Keith Feldman, PhD

Associate Professor of Pediatrics, University of Missouri-Kansas City School of Medicine

Full Biography

Keith Feldman, PhD, Research Faculty, recently received a three-year, $299,997 Transformational Project Award grant from the American Heart Association for his project “Identifying Risk Factors for Severe Fontan Associated Liver Disease: A Computational Framework to Improve Matching in Case-Control Study Designs”.

The project will develop a computational framework for identifying optimal control patients for a given study based on all available medical history and clinical trajectory of patients, rather than a subset of hand-picked factors.

As for the focus on congenital single ventricle disease, Dr. Feldman explains that despite clinical improvements now extending the life expectancy of these patients, due to their compromised cardiovascular function, they tend to develop Fontan Associated Liver Disease (FALD). FALD is a progressive disorder that can include fibrosis to liver failure. A degree of FALD is present in nearly all single ventricle disease patients, but doctors and researchers continue to try to identify the risk factors that lead to end-stage disease. However, given how medically complex these patients are, disentangling specific factors is extremely challenging.

To better isolate the effect of a risk factor among the data, researchers often utilize retrospective matched case-control (MCC) study designs. By “matching” patients who exhibit an outcome (cases) to similar patients who do not (controls) it is possible to estimate the direction and magnitude of associations by comparing the incidence of a risk factor between groups.

“Matched case-control studies are powerful in removing bias from non-randomized studies. However, reliability of them is dependent on identifying controls sufficiently similar to a given case,” explains Dr. Feldman. “And the selection of your control is subject to human bias.”

The measures of similarity used to “match” patients are commonly based on a limited set of demographics or simple measures of disease state. The decision of which factors are used remains at the discretion of a study team and can vary widely between studies. This raises the possibility that different controls may have been selected if a more comprehensive set of confounders (a variable that influences both the dependent variable and independent variable) were considered.

Moreover, these approaches utilize data from only a single point in time, overlooking the importance of patients’ clinical trajectories. Dr. Feldman recognized the need to standardize and expand the matching process. His project will both develop the computational framework for case-control matching and evaluate the impact of case-control matching on the precision of study results.

“We expect this approach to allow for more reliable identification risk factors of severe FALD from the heterogenous data of patients with single ventricle disease. Furthermore, given that the framework representations can be generalized to any specified outcome, we expect this approach will facilitate the study of a range of congenital heart conditions for which matched case-control studies remain the only practical design,” he said.

Together with Dr. Feldman, the study team consists of co-investigators Ryan Fischer, MD, Lori Erickson, PhD, MSN, CPNP, and Hung-Wen (Henry) Yeh, PhD, MS.