How accurate are the AHA PREVENT equations at predicting heart disease?
Answered November 2, 202610 sources
EvidenceBModerateGood evidence in people with one real weakness. New studies could change the answer.
The PREVENT equations rank people by heart disease risk well. They can still under- or over-estimate actual risk, and in some US health systems they predicted far fewer events than occurred.
- Who it's for
- US adults aged 30 to 79 without known heart disease.
- What to expect
- good ranking of who is higher or lower risk; the exact percentage can be off.
- Watch out for
- risk estimates can run too low in some health systems.
Each dot is the C statistic (1 = perfect ranking of who gets disease) with a line for its 95% range.
- Asian patients0.83 (0.82 to 0.84)Clear improvement
- Hispanic patients0.8 (0.79 to 0.81)Clear improvement
- US adults, CVD deaths0.89 (0.88 to 0.9)Clear improvement
- Southern California adults0.74 (0.74 to 0.75)Clear improvement
| Result | Estimate | 95% range | Reading |
|---|---|---|---|
| Asian patients | 0.83 | 0.82 to 0.84 | Clear improvement |
| Hispanic patients | 0.8 | 0.79 to 0.81 | Clear improvement |
| US adults, CVD deaths | 0.89 | 0.88 to 0.9 | Clear improvement |
| Southern California adults | 0.74 | 0.74 to 0.75 | Clear improvement |
Sources 4 JAMA Cardiol, 2025; 5 JAMA Netw Open, 2024; 6 J Am Heart Assoc, 2025
Figure 1. Kaplan-Meier Survival Curve Stratified by Predicting Risk of Cardiovascular Disease Events (PREVENT) Risk Score Category The risk score categories are defined as low (<5.0%),…
Figure 1 from External Validation of the American Heart Association PREVENT Cardiovascular…, JAMA Netw Open 2024. Licence: CC BY 4.0
Figure 2. Receiver-Operator Characteristic Curves Demonstrating Sensitivity and Specificity of Predicting Risk of Cardiovascular Disease Events Score for Cardiovascular (CV) and Competing…
Figure 2 from External Validation of the American Heart Association PREVENT Cardiovascular…, JAMA Netw Open 2024. Licence: CC BY 4.0
What guidelines say
The American Heart Association describes PREVENT as race-free equations giving 10- and 30-year risk of total heart disease for adults 30 to 792. The 2025 and 2026 American Heart Association and American College of Cardiology guidelines recommend PREVENT to guide starting and intensifying blood pressure and cholesterol treatment8.
What the studies show
- In the original development work, the equations were tested in 3.3 million adults from 21 data sets. Ranking was good: scores of 0.794 in women and 0.757 in men (1.0 is perfect, 0.5 is a coin flip). Predicted risk closely matched observed risk3.
- A study of 270,320 patients in four US health systems found only moderate ranking (0.69 to 0.74). PREVENT underestimated 10-year risk of heart attack and stroke by 71% in one system, 40% in another and 36% in a third. It was close in the fourth (1.3% too high)9. Underestimation was larger in women in two systems9. These results conflict with the more favorable validation studies, and they come from academic medical centers.
- A study of 559,241 Kaiser Permanente members followed for a median of 10 years found PREVENT well calibrated. The older Pooled Cohort Equations overestimated risk, with predicted events about 1.8 to 2.2 times those observed. Ranking was about the same for both (0.741)6.
- A national US survey analysis (172.9 million people represented, deaths from heart disease over 10 years) found excellent ranking (0.89). Risk was slightly underfit, meaning predictions were a little too spread out5.
- In 361,778 patients in one health system, ranking was good in Asian (0.83) and Hispanic (0.80) patients. Predicted risk ran too high in several Asian subgroups A slice of the people in a study, such as only the women or only those over 65. Results in one slice are less reliable than the main result., except Asian Indian patients4.
- In 314,783 people, observed event rates generally fell within predicted risk categories. People with high lipoprotein(a), a blood fat not in the equations, had about 30% more events than others. Adding it improved prediction only modestly1.
Key studies
| Study | Who | What it found |
|---|---|---|
| Development and validation, 20243 | 6.6 million US adults | Good ranking and close calibration in outside data |
| Health-system study, 20259 | 270,320 patients, four systems | Moderate ranking; risk underestimated in three of four |
| Study following people, 20256 | 559,241 Kaiser members | Well calibrated; older equations overestimated risk |
| Study following people, 20245 | US national survey sample | Excellent ranking; slight miscalibration |
| Cohort study Researchers follow a group of people over time without giving them a treatment. It can show links, not cause and effect., 20254 | 361,778 primary care patients | Good ranking across Asian and Hispanic subgroups |
What we don't know
- Why accuracy differs so much between health systems, and how to correct it locally.
- Whether treating people based on PREVENT scores leads to better health outcomes than older tools.
Try it yourself: 10-year heart risk (PREVENT)
Ask a follow-up
Written by AI from the studies listed, without review by a clinician, and graded with the GRADE method. It can be wrong, so check the studies it cites. Education, not medical advice.