A Pilot Study of the Generalizability of Preclinical Entrustment Assessments in Undergraduate Medical Education

The reproducibility and consistency of assessments of entrustable professional activities (EPAs) in undergraduate medical education (UME) have been identified as potential areas of concern. EPAs were designed to facilitate workplace-based assessments by faculty with a shared mental model of a task who could observe a trainee complete the task multiple times. In UME, trainees are frequently assessed outside the workplace by faculty who only observe a task once.

In November 2019, the authors conducted a generalizability study (G-study) to examine the impact of student, faculty, case, and faculty familiarity with the student on the reliability of 162 entrustment assessments completed in a preclerkship environment. Three faculty were recruited to evaluate 18 students completing 3 standardized patient (SP) cases. Faculty familiarity with each student was determined. Decision studies were also completed. Secondary analysis of the relationship between student performance and entrustment (scoring inference) compared average SP checklist scores and entrustment scores.

G-study analysis revealed that entrustment assessments struggled to achieve moderate reliability. The student accounted for 30.1% of the variance in entrustment scores with minimal influence from faculty and case, while the relationship between student and faculty accounted for 26.1% of the variance. G-study analysis also revealed a difference in generalizability between assessments by unfamiliar (φ = 0.75) and familiar (φ = 0.27) faculty. Subanalyses showed that entrustment assessments by familiar faculty were moderately correlated to average SP checklist scores (r = 0.44, P < .001), while those by unfamiliar faculty were weakly correlated (r = 0.16, P = .13).

While faculty and case had a limited impact on the generalizability of entrustment assessments made outside the workplace in UME, faculty who were familiar with a student's ability had a notable impact on generalizability and potentially on the scoring validity of entrustment assessments, which warrants further study.