Clinical Report: An Innovative Nutritional Assessment Method for Detecting Infants at Risk of Stunting
Overview
This study developed a novel scoring system and prediction algorithm to identify infants at risk of stunting using growth, nutritional, and biochemical indicators. The Gradient Boosting model demonstrated robust performance with an AUC of 0.861 in the training set and 0.850 in the validation set.
Background
Infant stunting is a significant public health challenge, impacting physical and cognitive development. Traditional assessments often rely on single measurements, which may delay early identification and intervention. This study addresses the need for a multidimensional tool that integrates various indicators to assess stunting risk effectively.
Data Highlights
Indicator
OR (95% CI)
P-value
Length growth velocity
0.340 (0.211–0.548)
<0.001
Gradient Boosting AUC (Training Set)
0.861 (0.784–0.938)
Gradient Boosting AUC (Validation Set)
0.850 (0.699–1.000)
Key Findings
Five indicators significantly associated with stunting were identified: infant weight Z-score, length Z-score, length growth velocity, diversity of complementary foods, and hemoglobin.
Length growth velocity was the strongest predictor of stunting.
LASSO regression confirmed the optimal variable combination for predicting stunting risk.
The Gradient Boosting model outperformed other machine learning models in predicting stunting risk.
SHAP analysis indicated that infant weight Z-score was the most critical predictive variable.
A visual nomogram was developed for practical risk assessment.
Clinical Implications
The developed scoring system and prediction algorithm can facilitate early identification of infants at risk of stunting, allowing for timely clinical interventions. This multidimensional approach may enhance routine nutritional assessments in clinical practice.
Conclusion
The nutritional scoring system and Gradient Boosting algorithm provide a robust framework for identifying infants at risk of stunting, supporting early intervention strategies.
