Abstract:

Objective To explore the application value of a machine learning-based model using XGBoost algorithm in predicting 90-day death after ischemic stroke. Methods Ischemic stroke patients from China National Stroke Registry (CNSR) database were selected as subjects, all of them were randomly divided into a training set and test set at a ratio of 7:3. The training set was used to build the prediction model, and the test set was used to evaluate the performance of the model. XGBoost and logistic regression were used to construct the 90-day death prediction model of ischemic stroke. The prediction value of the two models were evaluated by the area under the ROC curve (AUC). Results A total of 10 645 patients were included, with an average age of 65.18±12.23 years old and 4045 females (38.0%). The NIHSS score at admission was 4 (2-9) points. There were 447 deaths (4.48%) at 3 months. The AUC of XGBoost and logistic regression model was 0.8539 and 0.8278, respectively (P =0.0835); the sensitivity was 0.7413 and 0.7133, respectively, and the specificity was 0.8286 and 0.8040, respectively. Conclusions The prediction model of 90-day death after ischemic stroke based on XGBoost machine learning algorithm had a good prediction value.

Key words: Ischemic stroke; Prediction model; Machine learning; Death

摘要：

目的 探讨基于机器学习算法XGBoost构建缺血性卒中发病3个月死亡预测模型的应用价值。 方法 选择中国国家卒中登记（China National Stoke Registry，CNSR）数据库中缺血性卒中患者为研 究对象。按照7∶3比例随机分为训练集和测试集，训练集用于构建预测模型，测试集用于评价模型效 果。分别采用XGBoost和Logistic回归方法构建缺血性卒中发病3个月死亡预测模型，通过ROC曲线下面 积（area under the curve，AUC）评价两种模型的预测价值。 结果 共纳入10 645例缺血性卒中患者，平均年龄65.18±12.23岁，女性4045例（38.0%），入院 NIHSS评分4（2～9）分,3个月死亡患者447例（4.48%）。XGBoost和Logistic回归预测模型的AUC分别为 0.8539、0.8278（P =0.0835），灵敏度分别为0.7413、0.7133，特异度分别为0.8286、0.8040。 结论 基于机器学习算法XGBoost构建的缺血性卒中死亡预测模型表现良好且稳定。

关键词: 缺血性卒中; 预测模型; 机器学习; 死亡