Machine Learning Prediction of Pharmacogenetic Testing Uptake Among Opioid-Prescribed Patients Using Electronic...

Area of Science:

• Pharmacogenomics
• Machine Learning
• Health Informatics

Background:

• Opioid therapy for pain management shows variable efficacy and adverse effects.
• Pharmacogenetic testing offers personalized opioid therapy but faces low uptake.
• Barriers to pharmacogenetic testing include patient, provider, infrastructure, and financial factors.

Purpose of the Study:

• Develop machine learning (ML) models to predict pharmacogenetic testing uptake in opioid users.
• Identify key demographic, clinical, and social determinants influencing testing likelihood.
• Improve resource allocation and patient outcomes in pain management.

Main Methods:

• Utilized electronic health record data from a single healthcare system.
• Extracted patient demographics, clinical variables, medication use, and social determinants of health.
• Developed and validated multiple ML models, including ensemble methods, for uptake prediction.

Main Results:

• An ensemble model combining XGBoost and SVM achieved the highest C-statistic (79.61%).
• The ensemble model demonstrated high accuracy (67.38%) and recall (76.50%).
• Age, hypertension, and household income were identified as key predictors of pharmacogenetic testing uptake.

Conclusions:

• The developed ensemble model effectively predicts pharmacogenetic testing uptake in opioid-prescribed patients.
• This ML model can serve as a decision support tool for clinicians.
• Optimizing pharmacogenetic testing can enhance personalized pain management strategies.