PREVENT-HF Predicts HFpEF More Accurately Than HFrEF in Three Data Samples

Researchers evaluated the PREVENT-HF equations in adults aged 30–79 years without prevalent cardiovascular disease or heart failure. The analysis included five pooled longitudinal cohort studies (23,652 participants) and two primary-care electronic health record samples: Beth Israel Deaconess Medical Center (29,815) and Mass General Brigham (138,999). The primary assessment was prediction of 10-year incident HFpEF and HFrEF, including discrimination and calibration.

Discrimination was consistently higher for HFpEF than HFrEF. The AUROC was 0.824 versus 0.789 in the pooled cohorts, 0.897 versus 0.833 at BIDMC, and 0.864 versus 0.797 at MGB. Calibration was similar between subtypes; in the pooled cohorts, the integrated calibration index was 0.041 for HFpEF and 0.038 for HFrEF.

The findings suggest that a model developed for overall heart failure risk can distinguish risk of both subtypes in community and ambulatory-care samples, with stronger discrimination for HFpEF. However, PREVENT-HF was not developed to estimate subtype-specific risk, and the authors note that further work is needed to establish subtype-specific risk thresholds. The reported results assess prediction performance and do not establish that using the model improves clinical outcomes.