Department of Statistics
University of Wisconsin-Madison
Abdominal aortic aneurysm (AAA) is a pervasive condition with high morbidity, affecting 2-4% of adults in the U.S., with 85-90% mortality in ruptured AAA cases. Surgical repair can mitigate the risk of rupture; however, open surgery is associated with high risk of complication. Endovascular aneurysm repair (EVAR) is a less-invasive repair procedure and associated with lower short-term mortality, but it is not clear whether it has long-term benefits. There are concerns that EVAR is less effective inthe long term, leading to reinterventions. Clinical trials comparing the procedures are limited in size, scope, or follow-up. Hence, we use a large Medicare enrollment data set with long-term follow-up in our analysis. To establish causal estimates of the differences in long-term mortality outcomes, we develop novel instrumental variable approaches to survival analysis. In particular, we analyze the Medicare data based on the semiparametric accelerated failure time model. Inference regarding the causal effects is carried out using a weighted bootstrap approach. Our analysis indicates there is little difference between open repair and EVAR in long-term survival.