Refining Adverse Drug Reactions using Association Rule Mining for Electronic Healthcare Data

摘要

Side effects of prescribed medications are a common occurrence. Electronichealthcare databases present the opportunity to identify new side effectsefficiently but currently the methods are limited due to confounding (i.e. whenan association between two variables is identified due to them both beingassociated to a third variable). In this paper we propose a proof of concept method that learns commonassociations and uses this knowledge to automatically refine side effectsignals (i.e. exposure-outcome associations) by removing instances of theexposure-outcome associations that are caused by confounding. This leaves thesignal instances that are most likely to correspond to true side effectoccurrences. We then calculate a novel measure termed the confounding-adjustedrisk value, a more accurate absolute risk value of a patient experiencing theoutcome within 60 days of the exposure. Tentative results suggest that the method works. For the four signals (i.e.exposure-outcome associations) investigated we are able to correctly filter themajority of exposure-outcome instances that were unlikely to correspond to trueside effects. The method is likely to improve when tuning the association rulemining parameters for specific health outcomes. This paper shows that it may be possible to filter signals at a patient levelbased on association rules learned from considering patients' medicalhistories. However, additional work is required to develop a way to automatethe tuning of the method's parameters.