University of Chicago
Sarah Urbut is a current cardiology fellow at Massachusetts General Hospital in Boston, MA with an interest in statistics, genomics, and preventive cardiology. For her postdoctoral years, she plans to combine new efforts in Bayesian risk modeling with causal inference to infer individualized lifetime risk prediction and therapeutic benefit for truly precision medicine.
Outside of science and medicine, she enjoys road cycling, traveling, comedy and supporting the Chicago White Sox.
Lifetime risk prediction; Heterogeneity of Treatment effects; Bayesian statistics, variable selection, method development
Existing guidelines for cardiovascular risk-reducing therapy calculate static fixed window estimates of risk for coronary artery disease. Under the current approach using a limited set of known cardiovascular risk factors the covariate with the greatest influence on this 10-year risk is age. These 10-year equations fail to accurately estimate risk in select populations, in particular younger individuals and those with hereditary risk. Critically, even the revised Pooled Cohort Equations (PCE) of 2018 critically mention that reducing overestimation of certain groups it was not sufficient to correct misestimation problems; the statistical methods also required revision to improve equation accuracy but do little to resolve the static and fixed versus dynamic longitudinal modeling framework. This leaves a sizable portion of the population without adequate prevention. Such models and the inherent assumptions within assume a fixed effect of traditional risk factors, or at best, a linear interaction with time. Furthermore, a sizable number of people are also classified as being at intermediate risk, for whom the optimal preventive strategy should be more precise.
Our goal is to ultimately provide individual trajectories of overall coronary artery disease risk for patients conditional on their time-varying covariates, recognizing time varying effects of the predictors, updated predictors, competing risks and accumulated exposure.The goal will be to add the importance of accumulated risk to inherent primordial ‘genomic risk’ and to see how the additive importance of acquired risk changes total prediction over the life course. Critically, we aim to provide a trajectory over time in contrast to commonly available static five- or ten-year risk projections, and critically, to offer covariate specific counterfactual representations of the alternative course. These counterfactual predictions provide ideal proposition for practical therapeutic preventive strategies, which has been critically absent from existing work.