Prediction Model for Severe Mycoplasma pneumoniae Pneumonia: Treatment Guidance for Pediatric A2063/2064G-mutated Infections

Background Mycoplasma pneumoniae pneumonia (MPP) is a common respiratory infection in children, with 10–40% progressing to severe MPP (SMPP), causing serious complications and mortality. Early SMPP identification is crucial for treatment and prognosis. Moreover, infections caused by macrolide-resistant MP (MRMP) significantly contribute to SMPP development. Selecting appropriate antibiotics for pediatric patients with MRMP has become a major clinical challenge. This study aims to develop an early prediction model for SMPP and explore personalized treatment strategies based on the model for pediatric patients infected with A2063/2064G mutated MP. Methods Clinical data from 2381 pediatric patients diagnosed with MPP at Shanghai Children's Hospital between November 2019 and December 2023 were analyzed. The study compared clinical characteristics, laboratory parameters, and therapeutic outcomes between SMPP and general MPP (GMPP) cohorts. A predictive model for SMPP was developed using multivariate logistic regression analysis, with its diagnostic accuracy evaluated through receiver operating characteristic (ROC) curve analysis, including sensitivity and specificity assessments. Additionally, we analyzed medication patterns and hospital stay among patients infected with A2063/2064G mutated MP, leading to the development of early personalized treatment strategies that integrate age factors and model prediction results. Results Among 2381 MPP patients, 71.3% were SMPP, and 46.9% had A2063/2064G mutated MP. The mutation was more prevalent in SMPP than GMPP patients (54.7% vs 27.5%, P < 0.001). A seven-indicator model (fever duration, LDH, ALB, CK-MB, Neu%, WBC and D-dimer) showed excellent diagnostic performance (AUC = 0.899, 95% CI=[0.861, 0.937], sensitivity = 0.827, specificity = 0.861). In A2063/2064G mutated MP patients, tetracyclines (TCs) or fluoroquinolones (FQs)-requiring cases showed higher SMPP rates than macrolide antibiotics (MACs)-responsive cases (89.6% vs 78.0%, P < 0.001). Early TCs/FQs administration reduced hospital stays (7.30 ± 1.96 vs 8.38 ± 2.20 days, P < 0.001). The prediction model demonstrated similar accuracy across treatment groups, with comparable predicted and actual SMPP proportions. Age-stratified analysis showed peak TCs/FQs usage in patients with both A2063/2064G mutated MP infection and model-predicted SMPP. Conclusion The prediction model effectively identifies SMPP and guides interventions when combined with MP mutation status. Early TCs/FQs may benefit children with A2063/2064G mutated MP when predicted as SMPP. Clinical trial number : not applicable.