Using machine learning to identify key subject categories predicting the pre-clerkship and clerkship performance:...

Area of Science:

• Medical Education Research
• Machine Learning in Healthcare
• Academic Performance Prediction

Background:

• Medical students require a strong knowledge base for physician development.
• Clerkship performance is a critical developmental milestone.
• Identifying predictors of clerkship success can optimize medical education.

Purpose of the Study:

• To identify academic subjects that predict medical student clerkship performance.
• To develop machine learning models for performance prediction.

Main Methods:

• A cohort study analyzed data from medical students graduating between 2011 and 2019.
• Machine learning techniques, including random forest, were employed.
• Models were trained and validated using 10-fold cross-validation.

Main Results:

• Thirteen pre-med and 10 basic science subjects showed predictive ability (AUC > 0.7).
• Medical humanities, sociology, chemistry, physician scientist training, pharmacology, immunology-microbiology, and histology were significant predictors.
• The random forest model achieved 95% accuracy and 88% AUC for predicting clerkship performance.

Conclusions:

• Pre-med and basic science subjects can predict clerkship performance.
• Understanding subject-performance relationships enhances student preparedness for clerkships.