ABSTRACT Despite declines in total cardiovascular mortality rates in the United States, heart failure (HF) mortality rates, as well as hospitalizations and readmissions, are increasing with the greatest increases in mortality rates observed among non-Hispanic Black adults under the age of 65 years. Identification of individuals at risk of HF and specific HF subtypes (HFrEF and HFpEF) within diverse samples is critical to inform much-needed strategies to reduce the burden of HF. Although guideline-directed medical therapies are increasingly available for HF with reduced ejection fraction (HFrEF), prognosis remains dismal with 50% survival at 5 years. Further, few effective disease-modifying therapies currently exist for patients with HF with preserved ejection fraction (HFpEF), which is the most common HF subtype. The significant and growing burden of heart failure highlights the need for preventive interventions prior to the development of clinical symptoms. As a result, risk prediction to target prevention of HF, particularly for HFpEF, is a critical next step to improve outcomes. Whereas risk-based prevention (matching the intensity of prevention with the absolute risk of the individual) is widely accepted in the primary prevention of atherosclerotic cardiovascular disease, no such prevention paradigm currently exists for HF, in part, due to the lack of a well-established and generalizable risk model. To address multi-society practice guideline recommendations, our group recently developed and validated the Pooled Cohort Equations to Prevent Heart Failure (PCP-HF) using classic statistical modeling techniques in a population-based cohort sample. The current proposal builds upon our prior work and expands it to leverage novel machine learning methods to efficiently integrate large, multidimensional data across multiple domains and from two integrated health systems (Northwestern Medicine and Kaiser Permanente). This will allow us to create a geographically, racially/ethnically, and socioeconomically diverse real-world cohort of approximately 800,000 individuals to inform effective and equitable risk-based prevention strategies focused on HF. We will analyze individual-level data from the two health systems (e.g., clinical risk factor levels, comorbidities, medication use, social determinants of health) alongside innovative statistical techniques (e.g., machine learning) to develop optimal risk prediction models. The aims of the current proposal are: (1) develop and validate sex-specific risk prediction models for incident HF and HF subtype (HFrEF and HFpEF) and (2) define the comparative effectiveness of preventive HF therapies (e.g., angiotensin converting enzyme inhibitors, sodium glucose co-transporter 2 inhibitors) stratified by predicted HF risk. This project will lay the groundwork for future dissemination and implementation of clinical decision support tools to personalize HF prevention strategies. Completion of these aims will directly a...