基于决策树及Logistic回归分析预测颅脑外伤手术患者术后病情恶化风险

打开文本图片集

【中图分类号】 R651.15 【文献标志码】 A 【文章编号】 1672-7770（2025)05-0553-06

Risk prediction posperative deterioration in patients with traumatic brain injury based on decision tree and logistic regression analysis GUAN Celin， ZHENG Jinliang， CHEN Tao ，LAN Fulin，SU Xinhong.

Corresponding author : SU Xinhong

Abstract:Objective Toconstruct and compare predictive models for posperative deterioration risk in patients undergoing craniocerebral trauma surgery based on decision tree and Logistic regression analysis.MethodsThe clinical data 3OO patients with craniocerebral trauma who were treated inLongyan Afiliated University from July 2020 June 2024 were analyzed retrospectively. Patients were grouped based on differences in posperative clinical outcomes. Independent variables were identified using multiple facr Logistic regression，and Logistic regression modeling was employed. A decision tree model was constructed using univariate analysis，and the stability the Logistic regression model was tested using Omnibus tests combined with 5-fold cross-validation. The ROC curve was used compare the discriminative effcts the two models.ResultsUnivariate analysis showed statistically significant differences between the two groups in age，body mass index(BMI），Glasgow coma scale(GCS） score，cause injury，type intracranial hemama，and surgical time( P

0.803 (95% CI=0.745-0.850) ，respectively. The comparison between the two models did not reach a significant level. After Hosmer-Lemeshow calibration,both models showed good fit(the values χ2 were 8.952 and 8.631 respectively，and the P-values were 0.442 and 0.385 respectively). ConclusionsBoth models constructed based on the aforementioned facrs can be used predict the risk posperative deterioration in patients undergoing craniocerebral trauma surgery. The decision tree model provides a clear clinical decision-making path through a tree-like structure and has greater visualization advantages. The results this study have reference significance for posperative risk management，but their clinical application still requires further verification.

KeyWords: traumatic brain injury；surgery；condition deterioration；decision tree; logistic regression

颅脑外伤(traumaticbraininjury，TBI)是全球致死致残的主要病因之一，年发病率约为250/10万[1]。（剩余14267字）