Reliability of Hospital-Level Mortality in Abdominal Aortic Aneurysm Repair

Abstract Objective The relationship between volume and surgical outcome has been shown for a variety of surgical procedures. The effect in abdominal aortic aneurysm repair continues to be debated. Reliability adjustment has been used as a method to remove statistical noise from hospital level outcomes but its impact on aortic aneurysm repair is not well understood. Methods We used prospectively collected data from the Vascular Quality Initiative to identify all patients undergoing abdominal aortic aneurysm repair from 2003 to 2019. We first calculated hospital-level risk-adjusted 30-day mortality rates. We subsequently used hierarchical logistic regression modelling to adjust for measurement reliability using empirical Bayes techniques. The effect of volume on risk- and reliability-adjusted mortality rates were then assessed using linear regression. Results Between 2003 and 2019, 67,073 abdominal aortic aneurysms were repaired, of which 11,601 (17.3%) were repaired with an open approach (OAR). The median annual volume was 7.4 (IQR 3.0-13.3) for open repairs and 35.4 (IQR 18.8-59.8) for endovascular repairs (EVARs). Of the 223 hospitals that performed open repairs during the study period, only 11 (4.9%) performed ≥15 open repairs per year and risk-adjusted mortality rates varied from 0-75% across all centers. After reliability adjustment, variability of risk-adjusted mortality rates decreased significantly to 1.3-8.2%. Endovascular repair risk-adjusted mortality rate variability shrunk from 0-14.3% to 0.3-2.8% after reliability adjustment. A decreasing trend in mortality was found with increasing annual case volume for OAR with each additional annual case associated with a 0.012% decrease in mortality (p=0.05) whereas the relationship was non-significant for EVAR (p=0.793). Conclusion The majority of hospitals do not perform a sufficient number of annual cases to generate a reliable center-specific mortality rate for open aneurysm repair. Center specific mortality rates for low-volume centers should be viewed with caution, as a substantial proportion of the variation for these outcomes is statistical noise rather than true center-level differences in quality of care.