早产预测多因素logistic回归模型的建立及预测价值的评价

摘要

目的：建立多因素logistic回归模型来预测早产。方法对孕24～35周有先兆早产症状的孕妇检测阴道分泌物中胎儿纤维连接蛋白（fFN），经腹部超声测量宫颈长度，记录妊娠终止的孕周。经单因素logistic回归分析筛选与早产有相关性的因素建立多因素logistic回归模型。结果逐步回归分析得出以双胎妊娠、受孕方式、fFN检测及宫颈长度为自变量的早产预测的logistic回归模型，拟合优度检验P＝0．99，ROC曲线下面积为0．97，以P＝0．25为分界点预测早产灵敏度为91．3％，特异度为90．3％。结论本研究建立的多因素logistic回归模型对早产具有较好的预测价值。%Objective To establish multivariate logistic regression model to predict preterm delivery.Methods Gestational women who got clinical visits to the first Affiliated Hospital of Anhui Medical University presenting with threatened preterm labor between 24 -35 weeks of gestation during Feb.2009 and Dec.2009 were recruited to this study.The fFN in the cervical-vaginal secretion was tested.Cervical length were measured using transabdorminal ultrasound.Univariate logistic regression analysis were carried out to extracted variables related to preterm delivery to establish mutivariable logistic regression model.Results Through stepwise logistic regression analysis,the predictive re-gression model composed by twin pregnangcy,ART,cervical length and fFN test was established with AUC =0.97.Homser-Lemeshow test showed the model had very excellent goodness-fit(P=0.99).The predictive sensitivity and specificity were 91.3% and 90.3% respectively while the cutoff point was 0.25.Conclutions Logistic regression model established by this study has good predictive value of preterm delivery.