The paper presents a review of current research to develop predictive models for automated detection of drug-induced repolarization disorders and shows a feasibility study for developing machine learning tools trained on massive multimodal datasets of narrative, textual and electrocardiographic records. The goal is to reduce drug-induced long QT and associated complications (Torsades-de-Pointes, sudden cardiac death), by identifying prescription patterns with pro-arrhythmic propensity using a validated electronic application for the detection of adverse drug events with data mining and natural language processing; and to compute individual-based predictive scores in order to further identify clinical conditions, concomitant diseases, or other variables that correlate with higher risk of pro-arrhythmic situations.
Keywords: Adverse Drug Events; Analytic-Decision Modelling; Clinical Decision Support System; Electrocardiography; Long QT; Machine Learning; Pharmacovigilance; Repolarization Disorders; Torsades-de-Pointes.