PROJECT ABSTRACT Almost half of the 1.7 million hospice enrollees have Alzheimer's disease and related dementias (ADRD) as the principal diagnosis or a comorbidity. Although hospice care is particularly beneficial to support ADRD patients and their caregivers at the end of life (EoL), one in five hospice enrollees with ADRD experience discharge from hospice prior to death, also known as “live discharge.” Hospice live discharge has a profound impact on EoL care of ADRD patients due to their complex EoL conditions, highly uncertain prognosis, and difficulty in communicating EoL care preference. Despite two recent policy changes, hospice live discharges remain persistently high among ADRD patients. The changing hospice market landscape may keep boosting live discharges among ADRD patients and lead to frequent care transitions and other adverse EoL outcomes after hospice live discharge. To date, critical gaps exist in understanding outcomes after hospice live discharge among ADRD patients. Little is known about the longitudinal patterns of care transitions in all care settings, functional decline, and overall healthcare utilization after hospice live discharge. Although Medicare Advantage (MA) covers half of Medicare beneficiaries, there is scant evidence about outcomes after hospice live discharge among MA enrollees. No validated tools are available to predict patient outcomes after hospice live discharge to facilitate discharge planning and post-discharge care coordination. This mixed-methods project aims to fill these gaps by characterizing and predicting patient trajectories after hospice live discharge among ADRD patients. In Aim 1, we will apply longitudinal clustering methods with national Medicare fee-for-service (FFS) claims and MA encounter data to identify patient subgroups with heterogenous trajectories after hospice live discharge. In Aim 2, we will develop and validate prediction models for patient trajectories and other clinically important outcomes after hospice live discharge. We will use Medicare FFS claims, MA encounter data, electronic health record (EHR) data from two large hospices, and linked claims/encounter data and EHR data to develop prediction models and ensure fairness of model accuracy by patient race/ethnicity and socioeconomic status. In Aim 3, we will conduct in-depth qualitative interviews with different stakeholders (e.g., physicians, patients and caregivers, and informatics staff) who are involved in the implementation of the prediction models in care delivery. Evidence from this aim will facilitate the transition from model development to real-world implementation in the future that can benefit ADRD patients and their caregivers. This project is well aligned with the National Institute on Aging's priority to better understand the burden of ADRD on patient outcomes. The proposed research will generate novel evidence to improve patient outcomes after hospice live discharge and inform the ongoing pilot programs to de...