Objective: To elucidate the clinical characteristics and risk factors associated with liver injury in children with severe Mycoplasma pneumoniae pneumonia (SMPP). Methods and materials: This 2-center retrospective study analyzed 1321 children with SMPP from Nanjing Medical University Affiliated Children’s Hospital (January–December 2023), divided into liver injury (alanine aminotransferase > 80 IU/L) and nonliver injury groups. Medical records were used to compare clinical features and prognoses. External validation used data from 640 patients at Nanjing Lishui People’s Hospital. Results: Of the 1321 patients, 55 had liver injury. These patients were typically older, had more severe pulmonary manifestations (eg, pulmonary consolidation, atelectasis and pleural effusion), and higher levels of white blood cell count, neutrophil percentage, neutrophil-to-lymphocyte ratio, lactate dehydrogenase (LDH), D-dimer, alanine aminotransferase, and aspartate aminotransferase. They also had longer hospital stays, higher costs and greater need for intensive care and oxygen support, along with higher risks of pulmonary embolism, necrotizing pneumonia and refractory Mycoplasma pneumoniae pneumonia. Multivariate logistic regression identified elevated LDH (odds ratio =1.040, 95% confidence interval: 1.027–1.055, P < 0.001) and D-dimer (odds ratio = 2.149, 95% confidence interval: 1.648–2.802, P < 0.001) as independent risk factors. The combined prediction model showed an area under the curve of 0.811. External validation confirmed the reliability of LDH and D-dimer as predictive biomarkers. Conclusions: SMPP with liver injury shows distinct clinical features. Affected children are often older and exhibit severe pulmonary symptoms. These patients face prolonged hospitalization, higher medical costs and increased need for intensive care and oxygen support. They are also at greater risk of adverse outcomes such as pulmonary embolism, necrotizing pneumonia and refractory M. pneumoniae pneumonia. External validation confirms LDH and D-dimer as reliable predictive biomarkers.