Construction and Validation of an In-hospital Mortality Risk Prediction Model for Acute Ischemic Stroke Patients

This research is mainly divided into two parts. The first part is to verify and optimize the existing prediction model. Through continuous collection of clinical data of acute ischemic Stroke patients hospitalized in Shenzhen Second People's Hospital from January 2017 to December 2021, including baseline indicators and end point events, based on the existing prediction model (SOAR, GWTG-Stroke), The predictive probability was calculated and compared with the actual mortality during hospitalization. The ROC curve, calibration curve and decision curve were used to evaluate the model's differentiation, calibration and clinical application value.

Using retrospective data, multivariate logistic regression was used to analyze the predictive value of baseline clinical indicators, screen risk factors, and optimize the prediction model of SOAR and GWTG-Stroke.

Extreme Gradient Boosting (XGBOOST) was used to select variables, and logistic regression model was used based on Akaike Information Criterion.

AIC) was used to construct an improved mortality risk prediction Model (New Model). Decision curves were used to compare the models. Combined with the clinical significance of the indicators, the construction of the prediction Model was improved.

The model was validated internally by resampling with computer simulation. The second part is to evaluate the clinical application effect of the improved prediction Model. The clinical data of acute ischemic stroke patients hospitalized in Shenzhen Second People's Hospital and Shenzhen Longhua District People's Hospital from January 2022 to December 2023 are collected continuously. The New Model is applied in the clinic, and the New Model is validated in the external time and space.

Evaluate prediction effectiveness and extrapolation.