Supervised Machine Learning Models Predicting Postoperative Low Cardiac Output Syndrome In Neonates.

Area of Science:

• Cardiovascular Surgery
• Neonatal Intensive Care
• Machine Learning in Medicine

Background:

• Low cardiac output syndrome (LCOS) is a critical complication in neonates post-cardiothoracic surgery.
• Early prediction of LCOS is essential for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and validate supervised machine learning (ML) models for predicting LCOS in neonates within 48 hours of cardiothoracic surgery.
• To identify key clinical and laboratory variables that predict LCOS development.

Main Methods:

• A retrospective observational study involving 181 neonates undergoing cardiothoracic surgery.
• Development of LightGBM ML models using hourly clinical and laboratory data from the first 48 postoperative hours.
• SHapley Additive exPlanations (SHAP) analysis for feature importance assessment.

Main Results:

• The ML models achieved high predictive performance, with AUC values ranging from 0.91 to 0.98.
• Key predictors for LCOS included higher vasoactive inotrope score, lower urine output, and higher serum lactate.
• 14.9% of neonates in the study experienced LCOS.

Conclusions:

• Supervised ML models can accurately predict LCOS in neonates, providing high interpretability.
• Findings support the integration of these models into clinical workflows for enhanced postoperative care.
• Further multicenter validation is recommended.