ABSTRACT: Direct acting oral anticoagulants (DOAC), specifically rivaroxaban (RIVA) and apixaban (APX) have emerged as the most commonly used oral anticoagulants, replacing warfarin in patients with atrial fibrillation (AF) and venous thromboembolism (VTE). However, hemorrhage (HEM) is the most feared adverse effect and is a critical barrier to institution of these efficacious therapies. Anticoagulants have consistently ranked as the top offenders for adverse drug event related hospitalizations in the US, with HEM accounting for ~80% of adm is s ions . To predict HEM risk, investigators have developed clinical prediction rules (CPRs) focused mainly on warfarin- related HEM. However, the lack of CPRs for DOAC-related HEM, the variable inclusion of antiplatelets, the lack of assessment of gastro-protective therapies or interacting drugs (influencing APX and RIVA pharmacokinetics), the lack of assessment of risk across the spectrum of kidney function, and the scarce (to nil) representation of Blacks are critical knowledge gaps. Identifying factors contributing to the increased HEM risk among APX and RIVA users and elucidating their impact (effect size) is of critical importance In this application, we aim to identify predictors of HEM risk and define the impact (effect size) in a clinical cohort of medically complex, racially diverse patients on APX (n=1500) and RIVA (N=1500) and develop CPRs for APX-related-HEM and RIVA-related-HEM. Through Aim 1, we will identify and assess the influence of predictors (e.g. demographics, comorbidities) on risk of HEM. In Aim 2, we will elucidate the influence concomitant use of antiplatelet therapy, gastro-protective therapies (proton pump inhibitors and H2 Receptor Blockers) and interacting drugs (amiodarone, diltiazem and verapamil) on risk of HEM among RIVA and APX users. Aim 3 will focus on determining the influence of chronic kidney disease (CKD) spectrum on HEM risk and elucidate steady state pharmacokinetics (PK) and pharmacodynamics (PD) across the spectrum of kidney function among RIVA and APX users. Finally, Aim 4 will incorporate results from Aims 1-3 into building CPRs to personalize the prediction of HEM among APX and RIVA users. The prediction models will validated in an independent cohort (n=500 APX, n=500 RIVA). Our focus on complex patients seen in clinical practice with robust representation of understudied patients (Blacks, patients with moderate and severe-CKD) and assessment of common co-medications improves generalizability and applicability to real world patients.