摘要
目的探讨对数几率回归模型(Logistic)构建联合预测因子在慢性乙型肝炎患者(CHB)早期肝硬化(LC)诊断中的临床应用。方法选择本院2018年1月-2019年10月的CHB患者179例、CHB/LC患者179例及测试集57例,检测血常规、血清肝纤维化、凝血功能、血清生化等实验室指标。单因素Logistic回归筛选并用多因素Logistic回归拟合得到联合预测因子,生成受试者工作曲线(ROC),在测试集中进行测试,评价模型预测效果。结果透明质酸(HA)、部分活化凝血酶原时间(APTT)、谷丙转氨酶(ALT)和血小板(PLT)纳入拟合得到新的联合预测因子表达式Logit(X)=39.621+1.365HA+1.213APTT+1.011ALT-0.723PLT。ROC曲线显示X的cut-off值取308.80时,YI指数最大为0.538,AUC为0.815,分别高于单个协变量因子。在模型准确率为70.12%的前提下,CHB合并LC患者测试组的预测LC准确率为72.41%,CHB患者测试组的预测为非LC的准确率为96.42%。结论对数几率回归模型构建的联合预测因子ROC曲线能对CHB患者的早期LC诊疗有更好的预测价值。
Objective To explore the clinical application of logistic regression model to construct combined predictors in the diagnosis of early cirrhosis( LC) in patients with CHB. Methods A total of 179 CHB patients,179 CHB/LC patients and 57 test sets were selected from January 2018 to October 2019 in our hospital to detect laboratory indicators such as blood routine,serum liver fibrosis,coagulation function,and serum biochemistry. Univariate Logistic regression screening and multi-factor logistic regression fitting were used to obtain the joint predictors,generate the receiver operating curve( ROC),conduct tests in the test set,and evaluate the prediction effect of the model. Results Hyaluronic acid( HA),partially activated prothrombin time( APTT),glutaminate alanine transaminase( ALT) and platelet( PLT) were brought into mathematical formula: Logit( X) =39. 621 + 1. 365 HA + 1. 213 APTT + 1. 011 ALT-0. 723 PLT. The ROC showed a maximum Youden index of 0. 538 and a maximum area of 0. 815 when cut-off was 308. 80. Under the premise of model accuracy( 70. 12%),accuracy rate in CHB/LC and CHB test suit was 72. 41% and 96. 42%. Conclusion ROC generated by joint prediction factor of Logistic model provide references to medical resources for CHB patients with LC.
作者
王晔恺
于倩
姚燕珍
罗雯斌
洪开听
鲍舟君
WANG Ye-kai;YU Qian;YAO Yan-zhen;LUO Wen-bin;HONG Kai-ting;BAO Zhou-jun(Clinical Laboratory,Zhoushan Hospital,Zhoushan,Zhejiang 316021,China;不详)
出处
《中国卫生检验杂志》
CAS
2021年第1期90-93,97,共5页
Chinese Journal of Health Laboratory Technology
