ABSTRACT
OBJECTIVE: Existing evidence demonstrates some benefit of regionalization on early postoperative outcomes following lung cancer resection, but data regarding the persistence of this effect in long-term mortality are lacking. We investigated whether previously reported improvements in short-term outcomes translated to long-term survival benefit. METHODS: We retrospectively reviewed patients undergoing major pulmonary resection (lobectomy, bilobectomy, or pneumonectomy) for cancer within our integrated health care system before (2011-2013; n = 782) and after (2015-2017; n = 845) thoracic surgery regionalization. Overall survival was compared by Kaplan-Meier analysis, and 1- and 3-year mortality was compared by the by χ2 or Fisher exact test. Multivariable Cox regression models evaluated the effect of regionalization on mortality adjusted for relevant factors. RESULTS: Kaplan-Meier curves showed that overall survival was better among patients undergoing surgery postregionalization (log-rank test, P < .0001). Both 1- and 3-year mortality were decreased after regionalization: to 5.7% from 11.1% (P < .0001) for 1 year and to 17.0% from 25.5% (P = .0002) for 3 years. The multivariable adjusted Cox regression analysis revealed that only regionalization (hazard ratio [HR], 0.57; 95% confidence interval [CI], 0.42-0.76), age (HR, 1.03; 95% CI, 1.02-1.04), cancer stage (HR, 1.72, 1.83, and 2.56 for stages II, III, and IV, respectively), and Charlson comorbidity index (HR, 1.80 for 1-2; 2.05 for ≥3) were independent predictors of mortality. CONCLUSIONS: We found that overall mortality as well as 1- and 3-year mortality for lung cancer resection were lower after thoracic surgery regionalization. The association between regionalization and reduced mortality was significant even after adjusting for other related factors in a multivariable Cox analysis. Notably, surgeon volume, facility volume, surgeon specialty, neoadjuvant treatment, and video-assisted thoracoscopic surgery approach did not significantly affect mortality in the adjusted model.