Title: Estimating Heterogeneous Causal Effect of Polysubstance Usage on Drug Overdose from Large-Scale Electronic Health Record

Abstract: Drug overdose has become a public health crisis in the United States with devastating consequences. However, most of the drug overdose incidences are the consequence of recitative polysubstance usage over a defined period of time which can be happened by either the intentional usage of required drug with other drugs or by accident. Thus, predicting the effects of polysubstance usage is extremely important for clinicians to decide which combination of drugs should be prescribed. Recent advancement of structural causal models can provide ample insights of causal effects from observational data via identifiable causal directed graphs. In this paper, we propose a system to estimate heterogeneous concurrent drug usage effects on overdose estimation, that consists of efficient co-variate selection, sub-group selection and heterogeneous causal effect estimation. We apply our framework to answer a critical question, can concurrent usage of benzodiazepines and opioids have heterogeneous causal effects on the opioid overdose epidemic? Using Truven MarketScan claim data collected from 2001 to 2013 have shown significant promise of our proposed framework's efficacy.