超早产儿早期死亡风险列线图预测模型的建立

Establishment of a nomogram prediction model for early mortality risk in extremely preterm infants

目的探讨超早产儿早期(生后7 d内)死亡的危险因素,并构建早期死亡风险预测模型。方法回顾性选择2017年1月至2022年12月于济宁医学院附属医院产科分娩的胎龄22+0~27周+6的244例超早产儿,根据生后存活是否≥7 d分为早期存活组(n=140)和早期死亡组(n=84)。采用LASSO、logistic回归筛选超早产儿早期死亡的危险因素。应用R软件绘制列线图预测模型。使用拟合优度检验、受试者工作特征曲线下面积(area under the curve,AUC)、校准曲线、决策曲线评估模型的效能和临床实用性。结果LASSO回归、多因素logistic回归分析显示臀位分娩(OR=3.055,95%CI:1.125~8.296)、产房内气管插管(OR=4.320,95%CI:1.328~14.053)、产房内应用肾上腺素(OR=10.706,95%CI:1.454~78.816)、生后6 h内诊断Ⅲ~Ⅳ级新生儿呼吸窘迫综合征(OR=11.552,95%CI:3.056~43.677)是超早产儿早期死亡的危险因素,而出生胎龄大(OR=0.234,95%CI:0.125~0.436)、产前使用糖皮质激素促胎肺成熟(OR=0.046,95%CI:0.014~0.145)和生后6 h内应用肺表面活性物质(OR=0.021,95%CI:0.004~0.122)是超早产儿死亡的保护因素。基于上述7个因素建立的列线图模型拟合优度检验表明拟合良好(P=0.702),预测超早产儿早期死亡的AUC为0.963(95%CI:0.943~0.983),灵敏度为0.904(95%CI:0.806~0.949),特异度为0.892(95%CI:0.829~0.938),准确性为0.880。决策曲线显示阈值概率>2%将产生总体净获益。结论臀位分娩、产房内气管插管、产房内应用肾上腺素、生后6 h内诊断Ⅲ~Ⅳ级新生儿呼吸窘迫综合征为超早产儿早期死亡的独立危险因素,出生胎龄大、孕母产前使用糖皮质激素促胎肺成熟及生后6 h内使用肺表面活性物质为超早产儿早期死亡的保护因素,基于上述因素构建的预测模型能定量、简便、直观地评估超早产儿早期死亡的风险。

Objective To identify the risk factors and to construct a predictive model for early postnatal mortality(with the first 7 days of life)in extremely preterm infants.Methods This retrospective study involved 244 extremely preterm infants with a gestational age of 22 to 27 weeks and 6 days,born at the Affiliated Hospital of Jining Medical College from January 2017 to December 2022.They were divided into an early survival group(n=140)and an early mortality group(n=84),based on survival for≥7 days after birth.LASSO and logistic regression were used to select risk factors for early mortality.A nomogram predictive model was constructed using the R software program.The goodness-of-fit tests,area under the curve(AUC),calibration curves,and decision curves were used to evaluate its performance and clinical usefulness.Results LASSO regression and multivariate logistic regression analyses showed that breech delivery(OR=3.055,95%CI:1.125-8.296),intubation in the delivery room(OR=4.320,95%CI:1.328-14.053),diagnosis of gradeⅢ-Ⅳneonatal respiratory distress syndrome within 6 h after birth(OR=11.552,95%CI:3.056-43.677),and use of adrenaline in the delivery room(OR=10.706,95%CI:1.454-78.816)were risk factors for early mortality in extremely preterm infants.Conversely,large gestation age(OR=0.234,95%CI:0.125-0.436),antenatal administration of corticosteroids to promote fetal lung maturity(OR=0.046,95%CI:0.014-0.145),and the use of pulmonary surfactant within 6 h after birth(OR=0.021,95%CI:0.004-0.122)were protective factors against mortality.The goodness of fit test of the early death risk nomogram prediction model for extremely preterm infants indicates a good fit(P=0.702).The AUC of the model was 0.963(95%CI:0.943-0.983),with a sensitivity of 0.904(95%CI:0.806-0.949),specificity of 0.892(95%CI:0.829-0.938),and accuracy of 0.880.Decision curve analysis indicated that a threshold probability>2%would yield a net benefit.Conclusions Breech delivery,intubation in the delivery room,use of adrenaline in the delivery room,and the diagnosis of gradeⅢ-Ⅳneonatal respiratory distress syndrome within 6 h post-birth are independent risk factors for early mortality in extremely preterm infants.Large gestational age,antenatal administration of corticosteroids to promote fetal lung maturity and use of pulmonary surfactant within 6 h after birth are protective factors.The constructed prediction model based on the aforementioned factors can quantitatively,conveniently,and intuitively assess the risk of early mortality in extremely preterm infants.