Factors affecting feeding ability in children with neonatal intensive care unit stay: a cluster analysis using...

Area of Science:

• Neonatal care
• Pediatric nutrition
• Medical informatics

Background:

• Oral feeding initiation is a complex process in Neonates Intensive Care Units (NICUs), often hindered by limited objective assessment tools.
• Critically ill neonates face particular challenges in establishing oral feeding.
• Understanding predictors of oral feeding success is crucial for optimizing care in the NICU.

Purpose of the Study:

• To identify and analyze oral feeding-related features in critically ill neonates admitted to the NICU.
• To determine the major predictors of successful oral feeding in this vulnerable population.
• To explore distinct profiles of oral feeding readiness.

Main Methods:

• A cohort of 1,419 critically ill neonates in the NICU was studied.
• The Preterm Infant Oral Feeding readiness assessment scale (PIOFRA) was used to gather data on oral feeding features.
• XGBoost models, utilizing Shapley Additive Explanation (SHAP) values, were developed for prediction, alongside ANOVA for group comparisons.

Main Results:

• Cluster analysis revealed three distinct profiles based on PIOFRA features.
• The XGBoost model demonstrated superior accuracy (85.2%) compared to other models like Logistic Regression and Random Forest.
• Sucking power (positive impact) and rooting reflex (negative impact) emerged as the most influential predictors for oral feeding outcomes.

Conclusions:

• Oral feeding difficulties are common in NICU neonates, necessitating detailed assessments, including gestational factors.
• PIOFRA features are robust predictors of achieving full oral feedings.
• Further research is required to elucidate the precise mechanisms by which sucking and rooting reflexes facilitate safe and efficient feeding in NICU neonates.