Ensemble Model Predicts Pharmacogenetic Testing Uptake
Researchers at the University of Florida develop and validate machine learning models to predict patients' likelihood of undergoing pharmacogenetic testing when prescribed opioids, using EHR data from 48,528 patients and multiple classifiers. An ensemble combining XGBoost and SVM achieved a 79.61% C-statistic and 67.4% accuracy, with age, hypertension, and household income as top predictors, suggesting a decision-support role to guide testing outreach and resource allocation.
Key Points
- 1Develops ensemble ML model using XGB and SVM achieving 79.61% C-statistic on 48,528 patients
- 2Identifies age, hypertension, and household income as top predictors via SHAP, highlighting demographic and clinical influences
- 3Enables clinicians and health systems to prioritize outreach, allocate resources, and guide pharmacogenetic testing decisions
Scoring Rationale
Solid ML performance and clinical applicability demonstrate practical utility, but findings are limited by single-center data and uncertain external generalizability.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
