Constructing a predictive model for early-onset sepsis in neonatal intensive care unit newborns based on SHapley Additive exPlanations explainable machine learning

Description

Xuefeng Tan and eight colleagues build an explainable machine learning model to predict early-onset sepsis in newborn intensive care patients, in a study from twenty twenty-four. Neonatal sepsis shows subtle, non-specific signs yet carries high mortality. Early-onset cases strike within seventy-two hours of birth. A retrospective cohort from a Chinese hospital supplied the data.

After exclusions, four hundred thirty newborns were split into training and test sets. A shrinkage regression selected salient features and guarded against overfitting. Six machine learning models were trained to forecast early-onset sepsis. Their predictive performance was rigorously evaluated against each other. Shapley additive explanations made the winning model's reasoning interpretable. The authors identify pivotal risk factors clinicians can watch. Interpretable prediction, they argue, supports earlier diagnosis in fragile newborns.

Catalog record

Source: https://europepmc.org/articles/PMC11621883