Data Management And Sharing System of China National Clinical Research Center for Neurological Diseases
GWTG Risk Model for All Stroke Types Predicts In-Hospital and 3-Month Mortality in Chinese Patients with Acute Stroke
First author：Sun Shichao,Pan Yuesong
Corresponding author：Guo Li
essential_information
- Person in charge：Sun Shichao,Pan Yuesong
- Data Sources： CNSR-I
- Number of papers：CNSR-I20190231
- Published time：2018-12-13
2018-12-13
Published articles
Journal：J Stroke Cerebrovasc Dis
Publication type: Original Contributions
IF(IF): 1.598
PMID: 30553646 Link To Pubmed Full text Download
Background And Purpose
We aimed to externally validate the Get With the Guidelines (GWTG) risk model for all stroke types to predict in-hospital stroke mortality in Chinese patients and moreover to explore its prognostic value in predicting 3-month mortality after stroke.
Method
The prognostic model was applied to patients with acute stroke from China National Stroke Registry II (CNSR II) to predict in-hospital and 3-month mortality. Model discrimination was estimated by calculating c-statistic and 95% confidence intervals (CIs). Calibration was assessed by Pearson correlation coefficient and Hosmer-Lemeshow test.
Conclusion
The GWTG risk model for all stroke types is a valid clinical tool to predict in-hospital and 3-month mortality in Chinese patients with acute stroke of any type.
Main results
Date from 21,684 stroke patients with complete data for in-hospital mortality prediction and 20,348 stroke patients with complete data for 3-month mortality prediction in the CNSR II were abstracted. The in-hospital and 3-month mortality were 1.4% and 5.6%, respectively. The c-statistics in the CNSR II were .86 (95% CI, .84-.88) and .83 (95% CI, .81-.84) for in-hospital and 3-month mortality, respectively. Calibration plot presented high correlation between the observed and predicted mortality rates (Pearson correlation coefficient, .996 for in-hospital and .998 for 3-month mortality; both P < .001). The Hosmer-Lemeshow statistics for the prediction of in-hospital and 3-month mortality were 0.21 and less than .001, respectively. The model performed nearly as well in each stroke type as in the overall model including all types.