Abstract: Objective To construct a nomogram and classification and regression tree (CART), to explore the application value of the two models in predicting the risk of stigma in elderly stroke patients, and to provide reference for early identification and intervention in high-risk patients. Methods The data of 252 elderly stroke patients hospitalized in a tertiary grade-A hospital in Baoding from May to October 2020 were retrospectively analyzed. The patients were divided into non-stigma group (n=125) and stigma group (n=127) according to whether stigma occurred or not, and a nomogram and CART were established. Internal validation was performed by Bootstrap resampling 1,000 times, and the predictive performance of the 2 models was compared in terms of AUC, sensitivity, specificity, positive predictive value, negative predictive value, and predictive accuracy. Results The AUC, sensitivity, specificity, positive predictive value, negative predictive value and predictive accuracy of CART and nomogram was 0.903 and 0.880, 90.6% and 86.8%, 86.0% and 80.0%, 85.1% and 77.8%, 90.7% and 87.4%, and 88.0% and 82.6% respectively. The differences were statistically significant when comparing the AUC value of the two models (t=19.814, P<0.001). Conclusion Both models have high accuracy and better guidance value, and the prediction performance of the CART is slightly better than that of the nomogram, which could guide the assessment of patient stigma and targeted interventions.

Key words: stroke in elderly people, stigma, nomogram, classification and regression tree, predictive model