Griffin KJ, Fleming SJ, Bailey MA, Czoski-Murray C, Baxter PD. Target setting for elective infra-renal AAA surgery: A single mortality figure?
Surgeon 2013;
11:191-8. [PMID:
23290747 DOI:
10.1016/j.surge.2012.12.003]
[Citation(s) in RCA: 1] [Impact Index Per Article: 0.1] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/01/2012] [Revised: 12/03/2012] [Accepted: 12/07/2012] [Indexed: 11/30/2022]
Abstract
OBJECTIVES
One of the key standards set by the UK NAAASP is that centres performing elective abdominal aortic aneurysm (AAA) repair have a mortality rate of <6%. In light of this, and the current aim to reduce elective AAA repair mortality to 3.5% by 2013, we sought to investigate the statistical validity of such targets.
METHODS
The National Vascular Database (NVD) was interrogated and the degree of AAA missing data and its geographical variation is described. Utilising published data from 2006 to 2008 a funnel plot was used to illustrate NHS Trust level data for current estimates of mortality rate. A binomial distribution model was applied to calculate variation in observed mortality rates in relation to number of patients treated, based on a "true" mortality rate of 3.5%. Funnel plots were constructed using simulated data-sets for units performing 10, 30, 50, 100 or 150 procedures annually with control-limits calculated using a cumulative probability distribution. Finally the effect of case-mix on mortality was modelled and shown graphically.
RESULTS
The NVD AAA data set shows a range of data missingness across variables (median 22%, IQR 10-64%). High levels of missingness typically coincide with non-required, non-preferred variables however this is subject to geographical variation. Funnel plots of simulated data demonstrate that smaller units have greater variability in 3-year mortality (range 0.0-10.0%) than the largest units performing 150 procedures annually (1.3-5.6%). Around 20% of NVD variables are described as "preferred", these typically relate to clinical measurements and patient medications and would inform any risk model of mortality. Data missingness amongst these variables ranges from 5 to 50%.
CONCLUSIONS
There are many problems with the use of a single mortality figure to assess performance. These include the natural statistical variability and the means by which "case-mix" is taken into consideration. This article calls for further research into mortality target setting and suggests strategies which may help provide solutions nationally and facilitate international comparison.
Collapse