Unconscious bias in students of health professions : an experimental vignette study

Abstract

Background: Unconscious bias refers to automatic, implicit attitudes or stereotypes that influence our understanding, decisions, and actions without our conscious awareness. It is recognised as a significant problem in healthcare contributing to disparities in treatment. To date it remains unclear how unconscious bias towards patients develops among health professionals. One hypothesis is that such bias is acquired during education, either through teaching content or by observing other health professionals interacting with patients and adopting their behaviour. We investigated whether health care students show an unconscious bias and whether there are indications that it develops during their professional education. Methods: We conducted a factorial survey experiment with bachelor’s and master’s students enrolled in various health professions’ programmes. Unconscious bias was assessed using three written vignettes describing clinical situations involving patients. Participants were asked to evaluate their likelihood of helping immediately, expected patient adherence, and expected quality of the patient relationship. Vignette dimensions contained common sources of bias that were experimentally manipulated: gender, age, socio-economic status, migration status, diagnosis (physical or mental illness), and sexual orientation. Multivariable regression models were used to estimate the causal effects of patient characteristics on vignettes on students’ responses. Additionally, an Implicit Association Test (IAT) on unconscious bias regarding homosexuality was used to measure implicit bias. Explicit attitudes were assessed via self-report. Results: A total of 470 students (response rate 21.5%) participated. Vignette analysis showed no differences in stated helping intention, adherence expectations, or relationship assessments with regard to patient characteristics such as gender, age, socio-economic status, foreign name, sexual orientation or diagnosis. No systematic differences were observed for subgroups of participants such as year of study, programme type, prior work experience, or reported exposure to bias behaviour by health care staff. Similarly, IAT results indicated no overall bias towards homosexuality. Conclusions: We found no evidence of systematic unconscious bias among students’ helping intentions, expected patient adherence, and expected patient relations across various patient characteristics. Comparisons across study years and programs provided no indication that such biases emerge or intensify during training. If replicated, these results would be encouraging, as it indicates an absence of unconscious bias in health care students.