Logistic回归分析在成人单侧急性创伤性硬膜下血肿预测模型中的应用

Application of Logistic regression analysis in prognosis of acute unilateral subdural hematomas in adults

关键词:  硬膜下血肿  死亡率  预后  Logistic模型  ROC曲线

Key words:  subdural hematoma  mortality rate  prognosis  Logistic model  ROC curve

摘要:

目的阐明与患者死亡率密切相关的影响因素，并建立预测预后的Logistic回归模型。方法对185例单侧急性创伤性硬膜下血肿患者的各项临床指标及其CT发现进行统计分析以明确影响死亡率的独立相关因素。研究的因素包括年龄、性别、合并伤、曈孔变化及对光反应、GCS评分、血肿厚度、血肿量，中线移位和脑肿胀。应用单因素、多因素Logistic回归分析和ROC曲线建立Logistic回归模型并确定概率预测值P的最适诊断界点及其诊断效率。结果年龄、GCS评分和脑肿胀这三项指标进入Logistic回归模型，诊断模型P值的曲线下面积为0.908，最适诊断界点为0.520,敏感性为89.8%，特异性为83.8% ,AUC和敏感性均优于各单项指标。结论单侧硬膜下血肿患者死亡为多因素影响的结果，本研究建立的Logistic回归模型可有效地预测亡率。

Abstract:

Objective To investigate the risk factors of mortality in patients with acute unilateral subdural hematomas by Logistic regression model. Methods One hundred and eighty five patients with acute unilateral subdural hematomas were enrolled in the study. The factors relevant to outcomes included age, gender, accompanying trauma, pupillary changes and light reflex, Glasgow Coma Scale (GCS)score，thickness of the hematoma, volume of the hematoma, the presence of midline shift and brain-swelling. All these factors were measured and analyzed by using univariate and multivariate Logistic regression to form a regression model. Then the optimal cut-off point of probability predictive value P and the diagnostic efficiency were obtained by using ROC curve. Results The age, GCS score and brain-swelling entered the Logistic regression model. Area under curve (AUC) of P value of the diagnostic model was 0.908，with the optimal cut-off point of 0.520，the sensitivity and specificity were 89. 8% and 83. 8% , respectively. Both AUC and sensitivity of the diagnostic model were superior to those of any single factor. Conclusion Mortality rate of unilateral acute subdural hematomas of the adult is influenced by many factors. The multiple Logistic regression model constructed in the study may be applied to predict the mortality rate.