Older adults (≥65 years old) constitute about 17% of the US population but account for ~57% of the population who use ≥3 drugs (polypharmacy). Currently, ~30million US adults use ≥3 drugs which increase their exposure to high-order drug-drug interactions (HDDIs), i.e. drug-drug interactions involving ≥3 drugs. HDDI is a major risk factor of serious adverse drug events (ADEs) that result in emergency department (ED) visits or hospitalization. Older adults, compared to younger adults, are 7-times more likely to be hospitalized and 2 to 3-times more likely to visit the ED for ADEs. However, knowledge on how to identify HDDI-ADE associations, data on the comparative safety of different 3-drug combinations and clinical and pharmacokinetic mechanisms for HDDI- ADE associations are all critically limited. The current project is designed to address these major gaps by specifically: (1) applying our state-of-art data mining techniques to discover HDDIs that are associated with higher risk of ADEs as well as those associated with lower ADE risk; (2) performing a comparative safety assessment of 3-drug combinations involving anticoagulants, antidiabetic agents and opioids; and (3) validating and evaluating the clinical validity and pharmacologic mechanisms of HDDI-ADE associations. These aims will be implemented by focusing on older patients who had an ED visit based on real-world data from health insurance claims (MarketScan and Medicare) and Electronic Health Records (EHR) data. We will focus on gastrointestinal (GI) bleeding, hypoglycemia and opioid-induced ADEs as the primary ADEs of interest for all three aims. These three ADEs account for 60% of all ADE-induced ED visits and are a target priority for prevention and surveillance by the US Health and Human Services. Our preliminary data from the MarketScan claims data (2012-2018) alone has confirmed that older adults are at higher risk of ADEs and that the risk of these ADEs increase with the counts of concomitant drugs used by patients. We will apply our mixture drug-count response (MDCR) and graphical models to identify the specific 3- or 4-drug combinations that have the highest risk and those with the lowest risk of ADEs in ED settings among older adults (Aim 1). For Aim 2, we will apply pharmacoepidemiologic study designs and casual inference approaches to assess the comparative safety of specific high and low-risk 3-drug combinations. We will control for the potential confounding effects of several patient-level (demographic, clinical, chronic comorbidities, drug characteristics, healthcare utilization, etc) and provider/healthcare system-level factors to assess the causal associations between high-risk versus low- risk 3-drug combinations. Finally, we will use a clinician expert team to evaluate the HDDI-ADE pairs and associations identified from Aims 1 and 2 for clinical validity and utility. The identification of high- and low-risk 3-drug combinations is vital knowledge that could guide the d...