Adaptive Machine Learning Framework for Real-Time Medication Adherence Prediction in Chronic Disease Management

Medication adherence is a critical factor in effective chronic disease management, yet a significant proportion of patients fail to follow prescribed treatment regimens, leading to poor health outcomes and increased healthcare costs. Existing adherence monitoring systems are largely based on static models and retrospective analysis, limiting their ability to provide timely and personalized interventions. To address these challenges, this study proposes an adaptive machine learning framework for real-time medication adherence prediction. The approach integrates temporal feature modeling, a sliding window mechanism, and an adaptive learning strategy to continuously update model parameters based on evolving patient behavior. A proactive risk prediction module is incorporated to estimate the likelihood of future non-adherence events, enabling early intervention. The framework is evaluated using a publicly available dataset transformed into time-series format to simulate real-world conditions. Experimental results demonstrate that the proposed model outperforms baseline methods, achieving an accuracy of 0.91, F1-score of 0.88, and ROC-AUC of 0.93, along with improved early prediction capability. These findings highlight the effectiveness of adaptive and real-time learning approaches in enhancing personalized healthcare systems and improving medication adherence outcomes.