ABSTRACT The prevalence of and mortality from cirrhosis have increased dramatically over the past twenty years. Atrial fibrillation (NVAF) disproportionately affects 15% of patients with cirrhosis and leads to substantial morbidity and mortality, including 5-fold higher rates of stroke. In the general population with NVAF, the risk-benefit balance typically favors initiation of oral anticoagulants (OAC). In contrast, over 50% of patients with cirrhosis and NVAF are not anticoagulated despite their markedly increased risk of stroke, due to concerns about bleeding, poor anticoagulation quality and falls. These concerns also complicate selection of an optimal OAC regimen, which involves choosing between warfarin and direct oral anticoagulants (DOAC: apixaban, rivaroxaban, dabigatran and edoxaban). While DOACs have potential benefits over warfarin in patients with cirrhosis, they have variable hepatic metabolism (between 20%-75%) and they are largely untested in cirrhosis, because such patients were excluded from all prior DOAC randomized controlled trials (RCTs) for NVAF. Moreover, RCTs are unlikely to recruit and retain vulnerable patients with advanced, decompensated cirrhosis or cognitive impairments in a manner representative of routine care. Thus, for patients with cirrhosis and NVAF – who are at higher risk of both bleeding and thrombosis – there is a pressing need for robust data regarding the optimal OAC strategy. This proposal seeks to define the comparative effectiveness and safety of warfarin and DOACs in a population- based cohort of over 250,000 U.S. adults with established cirrhosis and NVAF diagnosed between 2011-2019 (including over 100,000 new OAC initiators), from 4 healthcare databases (Medicare, Medicaid, Truven MarketScan and Optum Clinformatics). Within this cohort, we propose: (Aim 1) To define the effectiveness and safety of use vs. non-use of OACs (including warfarin and specific DOACs); (Aim 2) To define the comparative effectiveness and safety of initiating (2a) warfarin vs. DOACs; and (2b) specific DOACs (i.e. head-to-head comparisons of, apixaban vs. rivaroxaban vs. dabigatran vs. edoxaban); and (Aim 3) To identify key subgroups for whom OACs are particularly beneficial or hazardous – including patients with advanced or decompensated cirrhosis, high fall risk or poor predicted anticoagulation quality. To minimize confounding and optimize study validity, we will use innovative methods developed by our team, including: (i) high-dimensional propensity scores; (ii) our novel, validated algorithms for phenotyping cirrhosis severity, fall risk and anticoagulation quality; and (iii) linkage of a subset of our cohort with rich clinical information and laboratory data from electronic health records (EHR), for external adjustment and propensity score calibration. Completing these studies will provide the necessary evidence base for providers to optimize OAC selection and stroke prevention in vulnerable patients with cirrhosis and ...