Integrating epidemiologic modeling and explainable machine learning to evaluate body roundness index for WHO-defined high cardiovascular risk: evidence from the ChinaHEART-Luohe screening cohort

Background: Efficient identification of individuals at high cardiovascular disease (CVD) risk is essential for prevention in middle-aged and older adults. The body roundness index (BRI), derived from waist circumference and height, may capture body-shape-related risk beyond conventional measures. We examined the association of BRI with World Health Organization (WHO)-defined CVD high-risk status in a community-based screening population. Methods: This cross-sectional study used baseline data from the Luohe branch of the ChinaHEART cohort, a community-based health screening program in Luohe, Henan, China (March 2021 to February 2022), including adults aged 35-75 years. WHO-defined CVD high-risk status was determined using WHO CVD risk charts, with an estimated 10-year risk ≥20% classified as high risk. BRI was analyzed as a continuous variable (per 1-unit increase), quartiles, and a binary variable using a receiver operating characteristic (ROC)-derived threshold. Multivariable logistic regression, restricted cubic splines, ROC analysis with bootstrap confidence intervals, and subgroup/interaction analyses were performed. An explainable machine-learning workflow (LASSO, random forest, and SHAP) was also applied. Results: Among 6,858 participants, 1,489 (22%) were classified as WHO-defined CVD high risk. Higher BRI remained associated with high-risk status in fully adjusted models. ROC analysis showed only modest standalone discrimination, while subgroup analyses suggested heterogeneity by sex and cardiometabolic strata. In machine-learning analyses, BRI was retained among selected predictors and contributed meaningfully within the multivariable model. Conclusion: In this community screening population, BRI was positively associated with WHO-defined CVD high-risk status and may serve as a low-cost adjunct marker to prioritize individuals for comprehensive risk evaluation in primary-care screening settings.