Objective The individualized prediction of treatment regimens of macrolide-unresponsive Mycoplasma pneumoniae pneumonia (MUMPP) is scarce. The aim of this study was, therefore, to evaluate the relevant data of patients and construct a nomogram for identifying refractory Mycoplasma pneumoniae pneumonia (RMPP) among children continued to be treated with macrolide after the confirmation of MUMPP, providing a reference for the choice of treatment regimen. Methods We performed a retrospective study involving 162 children who continued to be treated with macrolide (azithromycin) after the confirmation of MUMPP without antibiotic changes between January 2020 and January 2022. We collected data on clinical feature, hospitalization period, treatments, laboratory data, extrapulmonary symptoms, parapneumonic effusion, and connections with other respiratory pathogens. In addition, the independent risk factors for RMPP were determined through univariate and multivariate analyses, and then a nomogram was constructed and validated. Results In this study, the multivariate logistic regression analysis showed that age, leukocyte count, neutrophil proportion, serum procalcitonin, and lactate dehydrogenase were independent risk factors for RMPP. Using the five independent associated factors, the nomogram for identification of RMPP was constructed. Moreover, the area under the ROC curve (AUC) was 0.925 (95% CI: 0.882–0.968) for the nomogram showing excellent discrimination. The calibration curve, close to the 45-degree line, exhibited good calibration of nomogram. Conclusion We constructed and validated a visual and user-friendly nomogram for individualized prediction of RMPP risk in children who continued to be treated with macrolide after the confirmation of MUMPP based on five variables. According to the nomogram model, continuation of macrolide should be considered rather than second-line antibiotics including tetracyclines (doxycycline or minocycline) and fluoroquinolones for MUMPP children with low predictive values.