(OR08-04-25) Stroke Risk Prediction Using a Nutrition and Health Cohort in China: The CHNS Study

Objectives: Stroke is a leading cause of mortality and disability worldwide, making accurate risk prediction crucial for effective prevention and management. This study aimed to develop and evaluate 5-year and 10-year stroke risk prediction models for Chinese adults aged 40 and older, using data from the China Health and Nutrition Survey cohort.

Methods: The study population included 9,801 participants in the 2018 survey, and 8,000 (18,325 observations) and 14,955 (57,725 observations) participants who were stroke-free at baseline and participated at least two waves in 2009-2018 and 1997-2018, respectively. Stroke occurrence was determined by self-reported doctor diagnosis. Various machine learning algorithms were used to identify predictors using data in 2018. The identified top predictors were incorporated into Cox proportional hazards models to develop the 5-year and 10-year stroke prediction models.

Results: The Multi-Layer Perceptron (MLP) model showed the best performance (F1=0.864, AUC =0.871). Predictors derived from MLP were used to fit the models. The 5-year stroke prediction model (without biomarkers) included 6 factors: age, urban-rural residence, systolic blood pressure, gender, overweight/obesity and fruit intake. The 5-year stroke prediction model (with biomarkers) included 7 factors: age, urban-rural residence, systolic blood pressure, gender, fruit intake, HDL-C, and TC. The 5-year models demonstrated good discrimination (C-index 0.767-0.770, AUC 0.734-0.758) and calibration. The 10-year model included 6 factors: age, gender, systolic blood pressure, physical activity, coarse grain intake, and overweight/obesity, with a C-index of 0.833 and AUC of 0.836.

Conclusions: High-risk predictors for stroke included older age, high systolic blood pressure, male, urban residence, dyslipidemia, overweight/obesity, and insufficient fruit /coarse grain intake. The developed 5-year and 10-year stroke risk prediction models showed good predictive performance and can be used for stroke risk assessment and intervention in primary care settings in China.

Funding Sources: The research received funding from the National Institutes of Health (NIH) and the Ministry of Finance of the Republic of China.