Predicting a successful match among applicants to reproductive endocrinology and infertility fellowship
Leigh A Humphries,Jeremy Applebaum,Monica A Mainigi,Caitlin E Martin,Divya K Shah
DOI: https://doi.org/10.1016/j.fertnstert.2024.08.337
2024-08-24
Abstract:Objective: To identify independent predictors of a successful match to reproductive endocrinology and infertility (REI) fellowships, and to develop and internally validate a prediction model for REI match results. Design: Retrospective cohort study. Setting: University-based institution. Participants: Reproductive endocrinology and infertility fellowship applications sent to the University of Pennsylvania from 2019 to 2023 (excluding 2020), which represented nearly all REI applicants nationally according to National Resident Matching Program data. Intervention(s): Demographics, education, training, and academic achievements. Main outcome measure(s): Match result, confirmed through online search and communication with program administrators. Univariate analyses identified variables associated with match, which were then included in multivariable models to identify independent predictors. Bootstrapping was used to assess model discrimination and calibration. The final model was integrated into a web-based tool. Result(s): Of 286 applications (99.0% of REI applications to the National Resident Matching Program), 199 (69.6%) resulted in a successful match. In univariate analyses, variables associated with match were younger age, attendance at an allopathic US medical school, United States Medical Licensing Examination (USMLE) and Council on Resident Education in Obstetrics and Gynecology scores, residency rank, residency affiliation with a fellowship, research experiences, first-author publications, abstracts/articles in progress, and poster presentations. In the adjusted model, independent predictors of match included residency affiliation with an REI fellowship (adjusted odds ratio [aOR], 5.43; 2.02-14.64), residency rank (aOR, 1.77; 1.25-2.50), USMLE score (aOR, 1.05; 1.02-1.08), at least one first-author publication (aOR, 2.32; 1.08-4.96), projects in progress (aOR, 1.26; 1.02-1.55), and poster presentations (aOR, 1.07; 1.00-1.15). Attendance at an international medical school was a negative predictor (aOR, 0.32; 0.11-0.88). The model achieved an area under the curve of 0.883, with 88.5% sensitivity and 65.8% specificity. A refined model without USMLE scores maintained strong performance (C-statistic, 0.85; 0.81-0.91; calibration slope, 0.91; 0.72-1.24). Conclusion(s): Affiliation with an REI fellowship, residency reputation, and research output strongly predicted match success. Gender, race, and ethnicity were not major predictors, yet underrepresentation of certain racial and ethnic groups limited the power to detect potential differences. Our prediction model correctly classified >75% of candidates' match results. These findings may help candidates optimize applications and estimate chances of a successful match into REI fellowship, as well as assist programs in critically reviewing their selection criteria for fellowship match.