Construction and validation of a predictive model for cognitive impairment risk in patients with first ischemic stroke
Objective To investigate the risk factors of cognitive impairment in patients with first ischemic stroke(IS),and to construct and validate a risk prediction model for post-stroke cognitive impairment(PSCI).Methods A total of 408 patients with first IS diagnosed at a third-level grade A hospital of integrated traditional Chinese and Western medicine in Shanghai from February 2020 to February 2022 by convenience sampling method was selected and followed up for 6 months,and were divided into training sets(n=286)and internal validation sets(n=122).The patients admitted to another third-level grade A hospital of integrated traditional Chinese and Western medicine in March 2022 were used as the external validation group(n=55).logistic regression analysis was used to screen the risk factors for cognitive impairment in patients with first IS.The nomogram model was constructed using R soft-ware,and the prediction effect was evaluated using the area under the curve(AUC)of receiver operating characteris-tic(ROC),C statistic,and calibration curve.Results The constructed model was Y=ez/(1+ez),where z=-9.346+0.101 × age+0.039 × Hcy+0.248 × NIHSS score+1.499 × temporal lobe infarction+1.626 × qi-xu syndrome.The AUC of model's ROC was 0.853 with 95％CI of 0.807-0.898,the optimal cutoff value was 0.503,the sensi-tivity was 74.8％,the specificity was 87.1％,and the accuracy was 80.1％.The C statistics for internal and exter-nal validation were 0.842(95％CI was 0.817-0.867)and 0.735(95％CI was 0.643-0.826),respectively,and the Brier scores were both less than 0.25.The calibration curve and DCA curve showed that the model had good fit-ting.Conclusion The risk prediction model constructed in this study has good discriminatory power and calibration,which can assist medical staff in screening high-risk populations for post-stroke cognitive impairment and provide ref-erence for early identification and management intervention.
万方数据
维普
