Author + information
- Suzanne Arnold1,
- Sean O'Brien2,
- Sreekanth Vemulapalli3,
- David Cohen1,
- Amanda Stebbins4,
- Rosemarie Hakim5,
- David Holmes Jr.6,
- Vinod Thourani7 and
- Fred Edwards8
- 1Saint Luke's Mid America Heart Institute, Kansas City, Missouri, United States
- 2Duke University, Durham, United States
- 3Duke Clinical Research Institute, Durham, United States
- 4Duke University, Durham, North Carolina, United States
- 5Centers for Medicare & Medicaid Services, Baltimore, United States
- 6Mayo Clinic, Rochester, Minnesota, United States
- 7Emory University Hospital Midtown, Atlanta, Georgia, United States
- 8University of Florida, Jacksonville, Florida, United States
Outcomes after TAVR have improved due to better patient selection, evolving technology, and provider experience. To fairly compare these outcomes across centers requires appropriate adjustment for patient risk. We sought to develop and validate a risk adjustment model that accounted for standard clinical factors as well as pre-procedural health status and frailty—factors known to impact outcomes beyond traditional clinical risk factors.
Using data from patients who underwent TAVR as part of the STS/ACC TVT Registry (6/2013-5/2016), we developed and internally validated a hierarchical logistic regression model with site-specific random intercepts to estimate risk of 30-day mortality after TAVR based only on pre-procedural factors. The model included all factors from the original TVT in-hospital mortality model (except NYHA) and added KCCQ (health status) and gait speed (5-m walk test).
Among 21,661 TAVR patients at 188 sites, 1025 (4.7%) died within 30 days. The model was able to stratify risk based on patient factors with good discrimination (c=0.71 derivation, 0.70 split-sample validation) and excellent calibration, both overall and in key patient subgroups. The predicted 30-day mortality risk ranged from 1.1% (lowest decile of risk) to 13.8% (highest decile of risk). Independent predictors of 30-day death included older age, low body weight, worse renal function, PAD, home oxygen, prior MI, left main disease, tricuspid regurgitation, nonfemoral access, worse baseline health status, and being unable to walk.
We developed a clinical risk model for 30-day death after TAVR using a large national database with rich clinical data, including health status and frailty. This model will facilitate tracking outcomes over time as TAVR expands to lower risk patients and to less experienced sites and will allow for an objective comparison of short-term mortality rates across centers.
STRUCTURAL: Valvular Disease: Aortic