JMIR Publications

ABSTRACT

Background:

The “8+2” Stroke Risk Score has been widely used to identify individuals with a high risk of stroke, despite there being insufficient evidence to confirm the predictive ability of stroke events.

Objective:

We aimed to validate the ability of the Risk Score to predict the risk of stroke within 10 years in community cohort populations, and attempt to optimize the scoring method to improve predictive ability.

Methods:

For this population-based cohort study, 22259 community residents were recruited in 2013 from one urban and rural monitoring site in Ningxia, China. The occurrence of stroke was established by a combination of self-reporting and review of hospitalization electronic records (the International Classification of Diseases-10: I60-63). A logistic regression model and a Rothman-Keller model were used to refine the eight-factor stroke risk score to predict the 10-year stroke risk. The performance of the model performance was assessed by net reclassification indices (NRI).

Results:

A total of 11692 community residents aged 40 years or older completed the 10-year of follow-up. According to the Risk Score, the stroke incidence of in the low-risk (n=8909), medium-risk (n=1074) and high-risk groups (n=1710) was 4.5%, 14.7%, and 12.3%, respectively. Relative to the Risk Score, the absolute NRI of the Rothman-Keller model was 0.051 (P<0.05) and of the logistic regression model was 0.010 (P>0.05).

Conclusions:

Our study confirmed that the“8+2” Stroke Risk Score does not have the desired effect to predict stroke event. The Rothman-Keller model may improve the ability to predict individuals with a high-risk for stroke.