A randomized trial involving a simulated clinical encounter suggests that physician gender and race may not significantly affect patient satisfaction with or confidence in physicians. The authors cautioned, however, that their results “should not be interpreted as contradicting the lived experiences of discrimination reported by physicians from underrepresented groups.”
Published in JAMA Network Open, the findings are from a study in which a geographically diverse sample of nearly 3,600 online US respondents were recruited from two crowdsourcing platforms: Amazon Mechanical Turk (MTurk) and Lucid. The participants were shown a clinical vignette involving diagnosis of gastroenteritis as evaluated by an emergency medicine physician and asked to play the role of the patient. They were shown the simulated physician’s image and a written diagnosis of gastroenteritis with a conservative treatment plan, along with a contradictory diagnosis of possible appendicitis from an Online Doc Symptom Checker with a more aggressive treatment plan. The Checker was a fabrication created for the study.
Participants were randomly assigned to see a simulated physician with different gender and race, with 823 assigned to black women, 791 to black men, 821 to white women, and 835 to white men. The images were from the Chicago Face Database, a research database for photographs of real human faces of varying gender/ethnicity. The authors used a set of 10 images for each treatment arm (e.g. 10 black women, 10 black men, 10 white women, 10 white men).
None of the participants were pregnant, had a current or previous diagnosis of cancer or a history of abdominal surgery. All MTurk participants were paid $1.00 as compensation.
Each study took approximately 10 minutes to complete and participants were blinded to the rationale for the research. They were asked to rate the simulated physician on confidence in him or her, satisfaction with care, likelihood of recommending the physician, trust in the physician’s diagnosis, and likelihood that they would request additional tests.
The authors found no significant differences in participant satisfaction and physician confidence for the white male physician control physicians, black male or black female physicians. Machine learning with Bayesian additive regression trees revealed no evidence of treatment effect heterogeneity as a function of participants’ race, gender, racial prejudice, or sexism.
Interestingly, approximately 40% of participants from Amazon Mechanical Turk and 34% of those from Lucid endorse some group-level superiority of white individuals compared with black individuals. The authors acknowledged that, “it is theoretically possible that some participants may have discerned the purpose of the study and censored their prejudice against female and black physicians to appear more socially desirable, thereby attenuating estimated treatment effects.” Nevertheless, they said that their results “add to a growing body of observational studies investigating patient bias against female and minority group physicians, most of which have not found evidence of systematic bias.”