Using standardized patients to assess professionalism: a generalizability study.

Background: Assessment of professionalism in undergraduate medical education is challenging. One approach that has not been well studied in this context is performance-based examinations. Purpose: This study sought to investigate the reliability of standardized patients' scores of students' professionalism in performance-based examinations. Methods: Twenty students were observed on 4 simulated cases involving professional challenges; 9 raters evaluated each encounter on 21 professionalism items. Correlational and multivariate generalizability (G) analyses were conducted. Results: G coefficients were .75, .53, and .68 for physicians, standardized patients (SPs), and lay raters, respectively. Composite G coefficient for all raters reached acceptable level of .86. Results indicated SP raters were more variable than other rater types in severity with which they rated students, although rank ordering of students was consistent among SPs. Conclusions: SPs' ratings were less reliable and consistent than physician or lay ratings, although the SPs rank ordered students more consistently than the other rater types.