Medicine Overdose Prediction

Abstract

Medication overdose continues to be a major global health crisis, leading to avoidable emergencies, bad drug reactions, and more hospital visits. The rising complexity of modern prescriptions, especially when patients take multiple drugs, makes it easier to misread doses, miss dangerous interactions, and cause accidental harm. This research introduces a smart automated tool designed to predict overdose risks by analysing prescriptions. Our framework combines Optical Character Recognition (OCR) with machine learning to improve prescription reading accuracy and keep patients safer. The process starts by using OCR to pull text from images of handwritten or printed prescriptions. Key details such as drug names, strengths, timing, patient age, and history are organised in a digital format. Then, machine learning models such as Random Forests, Decision Trees, and Neural Networks analyse the data to identify risks and calculate safety scores. Tests show that the OCR reaches 92% accuracy in reading text, while the prediction model hits 89% accuracy, with 87% precision and 91% recall for high-risk flags. This method greatly reduces human reading errors and makes dosage checks much clearer. Future goals include linking the tool to Electronic Health Records (EHRs), adding support for multiple languages, expanding the drug interaction database, and developing a mobile app for instant alerts. This system shows great promise as a helpful clinical support tool, promoting safer medication habits and lowering the number of overdose cases worldwide.