An interpretable boosting model to predict side effects of analgesics for osteoarthritis

Area of Science:

• Medical informatics
• Machine learning in healthcare
• Pharmacovigilance

Background:

• Osteoarthritis (OA) is a prevalent degenerative joint disease.
• Analgesic use in OA patients is linked to a 20-50% increased risk of cardiovascular diseases.
• Limited research exists on predicting analgesic side effects in OA, with a lack of interpretable models.

Purpose of the Study:

• To develop an accurate and interpretable prediction model for analgesic side effects in OA patients.
• To identify high-risk features associated with cardiovascular diseases from analgesic use.
• To provide clinical decision support for OA pain management.

Main Methods:

• Utilized the eXtreme Gradient Boosting (XGBoost) supervised machine learning algorithm.
• Trained the model using Electronic Medical Records (EMRs) from public knee OA studies.
• Evaluated model performance against four established machine learning algorithms.

Main Results:

• The XGBoost model demonstrated superior performance compared to other algorithms.
• Successfully predicted side effects of analgesics in OA patients.
• Identified key risk features for cardiovascular diseases linked to analgesic use.
• Provided interpretable rules for risk prediction.

Conclusions:

• The XGBoost model offers a valuable tool for predicting analgesic side effects in OA patients.
• The model's interpretability enhances its clinical utility for researchers and patients.
• This approach supports safer analgesic selection and management in osteoarthritis treatment.