Constructing the Prediction Model for the Risk of Stroke in a Chinese Population

Abstract

Background and Purpose— Prediction rules for the risk of stroke have been proposed. However, most studies were conducted with whites or for secondary prevention, and it is not clear whether these models apply to the Chinese population. The purpose of this study was to construct a simple points-based clinical model for predicting incident stroke among Chinese adults in Taiwan. Methods— We estimated the 10-year risk of stroke in a cohort study of middle-aged and elderly participants who were free from stroke at baseline. Multivariate Cox model-derived coefficients were used to construct the simple points-based clinical and biochemical model and the prediction measures using the area under the receive operating characteristic curve, net reclassification improvement, and integrated discrimination improvement statistics were applied. Results— Of the 3513 participants without stroke at baseline, 240 incident cases of stroke were documented for a median 15.9-year follow-up. Age (8 points), gender (1 point), systolic blood pressure (3 points), diastolic blood pressure (2 points), family history of stroke (1 point), atrial fibrillation (3 points), and diabetes (1 point) were found to significantly predict stroke events. The estimated area under the receive operating characteristic curve for this clinical points-based model was 0.772 (95% CI, 0.744 to 0.799). The discrimination ability of this clinical model was similar to the coefficients-based models and better than available stroke models. Conclusions— We have constructed a model for predicting 15-year incidence of stroke in Chinese adults and this model may be useful in identifying individuals at high risk of stroke.