Development and Validation of a Nomogram Prediction Model for New Stroke in Community Population
Yanyan Wang1, Fei Han1, Dingding Zhang2, Yicheng Zhu1
1Department of Neurology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, 2Center for Prevention and Early Intervention, National Infrastructures for Translational Medicine, Institute of Clinical Medicine, Peking Union Medical College Hospital, Chinese Academy of Medical Science and Peking Union Medical College, Beijing 100730
Objective:
To construct a model to predict new stroke based on community population.
Background:
The use of risk prediction models to optimize screening and interventions of stroke is recommended, but existing models mostly focused on a single dimension.
Design/Methods:
This was a prospective community-based cohort study in which eligible stroke-free participants were continuously recruited and were followed once a year for incident stroke. Sociodemographic information, serum biochemical markers and neuroimaging data were collected at baseline. Cox proportional regression method was used to construct the prediction model of new stroke. Bootstrap sampling was utilized for internal validation.
Results:
During a mean follow-up of 8.2 years, 54/1173 (4.6%) had an incident stroke. Current smoking (HR = 1.940, p = 0.023), diabetes mellitus (HR = 3.251, p < 0.001), LnsFA (serum folic acid) (HR = 0.401, p = 0.002), cerebral small vessel disease (CSVD) burden (HR = 2.356, p < 0.001) and intracranial artery stenosis (ICAS) ≥50% on MRA (HR = 5.070, p < 0.001) were significant and independent predictors of new stroke. A Cox proportional hazards model consisting of 5 important factors was constructed to predict the risk of new stroke. After 500 times Bootstrap sampling, the AUC value of this prediction model for 5, 7 and 9 years was 0.881 (0.809-0.944), 0.854 (0.768-0.925) and 0.855(0.771-0.931) respectively; the Brier score for 5, 7 and 9 years was 0.030 (0.020-0.041), 0.038 (0.026-0.049) and 0.040 (0.026,0.056) respectively. The decision curve analysis demonstrated that the model had a favorable clinical utility.
Conclusions:
Current smoking, diabetes mellitus, serum folic acid, CSVD burden and severe ICAS are independent predictors of new stroke in the general population. The combination of traditional stroke risk factors, serum biochemical index and neuroimaging index can predict new stroke well.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.