Benchmarking of surgical complications in gynaecological oncology: prospective multicentre study.

ObjectiveTo explore the impact of risk-adjustment on surgical complication rates (CRs) for benchmarking gynaecological oncology centres. DesignProspective cohort study. SettingTen UK accredited gynaecological oncology centres. PopulationWomen undergoing major surgery on a gynaecological oncology operating list. MethodsPatient co-morbidity, surgical procedures and intra-operative (IntraOp) complications were recorded contemporaneously by surgeons for 2948 major surgical procedures. Postoperative (PostOp) complications were collected from hospitals and patients. Risk-prediction models for IntraOp and PostOp complications were created using penalised (lasso) logistic regression using over 30 potential patient/surgical risk factors. Main outcome measuresObserved and risk-adjusted IntraOp and PostOp CRs for individual hospitals were calculated. Benchmarking using colour-coded funnel plots and observed-to-expected ratios was undertaken. ResultsOverall, IntraOp CR was 4.7% (95% CI 4.0-5.6) and PostOp CR was 25.7% (95% CI 23.7-28.2). The observed CRs for all hospitals were under the upper 95% control limit for both IntraOp and PostOp funnel plots. Risk-adjustment and use of observed-to-expected ratio resulted in one hospital moving to the >95-98% CI (red) band for IntraOp CRs. Use of only hospital-reported data for PostOp CRs would have resulted in one hospital being unfairly allocated to the red band. There was little concordance between IntraOp and PostOp CRs. ConclusionThe funnel plots and overall IntraOp (approximate to 5%) and PostOp (approximate to 26%) CRs could be used for benchmarking gynaecological oncology centres. Hospital benchmarking using risk-adjusted CRs allows fairer institutional comparison. IntraOp and PostOp CRs are best assessed separately. As hospital under-reporting is common for postoperative complications, use of patient-reported outcomes is important.