Development and multicenter validation of an explainable machine learning diagnostic criteria for pediatric abdominal sepsis - PubMed
12 hours ago
- #diagnostic criteria
- #machine learning
- #pediatric sepsis
- Development of an explainable machine learning model (ABSeD) for early diagnosis of pediatric abdominal sepsis (PAS).
- Retrospective data from 6566 pediatric patients used for model construction, with prospective data from 308 patients across seven hospitals for external validation.
- ABSeD integrates nine routine clinical variables, showing high diagnostic accuracy (AUC = 0.934 in training, 0.928 in validation) and robust multicenter generalizability.
- The model aims to address the gap in existing pediatric sepsis criteria by focusing on early intra-abdominal infections.
- Potential to enhance timely intervention for hospitalized children with suspected or clinically identified intra-abdominal septic pathology.