Assessment of Machine Learning Algorithms to Predict Medical Specialty Choice
David Vicente Alvarez1, Milena Abbiati1,2, Alban Bornet1
1University of Geneva.
Studies in Health Technology and Informatics
|May 17, 2025
Summary
Machine learning models can predict medical students' specialty choices. Key factors include surgical interest and personality traits, aiding future physician workforce planning.
Area of Science:
- Medical Education
- Public Health
- Computational Medicine
Background:
- Equitable physician distribution across specialties is a critical public health issue.
- Traditional statistical models have limitations in predicting medical students' career paths.
- Understanding factors influencing specialty choice is vital for effective workforce planning.
Purpose of the Study:
- To explore machine learning (ML) techniques for predicting medical students' specialty decisions early in their studies.
- To identify key factors influencing career choices using ML interpretability methods.
- To assess the potential of ML in improving physician workforce planning.
Main Methods:
- Evaluated supervised ML models: Support Vector Machines, Artificial Neural Networks, XGBoost, and CatBoost.
- Utilized data from 399 medical students in Switzerland and France.
- Employed post-hoc interpretability techniques to identify influential factors.
Main Results:
- Ensemble ML methods, particularly CatBoost, outperformed simpler models.
- CatBoost achieved a macro Area Under the Receiver Operating Characteristic Curve (AUROC) of 76%.
- Significant predictors included motivation for surgery and psychological traits like extraversion.
Conclusions:
- Machine learning offers a promising approach for predicting medical career paths.
- ML interpretability can reveal crucial factors influencing specialty selection.
- These insights can inform targeted interventions for better physician distribution and workforce planning.
