Risk Model Improves Practical Prediction of In-Hospital Mortality After TAVR

Gauging risk of mortality in patients eligible for TAVR procedures is not an easy task, but the availability of a new prediction model may help allay fears among patients and improve accuracy of decision making among clinicians.  

Take Home: Risk Model Improves Practical Prediction of In-Hospital Mortality After TAVR

“We’re about as confident as we can be in the results,” lead author Fred H. Edwards, MD (University of Florida College of Medicine–Jacksonville, FL), told TCTMD. “It’s designed for practical use as opposed to purely research.”

Edwards and colleagues developed their model using data on 13,718 consecutive patients in the Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy (TVT) Registry who underwent TAVR between November 2011 and February 2014—the development group. Their validation group then included 6,868 consecutive cases performed between March and October of 2014.

In an interview with TCTMD, Edwards said the TVT Registry model includes readily available risk factors and compared with the FRANCE-2 model—the only other risk-prediction model based entirely on a TAVR population—drew from a much larger cohort of patients. The risk factors in the new model are: higher age, lower glomerular filtration rate, need for hemodialysis, NYHA functional class IV, severe chronic lung disease, nonfemoral access site, and procedural acuity.

Good Predictive Ability

Published online March 9, 2016, in JAMA Cardiology, results showed that the model’s ability to discriminate, as measured by the C statistic, was higher for the TVT Registry model than what had been previously reported for the FRANCE-2 model (0.66 in the validation group vs 0.59 for FRANCE-2). Additionally, the model showed good agreement between predicted versus observed rates of in-hospital mortality in the total population and across some select subgroups, including those with reduced ejection fraction, higher NYHA class, and prior aortic procedures, among others.

“Everyone who undergoes an operation wants to know, ‘What are my chances?’” Edwards said. “It’s useful to have an objective index of predicted risk, . . . and a statistical model can give you exactly that.”

In a sense, he added, the model can be thought of like any other lab test, cautioning that while it should not dictate decision making, it is “one more piece of the puzzle to be evaluated in concert with all the traditional pieces of the puzzle when making a decision on management.”

Refinements Needed to Improve Accuracy, Reliability

Edwards and colleagues note some limitations of the model, including incomplete collection of frailty and quality-of-life data, questions about the registry’s auditing process, and risk factors taken from the early TAVR experience that may not include covariates found to influence mortality. They also note that it is only the first in a series of risk models the TVT Registry plans to develop. Future models, they add, will focus on 30-day and 1-year mortality outcomes.

Other factors could also be considered, write Laura Mauri, MD, MSc, and Patrick T. O’Gara, MD (Brigham and Women’s Hospital, Boston, MA), in an editorial accompanying the study.

“To this list should be added the lack of clarity regarding the process by which expert opinion (rather than incorporation of new findings from data analysis) drove development of the model and the absence of information regarding major in-hospital cardiac (need for permanent pacemaker or implantable defibrillator, atrial fibrillation, and cardiac arrest) and noncardiac (stroke and vascular access) complications, as well as patient-oriented outcomes (functional status and quality of life),” they write.

It is encouraging, they add, that the investigators plan to improve the model, “with the ultimate goal of creating a tool that provides a fuller picture of anticipated survival and functional outcomes for the TAVR population, the demographics of which may change considerably in the years ahead.”

The editorialists maintain that a reliable risk score could be used in the future as a way for sites to improve the accuracy by which they compare their patients and outcomes, which will in turn allow for continuous quality improvement of heart teams.

To TCTMD, Edwards said Mauri and O’Gara are “right on the mark” with their suggestions for improvements.

“This should be regarded as a first step. The model that we have a year from now very likely will be different, [because] we will have more complete data.” Edwards noted. “In a year, certainly in 2 years, we hope to add [quality-of-life and frailty] information into the model.”

Edwards concluded by saying that he is confident that there are enough data now to construct a 30-day model for mortality and another to predict the probability of stroke after TAVR. 


  • Edwards FH, Cohen DJ, O’Brien SM, et al. Development and validation of a risk prediction model for in-hospital mortality after transcatheter aortic valve replacement. JAMA Cardiol. 2016;Epub ahead of print. 
  • Mauri L, O’Gara PT. Predicting outcomes in individual patients after transcatheter aortic valve replacement: small steps on the path to improved decision making. JAMA Cardiol. 2016;Epub ahead of print.



  • Edwards and O’Gara report no relevant conflicts of interest. 
  • Mauri reports grants to her institution from Abbott, Boston Scientific, and Medtronic. 

Related Stories: