Abstract The purpose of the proposed research is to develop a suite of flexible statistical models and computationally scalable inferential methods to understand how variability of biomarkers may be associated with future health outcomes. While variance is typically understood as nuisance – the “noise” in “signal-to-noise” – there is increasing evidence that underlying variability in subject-level measures over time may also be important in predicting future health outcomes of interest. Previous work in this area has focused on using repeated measures on predictors, one-at-a-time, to develop subject-level mean and variance estimates to use as predictors in joint models of binary outcomes. The technological advances in scientific measurements have resulted in biomarkers that are multivariate, mixed-scale, and obtained at increasingly higher time resolutions. While the scope of scientific questions involving the use of biomarkers in clinical studies has greatly expanded, statistical method development has not kept apace. The proposed research will extend existing work to model time-to-event outcomes, to perform multi- outcome modeling of both scalar and multivariate outcomes as functions of multiple sets of longitudinal predictors, and to deal with high dimensional longitudinal predictors such as those provided by repeated long-term surveillance in prospective cohort studies and by biometric ecological momentary assessment (EMA) measures.