Do Estimates of Treatment Risk Based on Clinical Vignettes Differ by Physician Gender?

Abstract

Background: Clinical vignettes are frequently used as instructional and evaluative instruments for physicians. Physicians' gender is a source of unconscious bias in treatment recommendations. This study assessed whether interpretation of information in clinical vignettes differed by physicians' gender as a possible source of unconscious bias.

Methods: Thoracic surgeons and physicians in cardiothoracic surgical training were asked to provide estimates of major complication rates for lung resection on the basis of anonymized clinical vignettes of patients undergoing lung resection. Vignettes were categorized as low, average, and high risk by using a sum of Charlson Comorbidity Index (possible range, 0 to 37) and a combined physiologic score, EVAD (forced expiratory volume in 1 second, diffusing capacity of lung for carbon monoxide, age; possible range, 0 to 12); participants were not aware of the risk scores or vignette categories. Generalized estimating equation linear regression models were fit with risk scores treated as a continuous independent variable.

Results: A total of 247 physicians (105 practicing surgeons, 142 trainees; 203 men, 44 women) participated in one or more of the studies. Nearly all (103; 98%) of the practicing surgeons rated themselves as competent or expert in lung resection compared with 77 (54%) of the trainees (p < 0.001). Participants' complication estimates mirrored both vignette risk category and combined risk score. There was no significant difference between men and women physicians in their estimates of complication rates.

Conclusions: Unconscious bias related to physicians' gender is not associated with differential use of information in clinical vignettes. Any possible bias may arise from face-to-face interactions with patients. Research into physicians' and patients' gender differences during such interactions is warranted.