剖宫产产妇产褥期感染预测模型建立及验证

Establishment and verification of a predictive model of puerperal infection in cesarean section women

目的通过建立剖宫产产妇产褥感染预测模型并对其进行验证,以此提高对产褥感染的早期干预能力。方法选择武汉儿童医院2018年12月-2020年12月选择剖宫产进行分娩的产妇作为研究对象,将2018年12月-2019年12月剖宫产产妇共420例作为建模组,根据是否发生产褥感染将其分为感染组(n=36)和非感染组(n=384),选择2020年1-12月剖宫产产妇395例作为验证组。通过收集研究对象临床资料,建立剖宫产产妇产褥期感染预测模型,并以Hosmer-Lemeshow法分析模型拟合优度。结果感染组与非感染组分娩过程中失血量、合并妊娠阴道炎、合并胎膜早破情况以及C-反应蛋白(CRP)、降钙素原(PCT)、中性粒细胞淋巴细胞比值(NLR)水平比较差异具有统计学意义(P<0.05);多元Logistic回归分析结果显示,分娩过程中失血量、合并妊娠阴道炎、合并胎膜早破、CRP、PCT、NLR水平升高是产妇产褥期发生产褥感染的独立危险因素(P<0.05);ROC分析结果显示,与单指标评价相比,产褥感染模型用于产褥感染的预测价值(AUC)最高,提示该模型具有良好的预测效能;Hosmer-Lemeshow法分析结果显示,所建立预测模型与观测值具有较好的拟合度(χ^(2)=2.185,P=0.795);所建立感染模型在验证组检验总准确率为90.89%,总体判别能力较好。结论该模型用于剖宫产产妇产褥感染预测总体判别能力较好,可在临床应用。

OBJECTIVE To improve the early intervention ability of puerperal infection by establishing and verifying the prediction model of puerperal infection in cesarean section women.METHODS The women who chose cesarean section for delivery in Wuhan Children′s Hospital from December 2018 to December 2020 were selected as the research subject.A total of 420 cases of cesarean section women from Dec 2018 to Dec 2019 were selected as the modeling group and divided into infected group(n=36)and non-infected group(n=384)according to whether they had puerperal infection or not,and 395 cases of cesarean section women from January to December 2020 were selected as the verification group.By collecting the clinical data of the study subjects,a predictive model of puerperal infection in cesarean section women was established,and the goodness of fit of the model was analyzed by the Hosmer-Lemeshow method.RESULTS There were significant differences in blood loss during delivery,vaginitis during pregnancy,premature rupture of membranes,and CRP,PCT,and NLR levels between the infected group and the non-infected group(P<0.05).The results of multivariate logistic regression analysis showed that blood loss during delivery,combination of pregnancy vaginitis,combination of premature rupture of membranes,elevation of CRP,PCT,and NLR levels were independent risk factors for puerperal infection during the puerperium(P<0.05).ROC analysis results showed that compared with single index evaluation,the predictive value(AUC)of the puerperal infection model was the highest,indicating that the model had good predictive performance.The Hosmer-Lemeshow method analysis results showed that the established predictive model had a good degree of fit with the observed value(χ^(2)=2.185,P=0.795);The total accuracy of the established infection model in the verification group was 90.89%,and the overall discriminative ability was good.CONCLUSION This model had good overall discriminative ability for the prediction of puerperal infection in cesarean section women,and could be used in clinical applications.