Related Experiment Video
Updated: May 5, 2026
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
What Are the Odds? Utilization of Artificial Intelligence to Predict Success in the Integrated Plastic Surgery Match
Kerilyn N Godbe1, Anne M Meyer1, Cole L Bird2
1Department of Plastic Surgery, University of Kansas Medical Center, Kansas City, Kansas.
Objective:
Despite widespread consensus that quantifiable application characteristics impact match success, no prospective predictive algorithm has been created in any specialty since the implementation of USMLE Step 1 pass/fail scoring. The rise of artificial intelligence presents a novel method to analyze application factors to predict success in the ever-competitive integrated plastic surgery match.
Design:
Quantifiable data from Plastic Surgery Common Applications (PSCAs) submitted to a single academic center in the 2023 to 2024 and 2024 to 2025 match cycles were collected. About 70% of applications from the 2023 to 2024 cycle were utilized to train a random-forest classification model to predict match status. This model was subsequently evaluated using the remaining 30% of 2023 to 2024 applicants and prospectively validated with applicants from the 2024 to 2025 match cycle.
Setting:
University of Kansas Medical Center, tertiary center.
Participants:
About 713 integrated plastic surgery applicants in the 2023 to 2024 and 2024 to 2025 match cycles.
Results:
The single most important predictor of match rate in our model was graduation from a highly ranked medical school, followed by average recommendation letter strength and USMLE Step 2 score. The predictive model demonstrated strong performance in classifying matched and unmatched cases in the 2023 to 2024 data with an AUROC of 0.85 and a balanced accuracy of 82.7% (95% CI 73.7%-89.6%, p < 0.0001). Prospective validation yielded consistent results, with an AUROC of 0.84 and a balanced accuracy of 77.6% (95% CI 73.1-81.7, p < 0.0001), validating the longitudinal accuracy of this model. The final model was deployed within our newly designed Match Prediction App, allowing real-time probability estimation of match outcomes based on selected input features.
Conclusions:
This model provides a comprehensive analysis of integrated plastic surgery residency match predictors, aiding both applicants and program directors in understanding key factors that influence match outcomes. While academic metrics like Step 2 scores are critical, the significance of letters of recommendation highlights that holistic review remains essential for residency selection.
