Abstract
OBJECTIVES: To establish a predictive model for severe Mycoplasma pneumoniae pneumonia (SMPP) in children younger than 5 years. METHODS: Clinical data of 504 children younger than 5 years with Mycoplasma pneumoniae pneumonia admitted to Xiaogan Hospital of Wuhan University of Science and Technology from January to December 2023 were retrospectively analyzed. Based on discharge diagnosis, patients were classified into a non-SMPP group (n=345) and an SMPP group (n=159). Univariate analysis and LASSO regression were used to screen predictors of SMPP. The selected variables were then entered into a multivariable logistic regression to construct the prediction model, and its performance was evaluated. RESULTS: Multivariable logistic regression identified lung imaging findings (proportion with consolidation), duration of fever, high-sensitivity C-reactive protein, lactate dehydrogenase, creatine kinase, and lymphocyte-to-neutrophil ratio as predictors of SMPP (P<0.05). The model based on these six indicators achieved an area under the receiver operating characteristic curve of 0.862 (95%CI: 0.824-0.900), with a sensitivity of 85.8% and a specificity of 77.4%. The calibration curve was close to the ideal curve, and Spiegelhalter's Z test indicated good calibration (P=0.313). Decision curve analysis showed a net benefit across a threshold probability range of 0.75%-100%, indicating high clinical applicability. CONCLUSIONS: The predictive model based on lung imaging findings (proportion with consolidation), duration of fever, high-sensitivity C-reactive protein, lactate dehydrogenase, creatine kinase, and lymphocyte-to-neutrophil ratio shows good performance for predicting SMPP in children younger than 5 years.