摘要
目的 采用决策树算法构建老年高血压性脑出血(HICH)患者病情转归不良的预测模型。方法 回顾性分析2020年1月到2023年10月我院收治的358例老年HICH患者的临床资料,按7∶3的比例随机分为训练集(n=251)和验证集(n=107)。在发病90 d后根据患者病情转归情况将训练集患者分为转归不良组和转归良好组,收集并比较2组患者的临床资料;采用Logistic回归分析筛选老年HICH患者病情转归不良的影响因素;利用决策树算法构建老年HICH患者病情转归不良的风险预测模型,并通过受试者工作特征曲线(ROC)评价模型的预测效能。结果 老年HICH患者的病情转归不良发生率为27.09%(97/358);多因素Logistic回归结果显示,年龄(OR=1.998,95%CI 1.326~3.009)、吸烟史(OR=1.458,95%CI 1.087~1.956)、饮酒史(OR=1.623,95%CI 1.228~2.143)、合并高脂血症(OR=1.806,95%CI 1.192~2.736)、入院时格拉斯哥昏迷量表(GCS)评分(OR=0.480,95%CI 0.309~0.748)、颅内血肿量(OR=1.745,95%CI 1.281~2.379)、规律服用降压药(OR=0.617,95%CI 0.453~0.841)均是老年HICH患者病情转归不良的影响因素(P<0.05);构建的决策树模型预测训练集和验证集老年HICH患者病情转归不良的曲线下面积(AUC)值为0.909(95%CI 0.867~0.942)、0.795(95%CI 0.706~0.867),敏感度为84.72%、80.00%,特异度为89.94%、84.15%。结论 年龄、吸烟史、饮酒史、合并高脂血症、入院时GCS评分、颅内血肿量、规律服用降压药均为老年HICH患者病情转归不良的影响因素,根据上述因素构建的决策树预测模型对老年HICH患者病情转归不良的预测效能良好。
Objective To construct a prediction model of adverse outcomes in elderly patients with hypertensive intracerebral hemorrhage(HICH)using decision tree algorithm.Methods The clinical data of 358 elderly patients with HICH admitted to our hospital from January 2020 to October 2023 were retrospectively analyzed,and they were randomly divided into training set(n=251)and verification set(n=107)according to a ratio of 7∶3.90 days after the onset of the disease,the patients in the training set were divided into poor outcome group and good outcome group according to their disease outcomes.The clinical data of the two groups were collected and compared.Logistic regression analysis was used to screen the influencing factors of adverse outcomes in elderly patients with HICH.A decision tree algorithm was used to construct a risk prediction model for adverse outcomes in elderly patients with HICH,and the predictive efficacy of the model was evaluated by receiver operating characteristic curve(ROC).Results The incidence of adverse outcomes in elderly patients with HICH was 27.09%(97/358).Multivariate Logistic regression results showed that age(OR=1.998,95%CI 1.326~3.009),smoking history(OR=1.458,95%CI 1.087~1.956),drinking history(OR=1.623,95%CI 1.228~2.143),combined with hyperlipidemia(OR=1.806,95%CI 1.192~2.736),Glasgow Coma Scale(GCS)score at admission(OR=0.480,95%CI 0.309~0.748),intracranial hematoma volume(OR=1.745,95%CI 1.281~2.379)and regular use of antihypertensive drugs(OR=0.617,95%CI 0.453~0.841)were the influential factors for adverse disease outcomes in elderly patients with HICH(P<0.05).The area under the curve(AUC)values of the constructed decision tree model for predicting poor outcomes in elderly HICH patients in the training and validation set were 0.909(95%CI 0.867~0.942)and 0.795(95%CI 0.706~0.867),with sensitivities of 84.72%and 80.00%,and specificities of 89.94%and 84.15%,respectively.Conclusion Age,smoking history,drinking history,combined hyperlipidemia,GCS score at admission,intracranial hematoma volume,and regular use of antihypertensive drugs are all influential factors in the adverse prognosis of elderly HICH patients.The prediction model based on the above factors is effective in predicting the adverse prognosis of elderly HICH patients.
作者
夏文静
陈媛媛
李义沙
王倩
Xia Wenjing;Chen Yuanyuan;Li Yisha(Department of First Neurology,First Hospital of Handan,Handan,Hebei 056000,China)
出处
《四川医学》
2025年第2期192-198,共7页
Sichuan Medical Journal
基金
邯郸市科学技术研究与发展计划项目(编号:1623208039ZC)。
关键词
决策树算法
高血压性脑出血
病情转归
decision tree algorithm
hypertensive intracerebral hemorrhage
disease outcomes
