摘要
目的探讨影响新生儿早产的主要风险因素。方法选取2020年1月至2022年12月期间在南宁市武鸣区妇幼保健院进行妊娠管理并早产的900例产妇作为研究对象。采用随机森林算法构建新生儿早产风险评估模型,数据集按3∶1比例分为训练集和测试集。通过单因素和多因素Logistic回归分析早产的影响因素,并使用受试者工作特征(ROC)曲线、校准曲线和C指数对模型的准确度和稳定性进行评估。结果900例产妇中,新生儿早产风险评估模型共识别有早产风险孕妇81例(占9.00%),无早产风险孕妇819例(占91.00%),分别纳入有早产风险组和无早产风险组。两组年龄、身高、孕前体重指数(BMI)及孕期BMI比较,差异无统计学意义(P>0.05);有早产风险组的产检次数少于无早产风险组,两组多胎妊娠、早产史、子宫机能不全、妊娠高血压疾病、死胎史、前置胎盘和霉菌感染比较,差异均有统计学意义(P<0.05)。有早产风险组中,有54例早产,无早产风险组中均正常分娩;随机森林算法构建的新生儿早产风险评估模型评估新生儿早产风险的灵敏度为100.00%(54/54),特异度为96.81%(819/846),准确度为97.00%(873/900)。多因素Logistic回归分析结果显示,产检次数增加是孕妇发生早产的保护因素,多胎妊娠、早产史、妊娠期高血压疾病、死胎史、子宫颈功能不全、前置胎盘和霉菌感染为早产的独立危险因素。结论基于随机森林算法构建新生儿早产风险评估模型在预测新生儿早产风险方面表现出较高的准确度、灵敏度和特异度,能够有效识别早产高风险人群,有助于临床防控早产。
Objective To explore the main risk factors affecting neonatal preterm birth.Methods A total of 900 pregnant women who underwent pregnancy management and gave birth prematurely at Wuming District Maternal and Child Health Hospital in Nanning City from January 2020 to December 2022 were selected as the research objects.The risk assessment model of neonatal preterm birth was constructed by random forest algorithm.The data set was divided into training set and test set according to the ratio of 3:1.The influencing factors of preterm birth were analyzed by univariate and multivariate logistic regression,and the accuracy and stability of the model were evaluated by receiver operating characteristic(ROC)curve,calibration curve and C index.Results Among 900 pregnant women,the neonatal preterm risk assessment model identified 81 cases(9.00%)at risk of preterm birth and 819 cases(91.00%)without risk,which were respectively categorized into the preterm risk group and the non-preterm risk group.No statistically significant differences were observed in age,height,pre-pregnancy body mass index(BMI)and pregnancy BMI between the two groups(P>0.05).The preterm risk group required fewer prenatal check-ups than the non-preterm risk group.However,statistically significant differences were found in multiple pregnancies,preterm history,uterine insufficiency,gestational hypertension,stillbirth history,placenta previa,and fungal infections(P<0.05).There were 54 cases of preterm birth in the preterm risk group,while all in the non-preterm risk group delivered normally.The sensitivity,specificity and accuracy of the neonatal preterm birth risk assessment model constructed by random forest algorithm were 100.00%(54/54),96.81%(819/846)and 97.00%(873/900),respectively.The results of multivariate logistic regression analysis showed that an increase in the number of prenatal check-ups was a protective factor for preterm birth in pregnant women,while multiple pregnancies,history of preterm birth,hypertensive disorder complicating pregnancy,history of stillbirth,cervical insufficiency,placenta previa and fungal infection were significant risk factors for preterm birth.Conclusion The risk assessment model of neonatal preterm birth based on random forest algorithm shows high accuracy,sensitivity and specificity in predicting the risk of neonatal preterm birth,which can effectively identify the high-risk population of premature birth and contribute to clinical prevention and control of premature birth.
作者
潘喜英
谭小利
梁芬
黄丹丹
PAN Xiying;TAN Xiaoli;LIANG Fen(Department of Obstetrics,Wuming District Maternal and Child Health Hospital,Nanning,Guangxi 530199,China)
出处
《医师在线》
2026年第1期28-32,共5页
Journal of Doctors Online
关键词
随机森林算法
新生儿
早产风险
模型构建
Random forest algorithm
Neonates
Preterm birth risk
Model construction
