Abstract Background Heart failure (HF) with preserved ejection fraction (HFpEF) presents a high mortality risk due to its heterogeneous and multifactorial nature, coupled with limited treatment options. The incompletely understood pathophysiology is influenced by various comorbidities, resulting in different clinical phenotypes. Identifying these could be crucial for tailored HFpEF treatments. While previous studies focused on small, well-characterized patient populations, electronic health records (EHRs) offer the potential to unveil HFpEF phenotypes in larger real-world patient groups. Purpose We aimed to establish a sizable, unselected HFpEF patient population using real-world data. Employing cluster analysis, we intended to uncover distinct groups with varied disease aetiology and medical needs based on comorbidity profiles and prognoses. Methods From EHRs of n = 1,610,101 HF patients, we identified an HFpEF cohort of n = 32,009 patients, by applying thresholds for the LVEF and data availability, and by excluding confounding conditions. Clustering by the relative frequency of common comorbidities aimed to reveal patient populations with unique prognoses, comorbidity profiles, and clinical and demographic characteristics. Results Six clusters of different sizes, comorbidity prevalence, clinical presentation, and all-cause mortality were identified. About 50% of patients (n = 15,445) were predominantly hypertensive, with varying extents of coronary heart disease and diabetes mellitus (DM), and a 5-year mortality of about 30%. Around 22% (n = 7,089) exhibited a cardiometabolic comorbidity profile, with an increased prevalence of obesity and sleep apnea, but also DM and hypertension, and a 5-year mortality of 22.5%. Over 15% of patients (n = 5,298) had a 5-year mortality rate of 45.0%, marked by highest rates of anaemia and chronic kidney disease (CKD) but also DM, elevated N-terminal prohormone of brain natriuretic peptide levels, and a reduced estimated glomerular filtration rate. Conclusion Despite the favourable prognosis, a significant proportion of patients represent a cardiometabolic HFpEF phenotype with notable unmet medical needs regarding cohort size and multimorbidity. Patients simultaneously suffering from CKD, anaemia, and the highest mortality rates may represent a previously underrecognized phenotype. Our study demonstrates that distinct medical needs can be identified in a large, unselected real-world HFpEF cohort. In contrast to clinical trials, this approach enabled access to a broader patient population, revealing novel phenotypes to be targeted by precision medicine in the future.