Clinical Report: Utilizing Machine Learning to Estimate Serum Phosphate Levels in Extremely Preterm Infants

Overview

This study investigates the use of machine learning (ML) to estimate serum phosphate levels in extremely preterm infants. It highlights the significant risk of hypophosphatemia, particularly in small for gestational age (SGA) infants, and demonstrates that ML models can moderately predict phosphate levels using routine clinical data.

Background

Hypophosphatemia is a common condition in very preterm infants, associated with serious complications such as sepsis and neurodisability. The condition is particularly prevalent in infants born before 28 weeks of gestation and those classified as small for gestational age (SGA). Accurate monitoring and management of phosphate levels are crucial for improving outcomes in this vulnerable population.

Data Highlights

Key Findings

• 33.6% of infants developed hypophosphatemia in the first week.
• SGA infants had a significantly higher risk of hypophosphatemia (log-rank p < 0.0001).
• Six predictors were consistently retained across all ML prediction tasks.
• Elastic net models performed best in estimating serum phosphate levels.
• ML models can predict phosphate levels 12-24 hours in advance.

Clinical Implications

Further prospective validation is necessary before clinical implementation.

Conclusion

Continued research is essential to validate these models for clinical use.

Related Resources & Content

1. BMJ Paediatrics Open, 2023 -- Optimising transfusion practices in preterm neonates using gestational age and birth weight-based prediction model: a retrospective cohort study from India
2. BMC Pregnancy and Childbirth, 2026 -- Prediction of preterm and low birth weight risk using a physiology based artificial neural network integrating hematological, dental, and periodontal index markers: a cross sectional study based on machine learning
3. Critical Care, 2026 -- Hypophosphatemia as a biomarker of metabolic intolerance to enhanced nutrition in the PICU: a secondary analysis of the PEPaNIC RCT
4. Frontiers in Medicine, 2026 -- Machine learning model for predicting hypotension following continuous renal replacement therapy initiation in end-stage kidney disease patients: A SHAP-interpretable approach
5. NICE, 2023 -- Recommendations | Neonatal parenteral nutrition | Guidance
6. Early hypophosphataemia and refeeding syndrome in extremely low birthweight babies and outcomes to 2 years of age: secondary cohort analysis from the ProVIDe trial, 2025
7. Frontiers, 2026 -- Predicting serum phosphate levels in very preterm infants using machine learning
8. Recommendations | Neonatal parenteral nutrition | Guidance | NICE
9. Early hypophosphataemia and refeeding syndrome in extremely low birthweight babies and outcomes to 2 years of age: secondary cohort analysis from the ProVIDe trial
10. Frontiers | Predicting serum phosphate levels in very preterm infants using machine learning