Fig 1.
Incidence of pediatric brucellosis by year (a) and seasonal distribution (b) in Yunnan Province, China, 2015–2024.
Fig 2.
Geographic distribution of pediatric brucellosis in Yunnan Province, China, 2015–2024: a. distribution at the municipal level; b. distribution at the district level.
Table 1.
Demographic, clinical, and laboratory characteristics of pediatric brucellosis patients in Yunnan Province, China, 2015–2024.
Fig 3.
Common symptoms (A) and signs (B) of pediatric brucellosis in Yunnan Province, China, 2015–2024.
Fig 4.
Positive differences among four types of body fluid cultures in pediatric brucellosis in Yunnan Province, China, 2015–2024.
Fig 5.
Severe pediatric brucellosis prediction model in Yunnan Province, China, 2015–2024.
(a) ROC curve analysis, evaluating the AUC values of continuous variables to predict the risk of severe brucellosis. (b) Variable selection using the BORUTA algorithm to identify key variables associated with the prediction of severe disease, with the selected variables marked by the red dashed lines. (c-h) Comparison of AUC values for six prediction models (logistic regression, KNN, naive Bayes, MLP, XGBoost, and random forest), with the random forest model demonstrating the highest AUC value (AUC = 0.970), indicating optimal performance. (i) Decision curve analysis (DCA) to assess the net benefit of the six models at different threshold probabilities, with the random forest model showing the greatest net benefit. (j) A web-based calculator developed using ‘Shiny’ functions, designed to predict the risk of severe brucellosis based on clinical variables, providing real-time prediction results and the importance evaluation of each variable to assist clinical decision-making.