How Sarah Cannon Blood Cancer Network (SCBCN) Uses Historical Data to Benchmark Survival, Transplant Related Mortality, Engraftment and GVHD for Performance Improvement

Abstract

Purpose To describe how SCBCN uses CIBMTR Portal data to define internal program benchmarks. Effective benchmarking is a way of establishing standards to detect performance gaps and potential causes contributing to performance levels [1] . Despite a profusion of published hematopoietic cell transplantation (HCT) data, representative, reliable benchmarks are unavailable due to variation across geographies, centers and patient populations. The SCBCN consists of 6 programs that adopted a standardized quality systems, patient eligibility criteria, HCT pathways, training and credentialing. From 2013 to 2018, SCBCN relied on externally sourced data for expected performance levels (EPLs). Despite potential limitations of comparability and timeliness, EPLs were used as benchmarks in dashboards for quarterly quality meetings. Failure to meet EPLs prompted investigations for cause, improvement initiatives, and focused reporting to local and network quality committees. Methods SCBCN now performs > 1200 HCTs annually and has data available in the CIBMTR Portal [2] on 7,449 HCTs performed 2008 to date [3] for 6,827 patients (Figure 1); 2,698 allogeneic and 4,751 autologous HCTs. These metrics and associated benchmarks include 30 days, 100 days and 1-year post-HCT survival and transplant-related mortality (TRM), relative percentages of graft vs. host disease by grade (Figure 2), and median days to engraftment (Figure 3). The SCBCN Quality Management Committee, has endorsed use of these data for dashboard benchmarks and has discontinued use of external data. Findings Because the care for HCT patients is standardized within SCBCN, benchmarks derived from in-network data provide more relevant EPLs than out-of-network data. Even with the aggregated numbers, sample size diminished with each data filter added to a query; consequently SCBCN has lower confidence that network estimates are representative of the broader population that the sample is intended to represent. In reporting views built on small “n” (e.g. pediatrics, rare diseases), past performance will still be benchmarked for tracking and trending but not established as a formal EPL. Next steps will be to define additional available benchmarks and access benchmarks data in a more granular and timely manner with StafaCT, a workflow solution implemented in SCBCN.