超早产儿低Apgar评分风险预测模型的构建与验证

Development and validation of a risk prediction model for low Apgar scores in extremely premature infants

目的建立超早产儿生后发生低Apgar评分的早期临床预测模型。方法回顾性分析2017年1月至2021年12月北京妇产医院分娩的超早产儿临床资料,按照7∶3的比例随机拆分为训练集和验证集。以生后是否发生低Apgar评分为结局变量,选取17个临床指标为预测变量。在训练集内应用Lasso回归和多因素logistic回归筛选最终预测因子进入最终模型,并应用验证集对最终模型进行校准度、区分度和临床决策曲线的评价。结果共纳入169例超早产儿,训练集117例,验证集52例。训练集经过Lasso回归和多因素logistic回归筛选后,将性别、胎儿宫内窘迫、受孕方式及分娩时间4个指标纳入最终模型。训练集与验证集均有较好的校准曲线,预测模型受试者工作特征曲线下面积为0.731,灵敏度为72.2%,特异度为60.5%,外部验证曲线下面积为0.704。临床决策曲线显示该模型在2%~75%的阈值范围内对判断超早产儿生后发生低Apgar评分的受益更大。结论本研究构建的临床预测模型具有较好的区分度、校准度和临床适用度,可作为预测超早产儿生后低Apgar评分的参考工具。

Objective To establish a risk prediction model for the occurrence of low 1 min Apgar scores in extremely premature infants(EPIs).Methods From January 2017 to December 2021,EPIs delivered at our hospital were retrospectively analyzed and randomly assigned into training set group and validation set group in a 7∶3 ratio.17 clinical indicators were selected as predictive variables and low Apgar scores after birth as outcome variables.Lasso regression and multi-factor logistic regression were used within the training set group to select the final predictors for the final model,and the calibration,distinguishability and clinical decision making curves of the final model were evaluated in the validation set group.Results A total of 169 EPIs were enrolled,including 117 in the training set group and 52 in the validation set group.4 indicators including gender,fetal distress,assisted conception and delivery time were selected as the final predictors in the final model.Both the training set group and the validation set group had good calibration curves.The area under the receiver operating characteristic curve(AUC)of the prediction model was 0.731,the sensitivity was 72.2%,the specificity was 60.5%and the AUC of the external validation curve was 0.704.The clinical decision making curve showed that the model had a greater benefit in predicting the occurrence of low Apgar score in EPIs within the threshold of 2%to 75%.Conclusions The clinical prediction model established in this study has good distinguishability,calibration and clinical accessibility and can be used as a reference tool to predict low Apgar scores in EPIs.