Risk-adjustment of diabetes health outcomes improves the accuracy of performance benchmarking

Eleanor Danek, Arul Earnest, Natalie Wischer, Sofianos Andrikopoulos, Anthony Pease, Natalie Nanayakkara, Sophia Zoungas

Research output: Contribution to journalArticleResearchpeer-review

6 Citations (Scopus)


Benchmarking clinical performance by comparing diabetes health outcomes across healthcare providers drives quality improvement. Non-care related patient risk factors are likely to confound clinical performance, but few studies have tested this. This cross-sectional study is the first Australian investigation to analyse the effect of risk-adjustment for non-care related patient factors on benchmarking. Data from 4,670 patients with type 2 (n = 3,496) or type 1 (n = 1,174) were analysed across 49 diabetes centres. Diabetes health outcomes (HbA1c levels, LDL-cholesterol levels, systolic blood pressure and rates of severe hypoglycaemia) were risk-adjusted for non-care related patient factors using multivariate stepwise linear and logistic regression models. Unadjusted and risk-adjusted funnel plots were constructed for each outcome to identify low-performing and high-performing outliers. Unadjusted funnel plots identified 27 low-performing outliers and 15 high-performing outliers across all diabetes health outcomes. After risk-adjustment, 22 (81%) low-performing outliers and 13 (87%) high-performing outliers became inliers. Additionally, one inlier became a low-performing outlier. Risk-adjustment of diabetes health outcomes significantly reduced false positives and false negatives for outlier performance, hence providing more accurate information to guide quality improvement activity.

Original languageEnglish
Article number10261
Number of pages11
JournalScientific Reports
Issue number1
Publication statusPublished - 1 Dec 2018

Cite this