Discrepancies in Perioperative Mortality Definitions After Open and Endovascular Aortic Aneurysm Repair; Evaluation in United States Medicare Beneficiaries

Objective: Perioperative mortality after abdominal aortic aneurysm (AAA) repair has been reported in three different forms, in-hospital, 30-day, or combined in-hospital and 30-day deaths. Many administrative databases only have access to in-hospital mortality. Since length of stay after EVAR is typically shorter than open repair, this measure has been criticized as adding bias in favor of EVAR. We describe the various perioperative mortality rates in the US Medicare population.
Methods: All US Medicare patients undergoing intact AAA repair during 2000-2004 were identified from Medicare part A files (N=61,598). Matched cohorts undergoing EVAR or open repair were developed (part A and B files) using propensity score matching (n=45,660). Perioperative mortality was determined using denominator and part A files. In-hospital mortality was defined as any death during the hospital admission. 30-day mortality was defined as death within 30 days of AAA repair. Mortality rates were determined for both EVAR and Open and were compared with the chi-squared test.
Results: In hospital mortality was 4.6% vs 1.1% for Open and EVAR respectively (Table). 30 day mortality was 4.8% vs 1.6% respectively, and combined in-hospital and 30-day mortality was 5.3% vs 1.7% respectively. The absolute difference in mortality between EVAR and Open remained relatively constant between the 3 definitions (3.5%, 3.2%, and 3.7%). The absolute difference increased with age but did not vary by definition of perioperative mortality. Similarly, the relative risk of mortality remained relatively constant across definitions in all age groups.
Conclusions: The significant perioperative mortality benefit of EVAR is demonstrated using all definitions of perioperative mortality. All definitions are valid for direct comparison of the two methods. However, comparisons of different studies should specify when different definitions have been applied.