Development and validation of a risk prediction model for early-onset hypoglycemia in preterm infants

BACKGROUND: Early-onset hypoglycemia is a frequent metabolic complication in preterm infants and may lead to adverse neurodevelopmental outcomes. Accurate risk prediction is essential for timely clinical intervention. METHODS: = 131). Univariate analysis and LASSO logistic regression were used to screen predictive variables. A multivariate logistic regression model was developed and visualized as a nomogram. Internal validation was performed using calibration plots, ROC curves, and decision curve analysis (DCA). RESULTS: A retrospective cohort of 436 preterm infants was analyzed, among whom 124 cases (28.44%) experienced hypoglycemia within 48 h after birth. Multivariate logistic regression identified six independent risk factors: reduced gestational age, multiple births, cesarean delivery, maternal gestational diabetes, gestational hypertension, and abdominal distension observed on the second postnatal day. The predictive model exhibited solid discriminative power, with an AUC of 0.802 in the training group and 0.829 in the validation group. Model calibration was satisfactory across datasets. DCA further supported the model's clinical utility, indicating consistent net benefit over a wide spectrum of risk thresholds. CONCLUSION: The model demonstrated promising predictive ability for early-onset hypoglycemia in preterm infants and could potentially serve as a preliminary tool to inform risk stratification strategies. However, its clinical translation requires confirmation through external validation in prospective, multicenter studies before any consideration of widespread implementation.