A pooled cross-sectional analysis of 235,571 U.S. adults from the 2021–2023 Behavioral Risk Factor Surveillance System (BRFSS) identified diabetes (aOR 1.55), poor self-reported general health (aOR 1.58), current smoking (aOR 1.44), and Black race (aOR 1.31) as independent predictors of stroke after full multivariable adjustment. Conversely, regular physical activity (aOR 0.86), higher income (aOR 0.92), Asian race (aOR 0.65), and Hispanic ethnicity (aOR 0.65) were associated with meaningfully lower stroke odds. The final logistic model achieved an AUC of 0.781, indicating good discriminative ability.
While none of these associations are surprising in isolation — the diabetes-stroke link has been robust across decades of cardiovascular research — the strength of this dataset lies in its size and survey-weighted design, which corrects for the complex BRFSS sampling structure and improves generalizability to the broader U.S. adult population. The 55% elevated odds for diabetes and the income gradient reinforce that stroke remains deeply entangled with socioeconomic determinants, not merely biological ones. The protective signal for physical activity (14% lower odds), though modest, aligns with established dose-response evidence from prospective trials.
Key limitations are significant: the cross-sectional design precludes causal inference, stroke status is self-reported, and residual confounding by unmeasured variables like atrial fibrillation or medication adherence cannot be excluded. Critically, this is a preprint posted on medRxiv and has not yet undergone peer review — findings should be interpreted cautiously until formally vetted. Overall, this is a confirmatory, policy-reinforcing study rather than paradigm-shifting science.