A new study warns that the American Heart Association’s (AHA’s) PREVENT calculator may underestimate risk for cardiovascular disease in certain populations, potentially leading to missed opportunities for prevention.

The study, published in the Journal of the American College of Cardiology, evaluated the performance and algorithmic fairness of base and social deprivation index-augmented PREVENT equations in adults aged 30-39 years.

It found that the PREVENT calculator substantially underpredicted this risk in young adults — particularly among non-Hispanic Black individuals, for whom event rates were nearly double.

When the AHA introduced the PREVENT cardiovascular risk calculator in late 2023, it removed race as a separate predictive factor. This decision was deliberate, and it has become part of an ongoing conversation in the medical community. For example, an article published on the Harvard Medical School website in July 2024 outlined concerns about the downstream effects of the new calculator.

The article, which discussed an analysis published in JAMA, stated that removing race from the equation lowers estimated risk for more Black Americans: “The removal of race from the recalibrated calculator would assign lower risk to more Black Americans and render more of them ineligible for therapy compared with the previous version of the calculator, which did include race as a risk magnifier.”

Now, new research suggests that such concerns may be warranted in at least one population: young non-Hispanic Black adults.

Performance Across Populations

Investigators for the JACC study analyzed Kaiser Permanente Southern California patients aged 20-39 years between 2008 and 2009 without prior cardiovascular disease. The researchers followed them through 2019. They also conducted an exploratory analysis among people aged 20-29 years to assess use of PREVENT outside its intended age range. Those with a history of cardiovascular disease, incomplete information on model predictor variables, or values outside the required range at baseline were excluded.

Of the 161,202 adults in the study, about 52% were Hispanic, 27% were White, 13% were Asian/Pacific Islander, and 9% were Black. Participants were followed for up to 10 years, with cardiovascular disease, atherosclerotic cardiovascular disease, and heart failure identified through electronic health records and claims data. Predicted 10-year risk estimates generated by the PREVENT calculator were then directly compared with observed event rates.

Model performance was evaluated using standard measures of discrimination and calibration, which assess whether predicted risk levels accurately reflect real-world outcomes. Fairness was assessed by examining whether these performance metrics were consistent across racial and ethnic groups.

Shifts in Accuracy

In terms of risk rankings, the PREVENT model appeared accurate. Patients who later developed cardiovascular disease generally had higher predicted risk than those who didn’t, regardless of race and ethnicity.

However, when investigators examined how closely PREVENT’s risk estimates matched what occurred for Black adults, prediction accuracy changed. In this group, 1.4% experienced a cardiovascular event, but the PREVENT model predicted only about half as many events.

In contrast, predicted and observed risks were much closer for other groups. The 10-year incidence of cardiovascular disease was approximately 0.6%-0.7% among Hispanic, White, and Asian/Pacific Islander young adults, and PREVENT’s predicted risks closely matched those observed rates.

In these groups, mean calibration ratios were near 1.0 — indicating good agreement between predicted and observed risk. Specifically, mean calibration for total cardiovascular disease was 1.07 among Hispanic participants, 0.96 among White participants, and 0.96 among Asian/Pacific Islander participants, reflecting only slight over- or underestimation.

Clinical Implications

The authors note that this degree of underprediction could lead clinicians to perceive risk as lower than it truly is and potentially delay preventive interventions when absolute-risk thresholds are used to guide treatment decisions.

“Treatment decisions are based on the absolute risk estimate and based on thresholds set in clinical treatment guidelines. Thus, the fact that the absolute risk is underestimated is important because that may mean certain groups are being undertreated,” said study author Norrina Allen, PhD, MPH, who is a professor of epidemiology and pediatrics at the Northwestern University Feinberg School of Medicine in Chicago.

“The most equitable path forward is to enhance our models with social, structural, and life-course variables that more directly capture the conditions producing cardiovascular disparities,” Allen said.

However, Donald M. Lloyd-Jones, MD, ScM, past president of the American Heart Association, argues that with absolute differences so small, little change is needed in preventive protocols, if at all.

“The ‘miscalibration’ is really only seen in the top 10% group of predicted risk among Black patients,” noted Lloyd-Jones, who is also a professor at Boston University Chobanian & Avedisian School of Medicine and the principal investigator of the Framingham Heart Study in Boston.

“The overall absolute difference between the predicted risk and the observed event rate in all groups and all deciles is very, very small, indicating that there would be no difference that would drive a different action or recommendation for prevention,” he said.

“Furthermore, we aren’t supplied with the characteristics of the patients by decile of predicted risk, but reading between the lines of the baseline characteristics table(s), it is pretty clear that the Black patients in the top decile of predicted risk have a lot of hypertension requiring treatment, among other red flags, so there would be other ways of identifying their potential for higher risk,” Lloyd-Jones said.

Allen reported having no relevant financial relationships.

Lois Anzelowitz Levine is a medical and lifestyle writer in Dallas.