Interpretable machine learning-derived nomogram model for early detection of persistent diarrhea in Salmonella typhimurium enteritis: a propensity score matching based case-control study

Abstract

Background: Salmonella typhimurium infection is a considerable global health concern, particularly in children, where it often leads to persistent diarrhea. This condition can result in severe health complications including malnutrition and cognitive impairment. Methods: A comprehensive retrospective study was conducted involving 627 children diagnosed with Salmonella typhimurium enteritis. These children were hospitalized for Salmonella typhimurium enteritis between January 2010 and December 2022 at the Second Affiliated Hospital of Wenzhou Medical University. Propensity score matching was used to explore the potential risk factors and predictors of persistent diarrhea following S. typhimurium infection. Results: The study identified body temperature, C-reactive protein (CRP) levels, alanine aminotransferase (ALT) levels, white blood cell count, and lactose intolerance were significant predictors of persistent diarrhea. Nomogram models developed based on these predictors demonstrated robust performance in predicting persistent diarrhea risk, with an accuracy of > 90%. Conclusion: The developed nomogram models provide a practical tool for the early identification of children at high risk of persistent diarrhea, facilitating intervention, potentially preventing serious sequelae, and improving the prognosis of children with S. typhimurium enteritis.