Linear Modeling to Reduce Bias in Plastic Surgery Residency Selection.

Abstract

Consistently selecting successful, productive applicants from an annual candidate pool is the goal of all resident selection practices. Efforts to routinely identify high-quality applicants involve scrutiny of multiple factors and formulation of an ordinal rank list. Linear modeling offers a quantified approach to applicant selection that is strongly supported by decades of psychological research. For the 2019 residency application process, the University of Wisconsin Plastic Surgery Residency Program used linear modeling in their evaluation and ranking process. A linear model was developed using United States Medical Licensing Examination Step 1 and Step 2 scores, letters of recommendation, publications, and extracurricular activities as inputs. The applicant's total score was calculated from a maximum total score of 100. The mean and median scores were 49 and 48, respectively, and applicants were ranked according to total score. A separate rank list was maintained using our program's standard methodology for applicant ranking, which involves global intuitive scoring during the interview process. The Spearman rank correlation coefficient between the two lists was 0.532, and differences between the rank lists were used as a fulcrum for discussion before making the final rank list. This article presents the first known instance of the use of linear modeling to improve consistency, increase fairness, and decrease bias in the plastic surgery residency selection process. Transparent sharing of methodology may be useful to other programs seeking to optimize their own ranking methodology. Furthermore, it indicates to applicants that they are being evaluated based on fair, quantifiable criteria.