Design: Validation study in five pooled cohorts and two electronic health record samples
Population: About 192,000 US adults aged 30–79 without cardiovascular disease
Primary outcome: Discrimination and calibration for 10-year incident HFpEF and HFrEF
Effect: AUROC HFpEF 0.82–0.90 vs HFrEF 0.79–0.83; calibration similar

The PREVENT-HF equations estimate 10-year risk of heart failure overall. A Circulation analysis tested how well they predict each subtype in five pooled community cohorts (23,652 adults) and two US primary-care record samples (29,815 and 138,999 adults), all aged 30–79 without cardiovascular disease.

Discrimination was good for both subtypes but consistently better for HFpEF: area under the curve 0.82 versus 0.79 in the pooled cohorts, 0.90 versus 0.83 and 0.86 versus 0.80 in the two health-system samples. Calibration was similar for both.

This matters because the two subtypes are prevented and treated differently. The equations were not built for subtype prediction, and no subtype-specific thresholds exist yet, so the practical use for now is the overall score — with some reassurance that a high score captures future HFpEF well.

Use PREVENT-HF for overall 10-year heart failure risk in adults without cardiovascular disease
A high score captures future HFpEF at least as well as HFrEF
Treat obesity, hypertension and diabetes hard in high scorers — the main HFpEF drivers
No subtype-specific risk thresholds exist yet, so do not split the score into HFpEF and HFrEF risk

Why it matters

HFpEF has the fewest treatments, so predicting it early is where prevention counts most.

The statistics, in plain English

An area under the curve of 0.82 means that, for a random pair of one person who developed HFpEF and one who did not, the score ranks them correctly about 82% of the time. Good calibration means predicted and observed risks match.