感染相关肝功能障碍预测模型的建立和评价  

ESTABLISHMENT AND EVALUATION OF A PREDICTION MODEL FOR INFECTION-RELATED LIVER DYSFUNCTION

作　　者：王帅[1] 刘华敏[2] 辛兆瑞 王万腾 王芸芸[1] 孙运波[1] WANG Shuai;LIU Huamin;XIN Zhaorui;WANG Wanteng;WANG Yunyun;SUN Yunbo(Department of Intensive Care Unit,The Affiliated Hospital of Qingdao University,Qingdao 266071,China)

机构地区：[1]青岛大学附属医院重症医学科,山东青岛266071 [2]青岛大学附属医院肿瘤科,山东青岛266071

出　　处：《青岛大学学报（医学版）》2021年第5期657-661,共5页Journal of Qingdao University(Medical Sciences)

基　　金：吴阶平医学基金会临床科研专项资助基金项目(HRJJ-20171017)。

摘　　要：目的通过数据分析,建立感染相关肝功能障碍的预测模型,并对模型进行评价。方法回顾性分析青岛大学附属医院重症医学科收治的302例感染病人的临床资料,按照是否发生肝功能障碍,将病人分为无肝损害组和肝损害组。对与肝功能障碍发生相关的危险因素进行Logistic回归分析后,建立感染相关肝功能障碍的预测模型,绘制联合预测因子和其他危险因素的受试者工作特征曲线(ROC曲线)进行评价。结果经过Logistic回归分析后得到4个与肝功能障碍发生相关的危险因素:急性生理与慢性健康状况评估Ⅱ(APACHEⅡ)评分、乳酸、C反应蛋白以及长期饮酒史。建立的感染相关肝功能障碍预测模型为:L=长期饮酒史+0.220 APACHEⅡ评分+0.011 C反应蛋白+2.707乳酸。对预测模型进行评价,模型系数的全局性Omnibus检验有统计学意义(χ^(2)=248.56,P<0.001);联合预测因子L的ROC曲线下面积为0.950,95%置信区间为0.925~0.975;当L>8.895时模型预测肝功能障碍发生的灵敏度为0.859、特异度为0.905。结论感染相关肝功能障碍预测模型可运用于临床,而且有较高的预测价值。Objective To establish a prediction model for infection-related liver dysfunction and to evaluate the model through data analysis. Methods A retrospective analysis was performed on the clinical data of 302 patients with infection who were admitted to Department of Intensive Care Unit, The Affiliated Hospital of Qingdao University. The patients were divided into normal liver function group and liver dysfunction group. After a logistic regression analysis of the risk factors associated with liver dysfunction, a prediction model was established to predict the occurrence of infection-related liver dysfunction, and the receiver operating characteristic(ROC) curve of the combined predictor and other risk factors was plotted to evaluate the predictor. ResultsThe logistic regression analysis identified four risk factors associated with liver dysfunction: Acute Physiology and Chronic Health Evaluation Ⅱ(APACHE Ⅱ) score, concentration of blood lactic acid, C-reactive protein, and the long history of drinking. The established prediction model for infection-related liver dysfunction was: L=long history of drinking+0.220 APACHE Ⅱ score+0.011 C-reactive protein+2.707 lactic acid. The prediction model was evaluated using the global Omnibus test of model coefficients, which was statistically significant(χ^(2)=248.56,P<0.001). The area under the ROC curve of the combined predictor L was 0.950(95%CI=0.925-0.975). When L>8.895, the model for predicting the occurrence of liver dysfunction had a sensitivity of 0.859 and specificity of 0.905. Conclusion The model for predicting infection-related liver dysfunction can be used in clinical practice and has a high predictive value.

关 键 词：感染 肝疾病 回归分析 预测 模型 统计学 

分 类 号：R575.3[医药卫生—消化系统]