External Validation of Two 10-Year Stroke Risk Prediction Models Using Korean Genome and Epidemiology Study Data

Abstract

This study aimed to externally validate two 10-year stroke risk prediction models: one developed in Korea (Model 1) and the other in China (Model 2), using community-based cohort data. Data from 8432 participants in Model 1 and 8915 participants in Model 2 were analyzed. Stroke risk was calculated based on each model's equations, and model performance was assessed using the area under the receiver operating characteristic curve (AUC). Age, blood pressure, and diabetes mellitus were significant common risk factors for stroke in both men and women. The AUCs of Model 1 were 0.72 for men and 0.68 for women, while those of Model 2 were 0.70 for men and 0.66 for women. When simplified to include 3-5 significant factors, the AUCs of Model 1 remained almost unchanged for both sexes, while those of Model 2 were 0.68 for men and 0.63 for women. In conclusion, the performance of both models was insufficient for the participants in this study. Therefore, it is suggested that stroke prediction models be developed using the most significant risk factors of each target population.

Keywords: risk assessment; stroke; validation study.