Group assessments of resident physicians improve reliability and decrease halo error

Background: Individual faculty assessments of resident competency are complicated by inconsistent application of standards, lack of reliability, and the "halo" effect. Objective: We determined whether the addition of faculty group assessments of residents in an ambulatory clinic, compared with individual faculty-of-resident assessments alone, have better reliability and reduced halo effects. Design: This prospective, longitudinal study was performed in the outpatient continuity clinics of a large internal medicine residency program. Main measures: Faculty-on-resident and group faculty-on-resident assessment scores were used for comparison. Key results: Overall mean scores were significantly higher for group than individual assessments (3.92±0.51 vs. 3.83±0.38, p=0.0001). Overall inter-rater reliability increased when combining group and individual assessments compared to individual assessments alone (intraclass correlation coefficient, 95% CI=0.828, 0.785-0.866 vs. 0.749, 0.686-0.804). Inter-item correlations were less for group (0.49) than individual (0.68) assessments. Conclusions: This study demonstrates improved inter-rater reliability and reduced range restriction (halo effect) of resident assessment across multiple performance domains by adding the group assessmentmethod to traditional individual faculty-on-resident assessment. This feasible model could help graduate medical education programs achieve more reliable and discriminating resident assessments.