Predictive value of an early comprehensive assessment model for refractory mycoplasma pneumoniae pneumonia and internal validation

Abstract Background Refractory mycoplasma pneumoniae pneumonia (RMPP) can result in severe complications and long-term effects. Early identification of RMPP and appropriate treatments can effectively alleviate complications and restrict the progression of sequelae. There is currently a dearth of a comprehensive and efficient model for predicting and evaluating RMPP. Methods The development cohort consisted of patients with mycoplasma pneumoniae pneumonia (MPP) who underwent fiberoptic bronchoscopy between January 2019 and October 2021. Multivariable logistic regression analysis was used to identify independent risk factors for RMPP, and a nomogram model was developed that included initial admission examinations and clinical characteristics. The accuracy of the model was validated using a validation cohort of patients enrolled between November 2021 and April 2023. Result 373 patients were enrolled including 229 cases of the development cohort and 144 cases of the validation cohort. Multivariable Logistic regression analysis showed that the six independent risk factors for RMPP were age (OR = 1.151, 95% confidence interval (CI) 1.014–1.306), acute physiology and chronic health evaluation (APACHE) II score(OR = 0.872, 95% CI 0.792–0.961), computed tomography (CT) total score(OR = 1.407, 95% CI 1.258–1.575), secretion color(OR = 2.719, 95% CI 1.562–4.734), mucosal edema(OR = 5.064, 95% CI 2.748-9.300), and procalcitonin (PCT) (OR = 0.871, 95% CI 0.806–0.941). The Area Under Curve (AUC) of the model in the development cohort and validation cohort was 0.913(95%CI 0.875–0.951) and 0.811(95%CI 0.739–0.883), respectively. Hosmer-Lemeshow test showed that the model’s goodness of fit had good consistency in the development cohort and validation cohort (χ2 = 10.546, P = 0.229; χ2 = 7.894, P = 0.342). The DCA of the development and validation cohorts showed clear net benefits using the nomogram to predict RMPP. Conclusion We developed a nomogram model that integrates clinical, imaging, and bronchoscopic features to enable early and accurate prediction of the risk of RMPP in children. This model provides a quantitative tool for personalized intervention and demonstrates significant clinical application value.