Cao Xincheng, Li Shujuan, Gu Xinyue, Chen Huiyao, Yang Chuanzhong, Qian Miao, Tian Xiuying, Xu Falin, Yang Zuming, Wang Yang, Guo Jinzhen, Lee Shoo K, Jiang Siyuan, Cao Yun
Department of Neonatology, Children's Hospital of Fudan University, National Children's Medical Center, Shanghai, China.
National Health Commission Key Laboratory of Neonatal Diseases, Fudan University, Shanghai, China.
Transl Pediatr. 2025 May 30;14(5):927-938. doi: 10.21037/tp-2025-33. Epub 2025 May 21.
Accurate prediction of outcomes for extremely preterm infants (EPIs) during the early stage is important to assist clinicians and parents in making decisions. This study aimed to develop and validate models for predicting mortality and survival without major morbidity for EPIs using information available on neonatal intensive care units (NICUs) admission.
Two of the largest contemporary cohorts of EPIs born at 24-28 weeks' gestation were included in China. Two predictive models were generated separately to predict mortality and survival without major morbidity at discharge. Potential predictors were identified if they had a well-established association with neonatal outcomes in literatures and could be easily obtained on NICU admission, including gestational age, birth weight, sex, inborn, antenatal steroids, 5-min Apgar score, and invasive ventilation on admission. Logistic regression was employed to develop the models. Model performance was assessed via area under the curve (AUC).
Among 2,438 EPIs in the development cohort, the mortality rate was 17.7% (431/2,438) and the rate of survival without major morbidity was 52.5% (1,281/2,438). Among the 5,045 infants in the validation cohort, 9.2% (463/5,045) died, and 59.1% (2,981/5,045) survived without major morbidity. Gestational age, birth weight, invasive ventilation on NICU admission, antenatal steroids use, and 5-min Apgar score were selected as predictors in the mortality model, yielding the AUC of 0.77 [95% confidence interval (CI): 0.75-0.79]. For the survival without major morbidity model, predictors were gestational age, birth weight, invasive ventilation on NICU admission, sex, and 5-min Apgar score, and the AUC was 0.72 (95% CI: 0.70-0.74). The validation cohort resulted in AUCs of 0.76 (95% CI: 0.73-0.78) and 0.70 (95% CI: 0.68-0.71) for the mortality and survival without major morbidity models, respectively.
Using commonly available predictors on NICU admission including gestational age, birth weight, invasive ventilation on NICU admission, antenatal steroids use, sex, and 5-min Apgar score, we successfully developed and validated two distinct models with acceptable performance, predicting mortality and survival without major morbidity for EPIs.
准确预测极早产儿（EPI）早期的预后对于帮助临床医生和家长做出决策非常重要。本研究旨在利用新生儿重症监护病房（NICU）入院时可获得的信息，开发并验证预测EPI死亡率和无严重并发症存活的模型。
纳入中国当代两个最大的孕24 - 28周出生的EPI队列。分别生成两个预测模型，以预测出院时的死亡率和无严重并发症存活情况。如果潜在预测因素在文献中与新生儿预后有明确关联且在NICU入院时易于获取，则将其纳入，包括胎龄、出生体重、性别、是否为足月儿、产前使用类固醇、5分钟阿氏评分以及入院时的有创通气情况。采用逻辑回归建立模型。通过曲线下面积（AUC）评估模型性能。
在开发队列的2438例EPI中，死亡率为17.7%（431/2438），无严重并发症存活的比例为52.5%（1281/2438）。在验证队列的5045例婴儿中，9.2%（463/5045）死亡，59.1%（2981/5045）无严重并发症存活。胎龄、出生体重、NICU入院时的有创通气、产前使用类固醇以及5分钟阿氏评分被选为死亡率模型的预测因素，AUC为0.77 [95%置信区间（CI）：0.75 - 0.79]。对于无严重并发症存活模型，预测因素为胎龄、出生体重、NICU入院时的有创通气、性别以及5分钟阿氏评分，AUC为0.72（95% CI：0.70 - 0.74）。验证队列中，死亡率和无严重并发症存活模型的AUC分别为0.76（95% CI：0.73 - 0.78）和0.70（95% CI：0.68 - 0.71）。
利用NICU入院时常用的预测因素，包括胎龄、出生体重、NICU入院时的有创通气、产前使用类固醇、性别以及5分钟阿氏评分，我们成功开发并验证了两个性能可接受的不同模型，用于预测EPI的死亡率和无严重并发症存活情况。
