ASSESSMENT OF A PATTERN-RECOGNITION EXAMINATION IN A CLINICAL CLERKSHIP

Purpose. To develop a pattern-recognition examination (called PAT) for use in assessing surgical clerkship students. Method. The PAT was dev eloped in 1992 - 93 by the surgery faculty at the Wright State Univers ity School of Medicine. The top 30 diagnoses of patients seen by clerk ship students in 1991-92 were identified, and possible presenting comp laints were selected for each diagnosis. Then for each complaint facul ty selected 18 common or catastrophic diagnoses, and a distinctive pat ient description was written for each diagnosis. An examination of 200 items was compiled covering 20 presenting complaints, each with a set of ten patient descriptions and 18 diagnoses from which to choose the most likely diagnosis. Scoring was by the percentage of correct respo nses. The PAT was given to 77 students during the 1992-93 clerkship al ong with other forms of evaluation. The PAT was also administered to 1 8 surgical residents. Results. The students' mean score on the PAT was 65%, SD, 12.7. Split-half reliability was high (Spearman-Brown r = .9 1). The students' PAT performances correlated strongly with other cued -response testing formats, e.g., r(s) = .79, PAT versus the surgery su bject examination of the National Board of Medical Examiners. Moderate correlations with essay scores (r = .49) and preceptors' ratings (r = .49) were found. The residents' PAT performances were significantly h igher for senior than junior residents (82% versus 63%, p = .004). Con clusion. Utilizing this examination format, it is possible to locally develop a reliable examination that is grounded in the actual clinical problems seen by clerkship students and that assesses more than factu al recall.