Analysis of the medical residency matching algorithm to validate and improve equity

by Briance Mascarenhas, Kartikeye Puranam

Algorithms are becoming prevalent but are often opaque and need external validation to assess whether or not they meet their purported objectives. The purpose of this study is to validate, using the limited information available, the algorithm used by the National Resident Matching Program (NRMP) whose intention is to match applicants to medical residencies based on applicants’ prioritized preferences. The methodology involved first using randomized computer-generated data to overcome the inaccessible proprietary data on applicant and program rankings. Simulations using these data were run through the compiled algorithm’s procedures to obtain match outcomes. The study’s findings are that the current algorithm’s matches are related to program input but not to applicant input, the applicant’s prioritized ranking of programs. A modified algorithm with student input as the primary factor is then developed and run using the same data, resulting in match outcomes that are related to both applicant and program inputs, improving equity.