A comprehensive review on Adverse Drug Reactions (ADRs) Detection and Prediction Models

Abstract

In medical domain ADRs are defined as unintended harmful reactions of drugs. Several incidences of ADR reports related to a medicinal product can lead to an intervention by higher medical authorities. It can result in label change or complete ban from consumer market. The main aim of this review paper is to elaborate different techniques and methodologies implemented for several ADR datasources using research works related to ADR detection and prediction domain. The relevant research works are collected from known sites like Pubmed & ResearchGate. The papers are selected on the basis of some research questions that are ‘Identify the different datasets used for ADR detection & prediction?’ ‘Why early detection of ADRs are important for better patient safety and healthcare? ’ and ‘How recent trends in artificial intelligence and machine learning domain are useful in accurate prediction of ADRs? On the basis of the research questions a total 172 research papers are collected. After analyzing thoroughly the authors had identified 87 research studies of actual interest that can be categorized into 51 research papers related to ADR detection theme and 36 research works are related to ADR prediction theme. Furthermore the authors present a gap analysis and based on it a novel deep learning framework have been designed. Through this review study the authors have successfully highlighted the fact that early detection and prediction of ADR is crucial for better patient safety and healthcare.