Development and validation of a model predicting preterm infant discharge in level 2 care

Area of Science:

• Neonatal Medicine
• Clinical Prediction Models
• Health Services Research

Background:

• Predicting discharge for preterm infants in level 2 neonatal wards is crucial for resource management and family communication.
• Existing models often focus on total length of stay, with limited prediction of imminent discharge.
• The level 2 neonatal population is underreported in predictive modeling studies.

Purpose of the Study:

• To develop and validate a predictive model for upcoming hospital discharge of preterm infants.
• To identify key clinical predictors associated with discharge within one week.
• To provide a tool for optimizing care and resource allocation in neonatal settings.

Main Methods:

• Retrospective cohort study of preterm infants admitted between January 2016 and December 2023.
• Development of a multivariable logistic regression model using backward selection.
• Prediction time set at day 7 of admission; primary outcome was discharge between days 7 and 14.

Main Results:

• The model included 1083 infants, with nine significant predictors identified (e.g., mode of delivery, gestational age, feeding method).
• Internal and external validation demonstrated excellent discrimination (AUC 0.93) and acceptable calibration.
• A probability threshold of 0.60 achieved 88% sensitivity and 89% specificity for predicting discharge.

Conclusions:

• A combination of perinatal and neonatal characteristics effectively predicts discharge for preterm infants in level 2 settings.
• The developed tool shows high accuracy and can aid in parent-provider communication and hospital capacity management.
• Further validation in prospective studies is recommended to confirm the tool's utility.