This proposal is to support a robust human cohort enrolling subjects with acute respiratory distress syndrome (ARDS), pneumonia, or sepsis (collectively termed APS) as part of the APS Consortium, and to identify clinical and molecular features that better predict, stratify, and explain organ failure, mortality, and disability following APS. We hypothesize that distinct and reproducible molecular subtypes are common and detectable across all 3 APS syndromes, and that we will identify the pathways that maximally contribute to organ failure and recovery trajectory through this well-powered molecular cohort. As part of the Consortium-Wide Longitudinal Cohort study, we propose to develop new tools for risk assessment, stratification, and recovery from APS. In aim 1a, we will use joint modeling to integrate multiple plasma markers of inflammation, vascular dysregulation, sarcopenia, and neural injury to identify the combinations most associated with organ failure, infer which plasma intermediates might contribute causally to organ failures, and identify the proportion of mortality risk mediated by specific organ failures. In aim 1b we focus on better prediction of long term disability post-APS, testing the ability of Katz- and Lawson-informed functional status features to predict persistent disability and asking whether prediction of disability is enhanced by added plasma markers of inflammation, vascular injury, or neuromuscular injury. Our Center-specific aims employ novel molecular phenotyping to better explain organ failure, death, and disability post-APS. In aim 2a we test specific hypothesis-driven candidate markers of immune dysregulation as potential organ failure markers. Aim 2b identifies patterns of host immune health during recovery using high dimensional flow cytometry to understand the peripheral blood host immune response, and asks whether immune cell trajectory associates with disability or recovery. Aim 2c focuses on vascular injury markers, and asks which components of vascular injury associate with specific organ failures and with post-APS disabilities. In aim 3, we integrate multiple streams of biologic data and identify patterns of response across APS acutely and during recovery, use machine learning to select the most informative features for joint modeling, and test the performance of the model of acute or long term disability in different APS states. Our site will make a lasting contribution to the APS Consortium and our completed aims will advance the prevention and personalized treatment of ARDS, pneumonia, and sepsis to improve overall health.