A three-factor nomogram predicts the use of invasive mechanical ventilation within 72 h in preterm infants

Early identification of preterm infants at risk for invasive mechanical ventilation (IMV) enables timely respiratory support and may reduce ventilation-related harm.

To develop and internally validate a parsimonious prediction model for IMV within 72 h after birth.

We conducted a single-center retrospective cohort study (July 2023-June 2024) including 1,059 preterm infants admitted within 72 h of life and randomly split them into training (n = 742) and validation (n = 317) sets. Exclusions included chorioamnionitis and deaths ≤ 72 h. Forty-five candidate variables were screened; after multiple imputation, least absolute shrinkage and selection operator (20-fold cross-validation, λ_1se) identified three predictors for multivariable logistic modeling: 1-min Apgar score, pulmonary surfactant administration within 72 h, and early-onset sepsis. The primary endpoint was endotracheal IMV lasting ≥ 12 consecutive hours within 72 h of birth. Discrimination, calibration, and decision-curve analysis (DCA) were assessed. Sensitivity analysis restricted early-onset sepsis to culture-proven cases.

In the validation set, the model achieved an AUC of 0.816; at the optimal probability threshold (0.224), sensitivity, specificity, and accuracy were 0.613, 0.914, and 0.855, respectively. Calibration was good (Brier score 0.096; Hosmer-Lemeshow P = 0.28; expected/observed ratio = 1), and DCA showed net benefit across thresholds 0.10-0.70. Culture-proven analysis yielded AUC 0.830 with similar calibration; a pulmonary surfactant × sepsis interaction was significant (β = -2.531, P = 0.028).

A three-factor model based on perinatal and early neonatal indicators provides accurate, well-calibrated prediction of IMV within 72 h and is readily implementable for bedside risk stratification; external validation is warranted.