MedSafe: AI-Driven Drug Side-Effect Detection and Alternative Recommendation
TL;DR
This work proposes an AI-driven system that consolidates disparate information sources, including patient demographics, medical history, and pharmacological properties of drugs, to enable early and accurate prediction of ADRs, and uses machine learning models to model complex relationships and generate predictive insights.
Abstract
: Adverse Drug Reactions (ADRs) are one of the main reasons for the increased rate of disease worldwide. Most of the conventional medication surveillance systems often identify ADRs only after their occurrences and hence cannot prevent patient harm. These approaches also continue to suffer from a variety of issues, including underreporting, biased reporting, and inability to record the complexity of drug-drug and drug-patient interactions. To overcome these limitations, this work proposes an AI-driven system that consolidates disparate information sources, including patient demographics, medical history, and pharmacological properties of drugs, to enable early and accurate prediction of ADRs. The proposed framework uses machine learning models to model complex relationships and generate predictive insights. Explainable AI (XAI) techniques such as SHAP are used to interpret model predictions. A dashboard displays the risk severity, assessment results, and suggested alternative drugs, offering useful and actionable insights. The system has been assessed using standard datasets such as FAERS, DrugBank, and SIDER, showing notable improvements in F1-score and ROC-AUC over their baselines. The integration of biomedical data with personalized patient information, this study, is supposed to revolutionize pharmacovigilance in the future by facilitating safer prescribing practices, improving patient safety, and encouraging the adoption of AI-powered decision support technologies in healthcare industry.