The effects of the change in the NRMP matching algorithm. National Resident Matching Program

Area of Science:

• Medical Education
• Algorithm Design
• Healthcare Management

Background:

• The National Resident Matching Program (NRMP) faced controversy regarding its matching algorithm.
• The NRMP decided to implement a new applicant-proposing algorithm.

Purpose of the Study:

• To design and evaluate an applicant-proposing algorithm for the NRMP match.
• To compare the new algorithm with the existing one regarding match outcomes and strategic behavior.
• To provide guidance to participants based on algorithmic performance.

Main Methods:

• Computational experiments were conducted using data from past NRMP matches (1987, 1993-1996).
• The newly designed applicant-proposing algorithm was compared against the existing NRMP algorithm.
• Rank order lists (ROLs) from applicants and residency programs were analyzed.

Main Results:

• The two algorithms produced very similar matchings, with fewer than 0.1% of applicants receiving different matches.
• Applicants generally fared better with the applicant-proposing algorithm, while programs slightly preferred the existing algorithm.
• Opportunities for strategic behavior were rare and unlikely to yield benefits for either applicants or programs under both algorithms.

Conclusions:

• The existing and applicant-proposing NRMP algorithms demonstrate comparable performance.
• Both algorithms support ranking preferences honestly without strategic manipulation.
• The NRMP's algorithm choice impacts a very small subset of applicants, suggesting a nuanced decision based on overall system effects.