Abstract:
|
A key goal of personalized medicine is to develop tools to accurately predict an individual's disease trajectory. This information can then be used to target therapies. Typically, predictive methods focus on personalization by conditioning on observed characteristics alone. However, in complex, chronic diseases there are many unobserved factors that affect progression. Examples include differences in disease pathways, past exposures, and family traits that may not be measured (or possible to measure). In this work, using Scleroderma --- a systemic autoimmune disease --- as an example disease, we explore a Bayesian framework for individualizing prognosis of disease trajectories. The proposed model incorporates the diverse set of markers that are routinely collected during an outpatient visit. We show promising results for improving prognosis using data tracking individuals over 20 years at the Hopkins Scleroderma Center, one of the largest centers nationally for treating scleroderma. We also discuss our experience with deploying a pilot based on our work at this center.
|