Hospital volume and perioperative outcomes for radical cystectomy: a population study

Abstract

Objectives: To evaluate the association between hospital volume and perioperative outcomes of radical cystectomy (RC) using state population data for a contemporary Australian cohort. Patients and Methods: Patients undergoing RC for urothelial malignancy in the state of Victoria, Australia between July 2003 and June 2014 were identified using the Victorian Admitted Episodes Dataset (VAED). Hospitals were divided into tertiles according to their caseload per year. Hospitals performing <4 RCs/year were defined as low-volume hospitals (LVH), 4–10 RCs/year as medium-volume hospitals (MVH), and >10 RCs/year as high-volume hospitals (HVH). Perioperative outcomes derived included: in-hospital mortality (IHM), prolonged length of stay (LOS; >14 days), prolonged intensive care unit (ICU) admission (>24 h), and requirement for blood transfusion. The relationship between hospital volume and perioperative outcomes was assessed using logistic regression. Results: During the 11-year study period, 803 patients underwent RC for bladder cancer. The overall IHM rate was 2.2% (LVH 3.7%, MVH 2.5%, HVH 0.9%). Other outcomes observed were prolonged LOS (45%), prolonged ICU admission (31%) and requirement for blood transfusion (56%). On multivariate analysis, LVH was found to be associated with increased IHM (odds ratio [OR] 5.74, P = 0.04) and prolonged ICU admission (OR 11.58, P < 0.001) when compared to HVH. There was a lower rate of prolonged LOS for LVH (OR 0.60, P = 0.01). No significant relationship was identified for LVH and blood transfusion. Conclusion: Perioperative outcomes in Victoria are comparable to international standards. Our results add further population study evidence to the volume–outcome relationship in RC. There was a significant association between LVH and both IHM and prolonged ICU admission. This subgroup of patients would appear to benefit from transfer of care to a HVH. The role of centralisation of RC in Australia should be further considered.