Abstract
OBJECTIVE: This study aims to compare real-world clinical response, safety, and institutional medication efficiency of intravenous (IV) azithromycin (AZI) versus erythromycin lactobionate (ERY) in hospitalized children with Mycoplasma pneumoniae pneumonia (MPP). METHODS: A retrospective cohort of 1,049 children with PCR- or serology-confirmed MPP was assembled (AZI: n = 672; ERY: n = 377). Propensity scores were estimated using prespecified baseline confounders (sex, age, severity phenotype, concomitant antibacterial agents, antiviral co-treatment). A 1:1 nearest-neighbor propensity score matching (PSM) without replacement cohort was built (364 matched pairs per arm). The primary endpoint was a three-level composite ordinal outcome (cure, improvement, ineffective) hierarchically assigned at 72 ± 12 h after IV macrolide initiation, without assuming that missing domains imply success. Two sensitivity cohorts tested missingness assumptions. Secondary endpoints included LOS, macrolide duration, and corticosteroid escalation, interpreted as adaptive process nodes. Sparse safety used bias-reduced likelihood inference. Institutional drug efficiency was evaluated by decomposing macrolide costs into dispensed, consumed, and wastage-related avoidable cost signals. RESULTS: Before matching, ordinal response distributions differed modestly (P = 0.040). After PSM, composite ordinal outcomes were similar (paired ordinal P = 0.599), with comparable cure rates (33.8% vs. 32.7%). A treatment × age interaction signal was observed (P_interaction=0.008). In smaller strata (<80 per arm), ORs for a higher ordinal grade with ERY vs. AZI were 0.75 (<8 years; P = 0.096) and 2.19 (≥8 years; P = 0.029). In the adjusted full cohort, ERY showed higher odds of mostly mild adverse events (adjusted OR 6.52, P = 0.006), driven by skin reactions (adjusted OR 17.90, P = 0.021) with wide CIs from sparse precision. Institutional macrolide costs were substantially higher with ERY (both P < 0.001). Duration was longer with ERY (P < 0.001), while LOS and escalation rates were similar post-match. CONCLUSIONS: IV AZI and IV ERY showed comparable overall clinical response in hospitalized pediatric MPP. The age interaction is a response-heterogeneity signal requiring confirmation, not causal proof of efficacy reversal. ERY carried higher odds of mostly mild adverse events, longer duration, and greater institutional macrolide cost burden. These results support future work on resistance-informed, sequence-aware, and child-appropriate formulation stewardship to improve interpretability, safety precision, and institutional antibiotic sustainability.