original study hypothesis, and the power of this study to detect significant three-way interaction effects was limited. Thus, it was not necessarily inappropriate for the authors or the journal to omit this information from the final paper.

Given the originality of the approach taken in this paper, the extent of corroborating data, and the strength of the findings (the OR for black females versus white males in referral for catheterization was 0.04, p<0.004), these data are of substantial importance despite the criticism.