Objective To investigate predictors of non-invasive positive pressure ventilation (NIPPV) failure in children with severe pneumonia using propensity score matching. Methods A retrospective case-control study included 79 children with severe pneumonia who experienced NIPPV failure at the Children's Hospital of Chongqing Medical University from January 2021 to December 2023 and 118 contemporaneous children who had successful NIPPV. Propensity score matching was used for 1∶1 matching on baseline variables. Multivariable logistic regression analysis was performed to identify predictors of NIPPV failure, and receiver operating characteristic (ROC) curves were used to evaluate the predictive performance of selected variables. Results After matching, 77 patients were included in each group, with no statistically significant differences in baseline characteristics (P>0.05). Multivariable logistic regression showed that respiratory rate (OR=1.109) and heart rate (OR=1.064) at 12-24 hours after NIPPV, Pediatric Early Warning Score (PEWS) score before NIPPV (OR=2.809), and lung consolidation on chest imaging before NIPPV (OR=16.144) were positive predictors of NIPPV failure (all P<0.05). Pulse oxygen saturation (SpO₂) at 12-24 hours after NIPPV was a negative predictor of NIPPV failure (OR=0.414, P<0.05). ROC analysis showed that SpO₂ at 12-24 hours after NIPPV predicted NIPPV failure with an area under the curve (AUC) of 0.906, sensitivity of 70.1%, specificity of 98.7%, and an optimal cutoff of 92%. The PEWS score before NIPPV predicted NIPPV failure with an AUC of 0.784, sensitivity of 64.9%, specificity of 84.4%, and an optimal cutoff of 3.5 points. Conclusions In children with severe pneumonia receiving NIPPV, close monitoring of respiratory rate, heart rate, and SpO₂ is essential, and risk stratification for treatment failure should incorporate pre-treatment chest imaging findings (consolidation area) and PEWS score.