Determining an Optimal Oxygen Saturation Target Range Based on Neonatal Maturity: Demonstration of a Decision Tree...

Area of Science:

• Neonatal intensive care
• Machine learning applications in medicine
• Pediatric respiratory monitoring

Background:

• Optimizing oxygen saturation (SpO2) targets in neonates is critical for preventing complications.
• Neonatal maturity, indicated by postmenstrual age (PMA), influences the risk associated with SpO2 levels.
• Understanding the interplay between SpO2 targets, PMA, and oxygenation risk is essential for clinical decision-making.

Purpose of the Study:

• To utilize decision tree machine learning to explore interactions between SpO2 target ranges, neonatal maturity, and the risk of extreme arterial oxygen levels (oxemic risk).
• To identify optimal SpO2 target ranges that minimize oxemic risk across different postmenstrual ages (PMA) in a neonatal intensive care unit (NICU) population.

Main Methods:

• An observational study analyzing 7500 paired SpO2 and PaO2 data points from a NICU over 3 years.
• The Classification and Regression Tree (CHAID) decision tree method was employed to assess risk across six SpO2 target ranges (88-97%).
• Oxemic risk was quantified using a severity-weighted average of arterial oxygen levels outside the neonate normal range (50-80 mmHg).

Main Results:

• The lowest SpO2 target ranges (e.g., 88-92%) were associated with the highest oxemic risk.
• SpO2 ranges of 91-95% and 92-96% demonstrated the lowest associated risks.
• Optimal SpO2 targets varied by PMA, with all ranges showing lowest risk at ≥42 weeks PMA, and preterm infants (≤34 weeks PMA) benefiting most from a 92-96% target range.

Conclusions:

• Decision tree analytics effectively demonstrate the utility in analyzing complex neonatal data.
• SpO2 target ranges deviating from typical values may significantly reduce neonatal morbidity and mortality.
• Further investigation through prospective randomized trials is recommended to validate these findings and guide clinical practice.