Machine Learning Improves the Predictive Utility of Lactic Acid in Hospitalized Infants.

Area of Science:

• Biomedical Informatics
• Neonatology
• Metabolic Disorders

Background:

• Hyperlactatemia is a frequent condition in hospitalized infants.
• Identifying the cause of hyperlactatemia is crucial for appropriate treatment and prognosis.
• Inborn errors of energy metabolism (IEEM) are a critical differential diagnosis in neonatal hyperlactatemia.

Purpose of the Study:

• To develop and validate a machine learning model for predicting IEEM in hospitalized infants with hyperlactatemia.
• To compare the diagnostic utility of machine learning models with lactate levels alone.

Main Methods:

• Retrospective cohort study of infants aged 0-90 days with lactate levels >= 5 mmol/L.
• Machine learning models (Random Forest, XGBoost) were trained and validated using clinical and laboratory data, including plasma amino acid and acylcarnitine levels.
• Model performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC).

Main Results:

• Machine learning models achieved an AUC-ROC of 0.81 in predicting IEEM, significantly outperforming lactate alone (AUC-ROC 0.56).
• Infants with IEEM had significantly higher lactate levels (median 12.6 mmol/L) compared to other causes.
• Mortality was high overall (30%) and particularly elevated in infants with IEEM (51%).

Conclusions:

• Machine learning demonstrates high diagnostic utility for identifying IEEM in neonatal hyperlactatemia.
• This approach facilitates computer-aided interpretation of complex data, enabling faster and more accurate diagnoses.
• Early and accurate diagnosis of IEEM is vital for improving outcomes in infants with hyperlactatemia.