Using case vignettes to study the presence of outcome, hindsight, and implicit bias in acute unplanned medical care: a cross-sectional study

BACKGROUND AND IMPORTANCE: Various biases can impact decision-making and judgment of case quality in the Emergency Department (ED). Outcome and hindsight bias can lead to wrong retrospective judgment of care quality, and implicit bias can result in unjust treatment differences in the ED based on irrelevant patient characteristics. OBJECTIVES: First, to evaluate the extent to which knowledge of an outcome influences physicians' quality of care assessment. Secondly, to examine whether patients with functional disorders receive different treatment compared to patients with a somatic past medical history. DESIGN: A web-based cross-sectional study in which physicians received case vignettes with a case description and care provided. Physicians were informed about vignette outcomes in a randomized way (no, good, or bad outcome). Physicians rated quality of care for four casevignettes with different outcomes. Subsequently, they received two more case vignettes. Physicians were informed about the past medical history of the patient in a randomized way (somatic or functional). Physicians made treatment and diagnostic decisions for both cases. SETTING AND PARTICIPANTS: One hundred ninety-one Dutch emergency physicians (EPs) and general practitioners (GPs) participated. OUTCOME MEASURES AND ANALYSIS: Quality of care was rated on a Likert scale (0-5) and dichotomized as adequate (yes/no). Physicians estimated the likelihood of patients experiencing a bad outcome for hindsight bias. For the second objective, physicians decided on prescribing analgesics and additional diagnostic tests. MAIN RESULTS: Large differences existed in rated quality of care for three out of four vignettes based on different case outcomes. For example, physicians rated the quality of care as adequate in 44% (95% CI 33-57%) for an abdominal pain case with a bad outcome, compared to 88% (95% CI 78-94%) for a good outcome, and 84% (95% CI 73-91%) for no outcome (P < 0.01). The estimated likelihood of a bad outcome was higher if physicians received a vignette with a bad patient outcome. Fewer diagnostic tests were performed and fewer opioids were prescribed for patients with a functional disorder. CONCLUSION: Outcome, hindsight, and implicit bias significantly influence decision-making and care quality assessment by Dutch EPs and GPs.