Delineating morbidity patterns in preterm infants at near-term age using a data-driven approach.

Area of Science:

• Neonatology
• Pediatric Critical Care
• Perinatal Medicine

Background:

• Premature birth survival is linked to cardio-respiratory and central nervous system morbidities.
• Existing research on infant morbidities is limited to pairwise associations, missing holistic co-occurrence patterns.
• Understanding comprehensive morbidity profiles is crucial for long-term outcomes in very preterm infants.

Purpose of the Study:

• To delineate and characterize comprehensive morbidity profiles in preterm infants at near-term age.
• To investigate prevalent morbidities including bronchopulmonary dysplasia (BPD), pulmonary hypertension (PH), cardiac defects, brain pathology, and retinopathy of prematurity (ROP).
• To identify shared morbidity patterns and risk profiles using a data-driven approach and machine learning.

Main Methods:

• Analysis of two independent prospective cohorts (AIRR and NEuroSIS) totaling 530 very preterm infants.
• Quantification of pairwise morbidity correlations and assessment of BPD's discriminatory power.
• Application of machine learning to identify infant subgroups with similar morbidity profiles.

Main Results:

• Bronchopulmonary dysplasia (BPD) and retinopathy of prematurity (ROP) showed the highest pairwise correlation, followed by BPD with PH and cardiac defects.
• BPD demonstrated limited capacity in discriminating overall morbidity occurrence.
• Machine learning identified distinct patient clusters with shared morbidity patterns (6 in AIRR, 8 in NEuroSIS).

Conclusions:

• The study provides a comprehensive characterization of preterm infant morbidity profiles at discharge, linked to shared pathophysiology.
• Identified morbidity patterns and patient subgroups can inform the development of personalized monitoring strategies.
• Future research should focus on refining risk profiles for tailored interventions in preterm populations.