B1579
Title: A Bayesian approach for modeling variance of intensive longitudinal Biomarker data as a predictor of health outcomes
Authors: Mingyan Yu - University of Michigan (United States) [presenting]
Zhenke Wu - University of Michigan at Ann Arbor (United States)
Margaret Hicken - University of Michigan (United States)
Michael Elliott - University of Michigan (United States)
Abstract: The development of intensive longitudinal biomarker data has led to the development of methods to predict health outcomes and facilitate precision medicine. Intensive biomarker data is measured at a high frequency and typically results in several hundred to several hundred thousand observations per individual measured over minutes, hours, or days. In longitudinal studies, the primary focus is often on the means of trajectories, and the variances are treated as nuisance parameters, although they may also be informative for the outcomes. A Bayesian hierarchical model is proposed to jointly and simultaneously model the cross-sectional outcome and the intensive longitudinal biomarkers. To model the variability of biomarkers and deal with the high intensity of data, subject-level cubic B-splines are developed and allow the sharing of information across individuals for both the residual variability and the random effects variability. Then, different levels of variability are extracted and incorporated into the outcome probit models to make an inference. An application of the joint model is demonstrated using bio-monitor data, including hertz-level heart rate data from a study on social stress.
