Background: Assessing residents by direct observation is the preferred assessment method for infrequently encountered subspecialty topics, but this is logistically challenging.
Objective: We developed an assessment framework for internal medicine (IM) residents in subspecialty topics, using tuberculosis diagnosis for proof of concept.
Methods: We used a 4-step process at 8 academic medical centers that entailed (1) creating a 10-item knowledge assessment tool; (2) pilot testing on a sample of 129 IM residents and infectious disease fellow volunteers to evaluate validity evidence; (3) implementing the final tool among 886 resident volunteers; and (4) assessing outcomes via retrospective chart review. Outcomes included tool score, item performance, and rates of obtaining recommended diagnostics.
Results: Following tool development, 10 infectious disease experts provided content validity. Pilot testing showed higher mean scores for fellows compared with residents (7 [SD = 1.8] versus 3.8 [SD = 1.7], respectively, P < .001) and a satisfactory Kuder-Richardson Formula 20 (0.72). Implementation of the tool revealed a 14-minute (SD = 2.0) mean completion time, 61% (541 of 886) response rate, 4.4 (SD = 1.6) mean score, and ≤ 57% correct response rate for 9 of 10 items. On chart review (n = 343), the rate of obtaining each recommended test was ≤ 43% (113 of 261), except for chest x-rays (96%, 328 of 343).
Conclusions: Our assessment framework revealed knowledge and practice gaps in tuberculosis diagnosis in IM residents. Adopting this approach may help ensure assessment is not limited to frequently encountered topics.