Preliminary study: Data analytics for predicting medication adherence in Malaysian arthritis patients

Area of Science:

• Health Sciences
• Artificial Intelligence in Medicine
• Pharmacology

Background:

• Medication adherence is critical for arthritis management in diverse Malaysian populations.
• Non-adherence can lead to increased disability and disease progression.
• Understanding predictors of non-adherence is essential for improving patient outcomes.

Purpose of the Study:

• To analyze and predict medication adherence in Malaysian arthritis patients.
• To evaluate the performance of 13 machine learning models for predicting adherence.
• To identify key factors influencing medication adherence.

Main Methods:

• 151 arthritis patients in Malaysia participated in the study.
• Recursive feature elimination was used for feature selection in machine learning models.
• Key variables identified included occupation, comorbidities, religion, income, medication aids, marital status, and daily medication count.

Main Results:

• 90.07% of respondents reported non-adherence to their arthritis medication.
• Machine learning models, particularly Random Forest and Gradient Boosting with selected features, showed high predictive performance (AUC up to 0.883).
• Occupation was identified as the most significant factor influencing adherence, followed by unemployment, comorbidities, income, medication aid, marital status, and daily medication count.

Conclusions:

• Arthritis medication adherence is influenced by a complex interplay of socioeconomic and clinical factors.
• Personalized interventions targeting identified predictors are necessary to improve adherence rates.
• Machine learning offers a promising approach for predicting and understanding medication adherence challenges.