Effects of rater selection on peer assessment among medical students.

Context: Although peer-assessment appears promising as a method to assess interpersonal skills among medical students, results may be biased by method of peer selection, particularly if different kinds of classmates are assigned systematically by different methods. It is also unclear whether students with lower interpersonal skills may be more negative towards their classmates than students with higher levels of interpersonal skills and, if so, how much bias this may introduce into the results of peer assessment. It is also unclear whether low-rated students are more likely to ask to rate one another.

Design, Setting, And Participants: Participants were 296 medical students from the classes of 2004, 2005 and 2006 at the University of Rochester School of Medicine and Dentistry. All participants were evaluated by between 6 and 12 classmates near the end of both their second and third years. In some classes peer-raters were assigned from class lists, while in other classes students were asked to recommend classmates to rate and be rated by.

Main Outcome Measures: Mean numerical ratings of interpersonal-attributes scores received from peers, compared with mean scores simultaneously assigned to peers. We also examined students' selection of peer-raters, as well as scores received by their peer-raters.

Results: For all 3 classes in both second and third years, students in the lowest quartile of received scores were significantly more likely to simultaneously assign lower scores to their peers. Among third-year students, such low-scoring students were significantly more likely to ask to rate and be rated by other low-scoring students. Among third-year students, low-rated students were more likely to rate and be rated by other low-rated students, irrespective of how rater assignments are made. Most students who were originally in the lowest quartile of received scores remained so after removing the scores assigned by low-rated raters. Among pre-clinical second-year students, however, low-rated students were no more likely than high-rated students to be rated by other low-rated students.

Conclusion: Students with low levels of peer-assessed interpersonal attributes are more negative in their judgements of classmates. Third-year medical students with low levels of interpersonal attributes are more likely to be rated by other students with low levels of interpersonal attributes, irrespective of method of rater assignment. This effect was not seen in the second year of medical school. Biases in selection of raters among third-year students, however, do not appear to affect the results of peer assessment. Thus, medical schools that choose to institute peer assessment can select the method of peer-rater assignment based on practicality and logistical considerations. Students' affiliations with one another in the third year appear to be related to peer-assessed interpersonal attributes.