Abstract
OBJECTIVE: To identify predictive factors for necrotizing pneumonia (NP) in children with refractory Mycoplasma pneumoniae pneumonia (RMPP) and develop a predictive nomogram. METHODS: A retrospective analysis was conducted on clinical data of children with RMPP admitted to the Affiliated Women and Children's Hospital of Dalian University of Technology between June 2023 and July 2024. The dataset was randomly split into a training set (70%, n = 197) for model development and a test set (30%, n = 77) for internal validation. The χ (2) test and Mann-Whitney U test were used to screen potential predictors, and multivariate logistic regression analysis was applied to establish a clinical prediction model. The Hosmer-Lemeshow test was used to evaluate model fit, and variance inflation factor was calculated to assess multicollinearity. The discriminatory and calibrative performance of the nomogram was evaluated using the receiver operating characteristic (ROC) curve and calibration curve, respectively. RESULTS: A total of 274 children with RMPP were analyzed. Of these, 51 who developed NP formed the necrotizing group, while the remaining 223 without NP were designated as the non-necrotizing group. The χ (2) text and Mann-Whitney U Test analysis indicated that ESR, CD4(+) T cells, NK cells, IL-4, duration of fever, and pleural effusion were significant predictors of NP in children with MPP (P < 0.05). Internal validation using the test set showed a consistency rate of 87.01% (67/77) between predicted and actual outcomes. The model demonstrated a sensitivity of 0.833, specificity of 0.877, and a Kappa coefficient of 0.590. Although predictive accuracy slightly decreased in the test set compared to the training set, the model still retained satisfactory predictive performance, indicating its potential generalizability. CONCLUSION: The prediction model incorporating ESR, CD4(+) T cells, NK cells, IL-4, duration of fever, and pleural effusion showed good predictive value for NP in children with RMPP.