Abstract
Adverse drug reactions (ADRs) are undesirable and potentially fatal outcomes resulting from the use of medications. The possibility of experiencing an ADR varies between individuals owing to dis- ease heterogeneity, genetic and demographic fac- tors, patient treatment history and disease trajecto- ries. Therefore, understanding the interplay among the multiple factors leading to ADRs is crucial to increasing drug effectiveness, individualising drug therapy and reducing incurred cost.
In this paper, we present the first step towards predicting ADRs based on patient profiles and treatment trajectories hidden within the Electronic Health Records (EHRs). We propose a flexible encoding mechanism that can effectively capture the dynamics of multiple medication episodes of a patient at any given time. We enrich the en- coding with a drug ontology and patient demo- graphics data and use it as a base for an ADR prediction model. We evaluate the resulting pre- dictive approach under different settings using real anonymised patient data obtained from the EHR of the South London and Maudsley (SLaM), the largest mental health provider in Europe. Using the profiles of 38,000 mental health patients, we iden- tified 240,000 affirmative mentions of dry mouth, constipation and enuresis and 44,000 negative ones. Our approach achieved 93% prediction accuracy and 93% F-Measure. Overall, we found that us- ing our encoding can improve prediction accuracy by 10% compared to static medication modelling settings.
In this paper, we present the first step towards predicting ADRs based on patient profiles and treatment trajectories hidden within the Electronic Health Records (EHRs). We propose a flexible encoding mechanism that can effectively capture the dynamics of multiple medication episodes of a patient at any given time. We enrich the en- coding with a drug ontology and patient demo- graphics data and use it as a base for an ADR prediction model. We evaluate the resulting pre- dictive approach under different settings using real anonymised patient data obtained from the EHR of the South London and Maudsley (SLaM), the largest mental health provider in Europe. Using the profiles of 38,000 mental health patients, we iden- tified 240,000 affirmative mentions of dry mouth, constipation and enuresis and 44,000 negative ones. Our approach achieved 93% prediction accuracy and 93% F-Measure. Overall, we found that us- ing our encoding can improve prediction accuracy by 10% compared to static medication modelling settings.
Original language | English |
---|---|
Title of host publication | IJCAI 2016 - Workshop on Knowledge Discovery in Healthcare Data |
Publication status | Published - Jul 2016 |