Presentation: Precision Medicine in the Data Science Age
Speaker: Daniel Luckett, Ph.D., Postdoctoral Fellow, UNC Gillings School of Global Public Health, University of North Carolina at Chapel Hill
Abstract: Precision medicine is the paradigm of improving patient outcomes by tailoring treatment to the individual. Precision medicine is formalized through treatment regimes, or maps from patient covariates into a recommended treatment. In this talk, I will discuss the estimation of treatment regimes in two different scenarios. The first is the emerging area of mobile health. Mobile technologies offer potential for developing precisely tailored treatment strategies as they allow for monitoring a patient’s health status in real time. However, techniques for analyzing mobile health data must accommodate observations at a fine granularity and an indefinite time horizon. I will present a novel machine learning method for estimating optimal treatment regimes from mobile health data, with an application to controlling blood glucose levels in patients with type 1 diabetes. Second, I will discuss precision medicine in the presence of multiple outcomes. Clinical decision making often involves balancing trade-offs between competing outcomes, precluding direct application of standard methods for estimating optimal treatment regimes. I will present a method for estimating a composite outcome from observational data under the assumption that clinicians approximately (i.e., imperfectly) act to optimize each patient’s individual utility function, with an application to balancing symptoms of depression and mania in patients with bipolar disorder.