Predicting serum phosphate levels in very preterm infants using machine learning
-
By
-
Åsbjørn S. Westvik
-
Oliver Tomic
-
Charlotte Tscherning
-
Sissel J. Moltu
-
July 15, 2026
Clinical Scorecard: Utilizing Machine Learning to Estimate Serum Phosphate Levels in Extremely Preterm Infants
At a Glance
| Category | Detail |
|---|
| Condition | Hypophosphatemia in very preterm infants |
| Key Mechanisms | High phosphate demands due to rapid postnatal growth and limited phosphate supply, particularly in SGA infants. |
| Target Population | Very preterm infants born <29 weeks' gestation, especially those who are small for gestational age. |
| Care Setting | Neonatal intensive care units managing preterm infants. |
Key Highlights
- 33.6% of infants developed hypophosphatemia in the first week.
- SGA infants have a higher and earlier risk of hypophosphatemia.
- Six predictors consistently identified for serum phosphate estimation.
- Elastic net models showed the best performance in predicting phosphate levels.
- Machine learning models can reduce unnecessary blood sampling.
Guideline-Based Recommendations
Diagnosis
- Diagnosis of hypophosphatemia is established through serum or plasma phosphate measurements.
Management
- Nutritional strategies must balance macro- and micronutrient provision to limit hypophosphatemia risk.
Monitoring & Follow-up
- Routine blood gas analysis can provide calcium and potassium values, aiding in phosphate estimation.
Risks
- Frequent blood sampling can lead to iatrogenic anemia and stress in preterm infants.
Patient & Prescribing Data
Infants born <29 weeks' gestation managed under a standardized nutritional protocol.
Nutritional regimens must accommodate high phosphate demands, especially in SGA infants.
Clinical Best Practices
- Utilize machine learning models for estimating serum phosphate to minimize blood sampling.
- Implement standardized nutritional protocols to optimize nutrient intake in very preterm infants.
Related Resources & Content
