Presentation Authors: Keith Lawson*, Katherine Daignault, Madhur Nayan, Bo Chen, Lisa Martin, Maria Komisarenko, Olli Saarela, Antonio Finelli, Toronto, Canada
Introduction: We have previously demonstrated widespread quality of care variations between hospitals for patients with renal cell carcinoma (RCC) utilizing a validated set of quality indicators (QI).1 However, the translation of this knowledge is limited by an incomplete understanding of the sources of this variation. Here we aimed to determine the relative contribution of surgeon versus hospital-level quality of care variations in order to inform the development of appropriate benchmarking strategies. Further, we sought to quantify the impact that hypothetical interventions on process-type indicators can have on outcome-type indicators.
Methods: Patients undergoing nephrectomy for RCC were identified using linked population-level administrative databases in Ontario, Canada. Hospitals and individual surgeons were benchmarked according to previously validated RCC-specific QIs using indirect standardization1, adjusting performance for clinicopathological variables inherent to their patient populations. Inter-hospital and -surgeon level variation in each QI was assessed via mixed effect models and meta-regression. Causal mediation analysis was employed to identify hospitals that could benefit from targeted quality improvement initiatives.
Results: A total of 10,111 nephrectomy patients with complete clinical and pathological data were identified between 1995 to 2014. Care was provided by 393 surgeons across 138 hospitals. After adjustment for case-mix, a total of 3-25% of surgeons and 4-32% of hospitals were statistically significant outliers for a given QI, performing worse than the provincial average. Significant variation at both hospital and surgeon levels was observed for four of the five indicators (P < 0.001). Surgeon-level variance was equal to or greater than hospital-level in four of five indicators. The causal mediation analysis identified hospitals where improvements in minimally invasive surgery rates could translate to shorter average length of stay.
Conclusions: Both surgeon- and hospital-level effects contribute to the observed variance in quality of care received by patients undergoing surgery for RCC. Further, we demonstrate the feasibility of identifying hospitals for targeted quality improvement interventions through causal mediation analysis. Collectively, these data demonstrate the utility of QI variance decomposition and mediation analysis methods for informing the development of data-driven quality benchmarking strategies for RCC. _x000D_
1. Lawson et. al. Eur Urol 2017. 72(3):379-386.
Source of Funding: Princess Margaret Cancer Centre Foundation, Canadian Institutes of Health Research