Enhanced predictive accuracy of mortality in VLBW infants with late-onset sepsis through a time-specific nomogram

This is the authors' abstract. We don't add key points for this paper.

Objective: This study aims to develop and validate a nomogram-based scoring system to predict mortality in very low birth weight (VLBW) infants with late-onset sepsis (LOS). Timely risk stratification in this vulnerable population is critical for optimizing clinical outcomes. Methods: We conducted a retrospective analysis on 202 VLBW infants diagnosed with LOS between January 2018 and December 2022. Predictive models were created at three key time points: 0 h, 6 h, and 12 h post-sepsis onset, utilizing Least Absolute Shrinkage and Selection Operator (LASSO) regression for variable selection and multivariable logistic regression for model construction. Internal validation was performed with 1,000 bootstrap resamples to correct for potential overfitting. External validation was conducted on an independent cohort of 71 infants from January 2023 to March 2024. Model performance was assessed using Area Under the Curve (AUC), calibration plots, and decision curve analysis (DCA). Results: The models exhibited excellent discrimination with AUCs of 0.83, 0.92, and 0.94 at 0 h, 6 h, and 12 h, respectively, in the development cohort, and 0.95, 0.95, and 0.97 in the validation cohort. Calibration plots showed strong agreement between predicted and observed outcomes. The significant disparity in maternal COVID-19 infection rates between cohorts (1 vs. 89%) may have contributed to the enhanced predictive accuracy in the external cohort. Conclusion: This dynamic, time-specific nomogram demonstrates high predictive accuracy and clinical utility for mortality in VLBW infants with LOS. The impact of maternal COVID-19 infection on neonatal outcomes offers a novel perspective for future research in sepsis prognostication.