JAMA Netw Open
PREVENT cardiovascular risk tool holds up in real-world EHR data
Clinical Takeaway: PREVENT can be used confidently for population-level cardiovascular risk stratification in routine EHR-based care, but clinicians should be cautious when using absolute risk thresholds, especially when key data are missing.
The American Heart Association’s PREVENT equations performed well in everyday clinical practice, maintaining strong cardiovascular disease (CVD) risk discrimination even when electronic health record (EHR) data were incomplete.
In this retrospective cohort study from the Duke University Health System, investigators evaluated PREVENT in two groups: a “relaxed” cohort of 406,230 adults allowing for missing labs and vitals with median imputation, and a “strict” cohort of 127,151 adults with complete data. Over up to 8 years of follow-up, the model’s 5-year CVD risk discrimination remained robust. C-index values were 0.75 in men and 0.77 in women in the relaxed cohort, and 0.77 for both sexes in the strict cohort.
Calibration was stronger when full clinical data were available. Models using imputed data tended to underestimate absolute risk, particularly in some socioeconomic subgroups. Locally refitted or machine learning–recalibrated models led to only modest calibration gains and little change in discrimination.
“These findings underscore the importance of evaluating models in the settings and populations where they will be used,” the authors wrote, noting PREVENT’s utility for outreach and preventive care planning in large health systems.
Source: Hong C, et al. (2026, April 14). JAMA Netw Open. Performance of PREVENT Cardiovascular Risk in Electronic Health Record–Based Clinical Practice