The 2019 U.S. practice guidelines recommend direct oral anticoagulants (DOACs) such as dabigatran, rivarox- aban, and apixaban over warfarin for stroke prevention in high-risk patients with atrial fibrillation (AF). However, the selection of a DOAC can be challenging for older ischemic stroke patients. The risk-benefit ratios in second- ary stroke prevention differ substantially from those in primary prevention. Not only are older stroke survivors at increased risk for recurrent ischemic events, older age and history of ischemic stroke are also major risk factors for bleeding complications. While it is known that drug selection should be individualized, it remains unknown how to tailor anticoagulant therapy according to the effectiveness/safety of the agents and patient unique char- acteristics. The long-term goal is to use comparative effectiveness research to improve quality of care and out- comes in patients with cardiovascular disease and stroke. Leveraging the American Heart Association (AHA) Get With The Guidelines-Stroke Registry (GWTG-Stroke) and Medicare inpatient and Part D database, the over- all objective of this application is to develop evidence-based strategies to improve appropriate anticoagulant therapy for secondary prevention in older ischemic stroke patients with AF. The central hypothesis is that differ- ences exist between DOACs in terms of effectiveness and safety, which have direct implications for therapeutic selection. Once the relative effectiveness/safety is known, the selection of an anticoagulant can be made based on patient risk profiles, making treatment safer and more effective. Guided by strong preliminary data, this hy- pothesis will be tested by pursuing two specific aims: 1) Determine the long-term clinical effectiveness and safety of different DOACs for secondary prevention in older ischemic stroke patients with AF; 2) Investigate DOACs dosing patterns and evaluate the potential impact of underdosing or overdosing on long-term outcomes. The proposed research is innovative in four key ways: 1) A patient-centered approach is used to address a real-life decisional dilemma facing stroke survivors and clinicians; 2) It shifts focus from selected samples in clinical trials to a nationwide representative stroke population, including traditionally underrepresented subgroups in commu- nity practice; 3) A novel propensity score inverse probability weighting method using generalized boosted models (a machine learning technique) will be employed to mimic a trial-type multi-treatment design, uncover treatment heterogeneity, and minimize selection bias in observational data; 4) Beyond traditional mechanisms of scientific publications, the evidence generated from this study will be disseminated to stroke survivors, clinicians, and relevant stakeholders through the AHA GWTG-led national quality initiatives and patient-led efforts to ensure a rapid translation of seminal findings into clinical practice. The proposed research ...