Confirmatory factor analysis of house office candidate appraisals

This article illustrates an application of confirmatory factor analysis, using the LISREL program, to retrospective analysis of rating data common in health professions education. We begin with specification of a causal model reflecting the multitrait-multimethod design of a surgical house-officer ranking system. The model is then tested using data provided by rating system functioning over two successive years. The results suggest that although most rating measures do cluster around latent traits as hypothesized, there is more method variance in the data set than trait variance. The trait and method factors are also found to be highly intercorrelated. Practical implications of these results include specific modifications in the rating system, potentially leading to improvement of the system.