Ó³»­´«Ã½

BACKGROUND: The American Heart Association's Predicting Risk of cardiovascular Events (PREVENT) aims to improve cardiovascular risk prediction. Whether PREVENT suitably predicts 10-year incident atherosclerotic cardiovascular disease (ASCVD) in health care populations is unknown.OBJECTIVES: This study sought to evaluate the calibration and discrimination of the PREVENT equations across integrated U.S. health care systems.METHODS: We retrospectively evaluated electronic health records of 270,320 patients (Mass General Brigham [MGB]: 136,654; Mount Sinai Health: 43,321; Penn Medicine: 56,889; Vanderbilt University Medical Center: 33,456) aged 30 to 79 years and without prior ASCVD from 2010 to 2014. We compared 10-year estimated ASCVD risk based on PREVENT and empirically observed first ASCVD event over a 10-year follow-up. Calibration was assessed based on Greenwood-Nam-d'Agostino test, discordance, and mean calibration. Discrimination was assessed with the use of the time-dependent Harrell's C-index.RESULTS: Based on PREVENT, the mean estimated 10-year ASCVD risk was 4.9% ± 4.7% in MGB (mean age 54 years, 42% female), 6.0% ± 5.6% in Mount Sinai (mean age 56 years, 54% female), 6.0% ± 5.2% in Penn (mean age 58 years, 55% female), and 4.8% ± 1.3% in Vanderbilt (mean age 60 years, 51% female). Although PREVENT underestimated the observed incidence rate in MGB (discordance: -71.0%), Mount Sinai (discordance: -36.2%), and Vanderbilt (discordance: -40.0%), it more closely mirrored the empirical rate in Penn (+1.3%). Overall, PREVENT yielded moderate discrimination C-index in MGB (0.70 [95% CI: 0.70-0.70]), Mount Sinai (0.74 [95% CI: 0.73-0.75]), Penn (0.69 [95% CI: 0.68-0.70]), and Vanderbilt (0.73 [95% CI: 0.72-0.74]). Nevertheless, calibration differed by sex, with greater underestimation among women in MGB (discordance: -80.7%) and Vanderbilt (discordance: -55.5%) but among men in Mount Sinai (discordance: -40.7%). The race and ethnicity-stratified predictive performance varied across health care systems. Compared with the pooled cohort equations, PREVENT demonstrated better overall calibration in Penn (+2.5% vs +93.6%) but worse in MGB (-70.0% vs -41.6%) and Mount Sinai (-36.4% vs 4.8%), notwithstanding comparable discrimination. The PREVENT predictive performance further differed with better discrimination among patients without diabetes mellitus or antihypertensives.CONCLUSIONS: The PREVENT model moderately discriminated ASCVD incidence across 4 geographically distinct academic health care systems in the United States. However, calibration metrics varied widely across health care systems, sociodemographics, and underlying cardiometabolic comorbidities.