Routine 36-week scan: prediction of prolonged neonatal intensive care unit admission

Area of Science:

• Maternal-fetal medicine
• Neonatal intensive care
• Predictive diagnostics

Background:

• Predicting prolonged neonatal intensive care unit (NICU) admissions is crucial for resource allocation and optimizing neonatal care.
• Maternal demographic characteristics and medical history are known risk factors, but their predictive power for prolonged NICU stays requires further investigation.
• The added value of third-trimester ultrasound findings and labor/delivery data to existing predictive models is not fully established.

Purpose of the Study:

• To assess the predictive contribution of maternal factors to prolonged NICU admissions.
• To evaluate the incremental value of 36-week ultrasound scan findings in predicting prolonged NICU admission.
• To determine the combined predictive performance of maternal characteristics, ultrasound data, and labor/delivery information for prolonged NICU stays.

Main Methods:

• A cohort of 107,762 women with singleton pregnancies underwent assessment including medical history, estimated fetal weight (EFW), and Doppler measurements (uterine artery PI, umbilical artery PI, fetal middle cerebral artery PI) at 35-36 weeks' gestation.
• Multivariable logistic regression models were employed to analyze the independent contributions of maternal factors, 36-week scan findings, and labor/delivery data to prolonged NICU admission.
• Detection rates (DRs) and areas under the receiver-operating-characteristics curve (AUCs) were calculated to compare the predictive performance of different models.

Main Results:

• Maternal factors like BMI, social deprivation, IVF conception, chronic hypertension, diabetes, and pre-eclampsia significantly predicted prolonged NICU admission (AUC 0.606).
• 36-week scan findings, including amniotic fluid indices and umbilical/middle cerebral artery PI, improved prediction (AUC 0.637 with maternal factors).
• Incorporating labor and delivery data (gestational age at delivery, mode of delivery, birth weight percentile) further enhanced prediction (AUC 0.709), reaching 57.2% DR at 10% FPR for non-intrapartum events (AUC 0.790).

Conclusions:

• Maternal factors, 36-week ultrasound findings, and labor/delivery data independently contribute to predicting prolonged NICU admission.
• Despite significant predictive contributions, over 40% of prolonged NICU admissions at term remain unpredictable prior to labor and birth.
• These findings highlight the multifactorial nature of NICU admissions and the limitations of antepartum prediction.