目的 分析新生儿肺部感染影响因素，并建立其预测模型。方法 选择2021年1月至2023年4月成都市双流区妇幼保健院住院救治的1714例日龄<7 d新生儿进行研究。采用R 4。1。3软件中"caret"包以7︰3比例随机分为模型组1199例与验证组515例。收集可能影响新生儿肺炎的新生儿资料与母体相关指标，根据有无肺部感染将模型组新生儿分为肺部感染组与非肺部感染组，比较两组新生儿资料与母体相关指标，建立新生儿肺部感染预测模型并用受试者工作特征曲线(ROC)法评估模型区分度，以校准曲线评估模型准确度。结果 模型组1199例新生儿中，33例确诊为肺部感染。肺部感染组与非肺部感染组新生儿出生孕周、出生体质量、出生5 min Apgar评分、胎膜早破情况、胎粪吸入情况、侵入性操作情况、呼吸机应用情况、静脉营养情况、肠道益生菌使用情况、抗生素使用情况均具有统计学差异(P<0。05)。套索回归(LASSO)筛选出6个潜在因素分别为:出生体质量、出生5 min Apgar评分、胎膜早破情况、胎粪吸入情况、侵入性操作情况、肠道益生菌使用情况。以多因素Logistic回归建立模型，结果显示:出生体质量(OR=0。218)、出生5 min Apgar评分(OR=0。573)、胎膜早破情况(OR=8。345)、胎粪吸入情况(OR=4。011)、侵入性操作情况(OR=4。056)、肠道益生菌使用情况(OR=3。637)为新生儿肺部感染的独立性影响因素(P<0。05)。ROC分析结果显示:模型组AUC为0。925，95％CI(0。891～0。959)，特异度为82。4％，敏感度为93。9％。验证组AUC为0。878，95％CI(0。815～0。943)，特异度为69。3％，敏感度为93。9％。校准曲线分析结果显示:模型组与验证组的预测曲线与标准曲线基本拟合。决策曲线结果显示:当模型预测值为0。179～0。889时患者净获益率>0。结论 新生儿肺部感染主要受出生体质量、出生5 min Apgar评分、胎膜早破情况等因素的影响，本研究以上述因素建立的模型用于预测新生儿肺部感染具有较高的准确度与区分度。
Construction and validation of a predictive model for neonatal lung infection
Objective To analyse the factors influencing neonatal lung infections and to develop a predictive model for them.Methods A total of 1714 neonates with day-old<7 d who were hospitalised and treated in Shuangliu District Maternal and Child Health Hospital in Chengdu City from January 2021 to April 2023 were se-lected for the study.They were randomly divided into 1199 cases in the model group and 515 cases in the vali-dation group in a 7︰3 ratio using the "caret"package in R 4.1.3 language software.The neonatal data and ma-ternal indicators that may affect neonatal pneumonia were collected,and the neonates in the model group were di-vided into a pulmonary infection group and a non-pulmonary infection group according to the presence or absence of pulmonary infection.The neonatal data and maternal indicators of the two groups were compared.A predictive model for neonatal pulmonary infection was established and the model differentiation was assessed using the sub-ject's work characteristic curve (ROC)method,and the model accuracy was evaluated using calibration curves. Results Of the 1199 birth cases in the model group,33 cases were diagnosed with pulmonary infection.There was a statistical difference between neonates in the pulmonary infection group and the non-pulmonary infection group in terms of gestational week of birth,body mass at birth,5 min Apgar score at birth,preterm premature rupture of membranes,meconium aspiration,invasive manipulation,ventilator application,intravenous nutri-tion,use of intestinal probiotics,and use of antibiotics (P<0.05 ).Least absolute shrinkage and selection oper-ator (LASSO)was used to screen for six potential factors:birth body mass,5 min appearance pulse grimace ac-tivity respiration (Apgar),preterm rupture of membranes,meconium aspiration,invasive manipulation,and in-testinal probiotic use.Modelling with multifactorial logistic regression showed that birth body mass (OR=0.218),5 min Apgar score at birth (OR=0.573),preterm rupture of membranes (OR=8.345),meconium aspiration (OR=4.011),invasive manipulation (OR=4.056),and intestinal probiotic use (OR=3.637)as independent influences on neonatal lung infection (P<0.05 ).The results of ROC analysis showed that the AUC of the model group was 0.925 with 95％CI (0.891-0.959),specificity was 82.4％and sensitivity was 93.9％. The AUC of the validation group was 0.878,95％ CI (0.815-0.943),specificity was 69.3％ and sensitivity was 93.9％.The results of calibration curve analysis showed that the prediction curves of the model group and validation group were basically fitted to the standard curve.The results of the decision curve showed that the net patient benefit rate was>0 when the model prediction value was 0.179-0.889.Conclusion Neonatal pulmonary infections are mainly influenced by factors such as birth body mass,5 min Apgar score at birth,and preterm rup-ture of membranes,etc.In this study,the model established with the above factors was used to predict neonatal pulmonary infections with a high degree of accuracy and discrimination.
万方数据
维普
