PROJECT SUMMARY Significance. Over 726,000 individuals in the U.S. have end-stage renal disease (ESRD) with about 87% of patients on dialysis, a life-sustaining treatment. Dialysis patients experience accelerated mortality and frequent hospitalizations (twice per year). These adverse outcomes are exacerbated at key time periods after transition to dialysis and show significant spatial variation across U.S. Our overarching goal is to identify modifiable patient, dialysis facility and region-level (spatial) risk factors and critical time periods for elevated hospitalization risk and mortality to guide patient care strategies. To reach this goal, we propose novel multilevel time-dynamic models applied to data from the national database, United States Renal Data System (USRDS), which contains data on nearly all patients on dialysis in the U.S. We consider three-level longitudinal data with longitudinal outcomes nested in patients, patients nested in dialysis facilities and facilities nested in geographic regions across U.S. The proposals include linking of multiple data domains to study effects from time periods prior to (prelude) and after (vintage) transition to dialysis and accommodate time varying effects of multilevel risk factors at the patient-, facility- and region-levels. Aims and Innovation. We propose three specific aims: Aim 1) To develop and apply multilevel spatiotemporal functional models (MST-FMs) of hospitalization and mortality rates. Initial efforts for identifying spatial “hot spots” with higher hospitalization and mortality rates have been largely descriptive. In addition, these approaches do not consider the critical temporal variation across time after transition to dialysis. We will develop estimation and inference procedures to model the spatially nested functional data of facility-level hospitalization and mortality rates for constructing spatiotemporal maps, identifying hot spot regions and post dialysis transition time periods of elevated hospitalization and mortality risk in the dialysis population for the first time. Aim 2) To develop and apply multilevel time-varying joint models (MT-JMs) of longitudinal hospitalizations and survival outcomes at the patient-level. The proposed MT-JMs will incorporate functional predictors from the prelude period, time-varying effects of both cross-sectional and time-dependent multilevel covariates as well as multilevel random effects. To date, joint modeling approaches mostly consider one-level hierarchies encompassing longitudinal outcome observed for a set of subjects, with only a few works considering a two-level hierarchy, intended for modeling time-static (fixed) covariate effects. The proposed MT- JMs address the need for flexible modeling features (to assess prelude data, monthly clinical longitudinal data post-transition, multilevel time-varying effects etc.), incorporation of three-level multilevel data structure, and scalability to complex large datasets. Aim 3) To characteriz...