Construction and efficacy evaluation of a model for early diagnosis of pediatric sepsis based on LASSO-logistic regression.

Objective: The aim of this study was to analyse the clinical characteristics and related risk factors of Pediatric Sepsis, construct a column-line diagram model to predict the likelihood of Pediatric Sepsis, and validate the model to facilitate primary care paediatricians to quickly and quantitatively assess the risk of Pediatric Sepsis. Methods: This single-center retrospective study included children hospitalized for infections at Gansu Provincial Maternity and Child-Care Hospital from January 2018 to June 2024. Data on 39 variables covering baseline characteristics, vital signs, and laboratory indicators were collected. The samples were randomized into training and validation groups in a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used for initial data screening and dimensionality reduction, followed by Logistic regression to identify independent risk factors for sepsis. Predictive modeling was then performed. The performance of the column-line plots was internally validated using ROC curves, calibration curves, and decision curve analysis (DCA). Results: The development dataset included 834 patients with severe infections, of whom 212 (25.4%) developed sepsis. Seven predictors were identified as independent risk factors: respiratory rate, temperature, immature granulocyte percentage, platelets, procalcitonin, fibrinogen, and lactic acid. A predictive column-line diagram was created using these predictors. Internal validation showed that the column-line diagrams had good discriminatory ability, calibration, and clinical applicability. Conclusion: A column-line diagram was successfully developed to predict the incidence of sepsis in children using seven commonly used clinical and laboratory indicators. The model demonstrated good performance and clinical validity through internal validation.