Quality indicators of end-of-life cancer care: What rate is right?
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
57 Background: Many publications use administrative health care data to describe quality of care indicators at the end of life (EOL). However, very little is available to help decide on optimal rates for these indicators. The purpose of this abstract is to develop data-driven and achievable benchmark rates for EOL quality indicators using administrative data from 4 provinces in Canada. Methods: Five quality indicators of EOL care were defined and measured using linked administrative data for each of the 33 regions across British Columbia, Alberta, Ontario and Nova Scotia. These were: emergency department (ED) use, intensive care unit (ICU) admission, physician house calls (MD) and nursing visits at home (RN) prior to death, and death in hospital (DH). First, an empiric benchmark was defined by determining indicator rates among the top ranked regions to include the top decile of patients overall. Second, funnel plots were used to graph the age and sex adjusted indicator rates for each region along with the overall average value and 95% confidence limits (CL) that accounted for region size. Results: There was significant variation in rates for each indicator among the regions. Minimum and maximum rates for ED, ICU, RN, MD and DH varied approximately 2 to 4 fold across the regions with MD showing the greatest variation. Benchmark rates based on the top decile performers were: ED 34%, ICU 2%, MD 34%, RN 63%, DH 38%. With the exception of ICU, funnel plots demonstrated that mean indicator rates and their 95% CL were uniformly worse than these benchmarks even after adjusting for age and sex. Additionally, few regions met the benchmark rates. Conclusions: There is significant variation in EOL quality indicators across regions in 4 provinces in Canada. The combination of these two methods allows each region to determine its performance relative to both a benchmark and the overall average. As a result, each region is then able to gauge their performance with greater context which facilitates priority setting and resource deployment. These two methods demonstrate how decreasing variation and striving for a target can drive quality improvement. Deriving benchmark values from ‘real world’ data offers the advantage of realistically achievable targets.
