Introduction of bias in residency-candidate interviews

We have developed, as part of our resident applicant ranking process, a computerized weighted scoring system incorporating, among other variables, the interview score (IS). This study was undertaken to test the IS as a dependent variable when the applicant file is available during the interview. Over a 3-year period, each of two faculty members interviewed 133 candidates in a "blinded" fashion (file unavailable) or 290 candidates with an "open" file. The mean discrepancy between paired interviewers (1 to 10 scale) was 1.34 +/- 0.10 for the blinded group versus 0.93 +/- 0.06 for the open group (p less than 0.001). Individual and multiple regression analysis were used to test the IS as a dependent variable in the open group, with the blinded group as a control. In each instance (except clerkship grades), the IS correlated more closely with other parameters (p less than 0.05) in the open group. Correlation coefficients were 0.27 (blinded) and 0.64 (open) for weighted scores combining all parameters (p less than 0.001) and 0.39 (blinded) and 0.65 (open) for multiple linear regression analysis (p less than 0.001). We conclude that the IS is significantly influenced if other objective variables are at hand, thus deflating the actual weight of the interview. To accurately quantify the value of the interview, a correction factor must be applied, or, more appropriately, the interview must be conducted blindly.