Analysis of sex-specific stroke risk factors in middle-aged and elderly Chinese population based on machine learning approach: A retrospective observational cohort study.

Stroke is a leading cause of global mortality and disability, yet comprehensive sex-specific stroke predictors are lacking. This study aimed to identify sex-specific stroke risk factors in middle-aged and elderly Chinese to improve early detection. 12,975 participants (7079 females, 5896 males) aged ≥45 from the China Health and Retirement Longitudinal Study 2011 to 2020 were analyzed. Sex-stratified correlations of 27 health indicators were examined. Eight machine learning algorithms identified significant stroke risk factors. Sex-specific associations between these risk factors and stroke risk were further analyzed using Cox proportional hazards models. Finally, a nomogram was developed to predict stroke risk. Males had higher stroke prevalence than females (P < .001). Nine key predictors were identified: triglyceride and glucose index, waist circumference, low-density lipoprotein-cholesterol, hematocrit, diastolic blood pressure, total metabolic output, mean-corpuscular volume, systolic blood pressure (SBP), waist-to-height ratio, and Cystatin C, with sex differences. High DBP, systolic blood pressure, total metabolic output, and Cystatin C in both males and females were still significantly associated with the risk of stroke (P < .05). The nomogram model exhibited better discrimination compared with other individual predictive factors. This study screened out 9 key parameters through machine learning algorithms and established a nomogram prediction model. It emphasized the importance of simultaneously considering the comprehensive risk score and sex-specific factors in clinical practice, thus providing a scientific basis for improving the prevention and treatment strategies for stroke in middle-aged and elderly populations.