Admission-Based Prediction of Macrolide Resistance in Children Aged >28 Days and ≤8 Years with Mycoplasma pneumoniae Pneumonia: A Retrospective Prediction Model Study

Weiwei Zhang,1,* Kelei Deng,2,* Zhiyue Deng,3,* Ying Nie,4 Yue Wu,5 Shuhe Yang,1 Xuewei Wang,6 Mei Mei,5 Jing Nie51College of Clinical Medicine, Jiamusi University, Jiamusi, Heilongjiang, People’s Republic of China; 2Department of Pediatrics, Gaozhou People’s Hospital, Gaozhou, Guangdong, People’s Republic of China; 3Department of Pediatrics Nagoya University Graduate School of Medicine, Nagoya, Japan; 4Department of Pediatric Rehabilitation, The Third Affiliated Hospital of Jiamusi University, Jiamusi, Heilongjiang, People’s Republic of China; 5Department of Pediatrics, The First Affiliated Hospital of Jiamusi University, Jiamusi, Heilongjiang, People’s Republic of China; 6Department of Neonatology, Jiamusi Central Hospital, Jiamusi, Heilongjiang, People’s Republic of China*These authors contributed equally to this workCorrespondence: Jing Nie, Email [email protected] Mei Mei, Email [email protected]: Macrolide-resistant Mycoplasma pneumoniae pneumonia (MRMP) complicates the management of pediatric Mycoplasma pneumoniae pneumonia (MPP), especially around the 8-year treatment threshold, where second-line antimicrobial choices require careful risk-benefit assessment. This study aimed to develop and internally validate an admission-based model for predicting macrolide resistance in children aged > 28 days and ≤ 8 years with MPP.Patients and Methods: This single-center retrospective prediction model study included hospitalized children with confirmed MPP at the First Affiliated Hospital of Jiamusi University from October 2023 to October 2024. Candidate predictors were restricted to demographic and laboratory variables available at admission or during early hospitalization. Univariable analyses were followed by multivariable logistic regression with stepwise selection. Model performance was assessed by receiver operating characteristic (ROC) analysis, calibration plots, decision curve analysis (DCA), and bootstrap internal validation.Results: A total of 724 children were included, of whom 146 (20.2%) were classified as having macrolide resistance. The final model included height, C-reactive protein (CRP), lactate dehydrogenase (LDH), and creatine kinase-MB (CK-MB). The model showed an apparent area under the ROC curve (AUC) of 0.752 (95% CI, 0.711– 0.793), and the optimism-corrected AUC after internal validation was 0.746. At the selected cutoff, sensitivity was 0.719, specificity was 0.680, and the negative predictive value was 0.906. Calibration and DCA suggested acceptable agreement and potential clinical usefulness within the observed sample.Conclusion: An admission-based model using height, CRP, LDH, and CK-MB showed moderate discrimination for predicting macrolide resistance in children aged > 28 days and ≤ 8 years with MPP. The model showed a high negative predictive value and may help identify children at lower risk of molecularly confirmed macrolide resistance; molecular testing remains necessary to confirm resistance in children at higher predicted risk, and external validation is required before routine clinical use.Keywords: Mycoplasma pneumoniae pneumonia, macrolide resistance, children, prediction model, nomogram