Single-stage open window thoracostomy with simultaneous muscle flap transposition and early negative pressure wound therapy for chronic empyema: a propensity score- and machine learning-based study.

backgroundTo compare the clinical efficacy of a single-stage surgical approach combining open window thoracostomy, muscle flap transposition, and early negative pressure wound therapy (NPWT) with that of conventional staged management for chronic empyema.

methodsThis retrospective, single-center cohort included 45 patients with chronic empyema (single-stage, n = 7; conventional, n = 38) who were treated from 2009 to 2024. The primary endpoint was the cavity reduction ratio measured via three-dimensional computed tomography (3D CT). To minimize selection bias, we applied propensity score matching (PSM) and inverse probability of treatment weighting (IPTW). We prespecified the average treatment effect on the treated (ATT) as the primary estimand, whereas the average treatment effect (ATE) was considered exploratory because treatment assignment was strongly related to bronchopleural fistula (all single-stage patients had BPF) and overlap was limited. The follow-up intervals were linearly normalized to adjust for time-dependent variability. Machine learning models were used as complementary predictive analyses to identify predictors of cavity reduction.

resultsAfter PSM (n = 5 per group), the single-stage group had significantly greater cavity reduction than the conventional group did (90% [89–94%] vs. 37% [32–43%], p = 0.008). The IPTW analysis demonstrated significant treatment effects: 0.38 (95% CI: 0.17–0.59, p = 0.005) for the ATT and 0.63 (95% CI: 0.12–1.14, p = 0.03) for the ATE (exploratory). Machine learning models consistently identified the treatment approach and time-related variables as important predictors of cavity reduction.

conclusionsCompared with conventional management, the single-stage approach was associated with significantly greater empyema cavity reduction and was feasible even in patients with bronchopleural fistulas. Causal interpretation is most defensible for the ATT, whereas generalization to the full population (ATE) should be considered exploratory due to limited overlap. Despite the small sample size, the consistent findings across diverse analytical frameworks support the single-stage strategy as a promising approach that warrants prospective multicenter validation and clearer criteria for patient selection.