Extracorporeal life support (ECLS) rescues critically ill patients with refractory cardiopulmonary failure. Hemorrhage and thrombosis are the most significant causes of morbidity and mortality in ECLS; stroke and intracerebral hemorrhage (ICH) are the most devastating among them. The use of ECLS has exploded, however studies to date have been unable to predict these complications. Amyloid beta (Al3) plays a central role in cerebral hemostasis and may contribute to neurologic complications during ECLS. Cerebral oximetry via near-infrared spectroscopy (NIRS) can detect cerebral ischemia. The long-term goal of the Pl's research is to harness rigorous bioinformatic techniques to identify and validate novel markers of hemorrhage and thrombosis in ECLS to improve outcomes. The Pl's central hypothesis is that machine learning will identify key and modifiable risk factors for bleeding and thrombosis and that incorporating innovative neurophenotyping (Al3 levels and NIRS) will enhance network prediction for neurological complications. The Pl will test this hypothesis by pursuing the following three aims: Aim 1: Generate and cross-validate machine learning models to predict hemorrhage (including ICH) within ELSO and collaborative datasets Aim 2: Generate and cross-validate machine learning models to predict thrombosis (including ischemic stroke) within ELSO and collaborative datasets Hypothesis: Novel networks of patient-, clinical-, and time-based factors will emerge as strong predictors of hemorrhage (Aim 1), and thrombosis (Aim 2), respectively.