Background: There is substantial global variation in demographics, disease burden, and treatment for gastric cancer patients. Benchmarking is an instrument to assess such variation and enables to investigate to which extent case-mix and treatments explain differences in outcomes. We aimed to evaluate hospital-level variation in surgical outcomes following gastrectomy for gastric cancer before and after adjusting for case-mix and treatment-related factors. Methods: Data were retrieved from the GastroBenchmark and GASTRODATA databases, including consecutive gastric cancer resections performed between 2017 and 2021 from 43 centers. Patients who underwent a (sub)total gastrectomy for adenocarcinoma were identified. Outcomes included 30-day mortality, severe complications (Clavien-Dindo grade ≥ 3a), > 15 lymph nodes retrieved, negative resection margin (R0), prolonged hospitalization (> 14 days), readmissions (< 30 days), reoperations, and escalation of care. We assessed absolute inter-hospital variation for outcomes, and estimated outcomes using mixed-effect logistic regression models with a random intercept. We estimated crude, case-mix adjusted, and case-mix and treatment adjusted hospital effects. The conditional and marginal pseudo-R2 were used to quantify the variance in outcome explained by case-mix and treatment-related factors. Results: A total of 7818 patients from 41 hospitals were included, with contributions ranging from 12 to 2554 patients per hospital (IQR: 49–146). Observed 30-day mortality and severe complications ranged from 0 to 9.7% (IQR: 3.2%) and 5.3 to 31% (IQR: 7.7%), respectively. Larger variation between hospitals was observed for retrieval of > 15 lymph nodes (IQR: 12.3%), prolonged hospitalization (IQR: 14.4%) and readmissions (IQR 11.3%). This variation was reduced in the crude model, while adjusting for case-mix and treatment-related factors did not significantly reduce variation for any outcome. Case-mix factors had a limited contribution to the explained variance, except for 30-day mortality (33.9%) and negative resection margins (31.7%). Adding treatment-related factors increased the explained variance for 30-day mortality by 40.8%, but had low impact (< 10%) on the variance in most surgical outcomes. Conclusions: Case-mix and treatment factors are not the primary drivers of variation in surgical outcomes following gastrectomy. Case-mix adjustment can improve the validity of global comparisons for 30-day mortality, but does not seem essential for comparing other investigated outcomes.
Hospital variation in surgical outcomes for gastric cancer: the impact of case-mix and treatment across a global cohort
Baiocchi G. L.;
2026-01-01
Abstract
Background: There is substantial global variation in demographics, disease burden, and treatment for gastric cancer patients. Benchmarking is an instrument to assess such variation and enables to investigate to which extent case-mix and treatments explain differences in outcomes. We aimed to evaluate hospital-level variation in surgical outcomes following gastrectomy for gastric cancer before and after adjusting for case-mix and treatment-related factors. Methods: Data were retrieved from the GastroBenchmark and GASTRODATA databases, including consecutive gastric cancer resections performed between 2017 and 2021 from 43 centers. Patients who underwent a (sub)total gastrectomy for adenocarcinoma were identified. Outcomes included 30-day mortality, severe complications (Clavien-Dindo grade ≥ 3a), > 15 lymph nodes retrieved, negative resection margin (R0), prolonged hospitalization (> 14 days), readmissions (< 30 days), reoperations, and escalation of care. We assessed absolute inter-hospital variation for outcomes, and estimated outcomes using mixed-effect logistic regression models with a random intercept. We estimated crude, case-mix adjusted, and case-mix and treatment adjusted hospital effects. The conditional and marginal pseudo-R2 were used to quantify the variance in outcome explained by case-mix and treatment-related factors. Results: A total of 7818 patients from 41 hospitals were included, with contributions ranging from 12 to 2554 patients per hospital (IQR: 49–146). Observed 30-day mortality and severe complications ranged from 0 to 9.7% (IQR: 3.2%) and 5.3 to 31% (IQR: 7.7%), respectively. Larger variation between hospitals was observed for retrieval of > 15 lymph nodes (IQR: 12.3%), prolonged hospitalization (IQR: 14.4%) and readmissions (IQR 11.3%). This variation was reduced in the crude model, while adjusting for case-mix and treatment-related factors did not significantly reduce variation for any outcome. Case-mix factors had a limited contribution to the explained variance, except for 30-day mortality (33.9%) and negative resection margins (31.7%). Adding treatment-related factors increased the explained variance for 30-day mortality by 40.8%, but had low impact (< 10%) on the variance in most surgical outcomes. Conclusions: Case-mix and treatment factors are not the primary drivers of variation in surgical outcomes following gastrectomy. Case-mix adjustment can improve the validity of global comparisons for 30-day mortality, but does not seem essential for comparing other investigated outcomes.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
