Construction and Validation of an Early Identification Model for Refractory Mycoplasma pneumoniae-Positive Lobar Pneumonia

Abstract Objective This study analyzed the relationship between clinical parameters and prognosis in children with Mycoplasma pneumoniae (MP)-positive lobar pneumonia and developed an early identification model. Methods Relevant clinical parameters were collected. Patients were then categorized into two groups based on their length of hospital stay: 116 cases in the refractory group (≥10 days) and 94 cases in the non-refractory group (<10 days). A univariate analysis of variance and binary logistic regression were utilized to develop a predictive model, accompanied by the construction of a nomogram. The model's performance was assessed using receiver operating characteristic (ROC) curves, diagnostic calibration curves, and decision curve analysis (DCA) curves. Furthermore, clinical data from 100 additional cases of MP-positive lobar pneumonia in children treated at other centers were gathered for external validation of the model. Results Binary logistic regression analysis identified four independent risk factors for prolonged disease duration in children with MP-positive lobar pneumonia: erythrocyte sedimentation rate (ESR), globulin, lactate dehydrogenase (LDH), and SF. We constructed a nomogram model based on these risk factors. In the training set, the area under the curve (AUC) was 0.869 (95% CI: 0.822–0.917), with a sensitivity of 68.54% and a specificity of 82.61%. For the test set, the AUC increased to 0.918 (95% CI: 0.866–0.971), demonstrating a sensitivity of 91.67% and a specificity of 78.69%. The DeLong test results indicated that the difference in AUC between the two datasets was not statistically significant (D = − 1.724, p = 0.086). Calibration curve analysis confirmed that the nomogram model exhibited a good fit in both the training set (Hosmer–Lemeshow test, χ2 = 8.120, p = 0.421) and the validation set (Hosmer–Lemeshow test, χ2 = 14.601, p = 0.067). DCA further demonstrated that the model performed significantly across a range of threshold probabilities. Conclusion The nomogram model developed for predicting refractory MP-positive lobar pneumonia in children has significant clinical value and can guide personalized treatment strategies.