新生儿肺炎继发腹泻发生风险调查及其预测模型的建立  

A Study of the Risk of Diarrhea Secondary to Pneumonia in Neonates and the Development of Risk Prediction Model

作　　者：吴玉萍 林平 邓以婷娜 唐静 WU Yuping;LIN Ping;DENG Yitingna;TANG Jing(Department of Neonatology,Chengdu Shuangliu Maternal and Child Health Hospital,Chengdu 610299,China)

机构地区：[1]成都市双流区妇幼保健院新生儿科,四川成都610299

出　　处：《标记免疫分析与临床》2024年第7期1281-1286,1297,共7页Labeled Immunoassays and Clinical Medicine

摘　　要：目的调查新生儿肺炎继发腹泻发生风险,建立其预测模型并进行验证。方法选择2022年7月至2023年12月接诊的新生儿肺炎患儿432例进行研究。以7∶3比例将患儿随机分为模型组302例,验证组130例。收集可能影响新生儿肺炎患儿继发腹泻的相关因素,根据有无出现继发性腹泻将模型组患儿分为腹泻组与无腹泻组,比较2组患儿一般资料、疾病与治疗相关因素及实验室指标,以套索回归(least absolute shrinkage and selection operator,LASSO)筛选变量后行多因素Logisitic回归,以此建立列线图模型并进行验证。结果模型组302例新生儿肺炎患儿中共有81例(26.82%)出现腹泻,根据LASSO回归结果行多因素Logisitic回归分析结果显示,出生孕周、喂养方式、侵入性操作、抗生素联用、糖皮质激素使用情况、微生态制剂使用情况、住院时间为新生儿肺炎继发腹泻的独立性影响因素(P<0.05)。模型组接收者操作特性曲线(receiver operating characteristic curve,ROC)下面积为0.779,95%CI为0.719~0.838,灵敏度为79.0%,特异性为63.8%;验证组ROC曲线下面积为0.751,95%CI为0.689~0.813,灵敏度为85.2%,特异性为54.3%。模型组与验证组校准曲线斜率为1,截距为0.000,模型曲线与理想模型基本拟合成对角线。H-L拟合优度检验(P>0.05)。临床有效性分析结果显示,当预测概率阈值为0.15~0.75时使用本研究模型预测新生儿肺炎继发腹泻的净获益最高。结论新生儿肺炎患儿继发腹泻主要受出生孕周、喂养方式、侵入性操作等因素的影响,本研究建立的列线图模型用于预测新生儿肺炎继发腹泻发生风险具有较高的准确度与区分度。Objective To investigate the risk of secondary diarrhea caused by pneumonia in newborns,and to develop and validate a risk prediction model.Methods A total of 432 neonates with pneumonia were selected from July,2022 to December,2023.These children were randomly divided into a model training group(302 cases)and a validation group(130 cases)with a ratio of 7∶3.The relevant factors that could affect the secondary diarrhea of neonatal pneumonia children were collected for model development,and the children in the model training group were divided into the diarrhea group and the diarrhea group according to the occurrence of secondary diarrhea.The general information,disease and treatment-related factors and laboratory indicators of the children in the two groups were compared.Multivariate logistic regression was applied by least absolute shrinkage and selection operator(LASSO)to train and validate the nomographic model.Results Among 302 children with neonatal pneumonia in the model training group,81 cases(26.82%)had diarrhea.According to the LASSO regression results,multivariate logistic regression analysis showed as follows:gestational age,feeding pattern,invasive operation,antibiotic combination,glucocorticoid use,microecological preparation use,and hospital stay were independent influencing factors for neonatal pneumonia secondary diarrhea(P<0.05).The area under receiver operating characteristic curve(ROC)of model group was 0.779,with 95%CI of(0.719-0.838),sensitivity of 79.0%,and specificity of 63.8%.In the validation group,the area under ROC curve was 0.751,with 95%CI of(0.689-0.813),sensitivity of 85.2%,and specificity of 54.3%.The slope of the calibration curve of the model training group and the validation group was 1,and the intercept was 0.000.The model curve and the ideal model were basically synthesized diagonals.H-L goodness of fit test was optimal(P>0.05).The results of clinical effectiveness analysis showed that the net benefit of using this model to predict neonatal pneumonia secondary diarrhea was t

关 键 词：新生儿肺炎 继发性腹泻 多因素分析 列线图模型 

分 类 号：R722.1[医药卫生—儿科]