Implementation of Continuum of Care Sepsis Phenotyping and Risk Stratification

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT This proposal outlines a 5-year research and career development plan for Dr. Gabriel Wardi, an emergency medicine intensivist and assistant professor at UCSD. The major objective of his research is the effective implementation of deep-learning algorithms to clinical practice to improve care of sepsis patients. This K23 proposal outlines and provides support for his career development plan, specifically focusing on (1) the ability to design meaningful sepsis studies and necessary statistical training, (2) strong understanding of machine- learning approaches, and (3) a focus on implementation science to improve care of sepsis patients with novel deep-learning algorithms. Dr. Wardi has assembled a diverse team of collaborative experts to support his career development and mentor him consisting of Dr. Atul Malhotra, an internationally recognized expert in critical care physiology and respiratory failure along with Dr. Shamim Nemati, a machine-learning expert with a strong focus in prediction of sepsis in real-time. Additionally, his training team includes experts in implementation science from the Dissemination and Implementation Science Center (DISC) at UCSD as well as an expert in clinical trial design and biostatistics (Dr. Sonia Jain). Despite decades of research, sepsis remains a major public health challenge. Current approaches to sepsis care emphasize “one-size fits all” bundles that may result in patient harm in certain subgroups. Newer approaches to data analysis, using multiple layers of non-linear arithmetic operations now allow for clustering of sepsis patients into novel clinical phenotypes that may provide for more personalized care. The PI will evaluate potential phenotypes of sepsis not present on admission (NPOA) in Aim 1. Prior investigations into phenotyping have been developed and validated in patients present in the emergency department. Patients with sepsis NPOA have high mortality and better quantification of phenotypes may help improve care by identifying novel groups. Dr. Wardi seeks to evaluate 2 inter-related hypotheses in this aim: one is that phenotypes may represent disease trajectories that are modifiable by accepted therapies (e.g. time to, and quantity of fluid resuscitation). The second is that novel phenotypes exist in the inpatient setting. In his second aim, Dr. Wardi seeks to determine clinical mechanisms of 30-day readmissions in sepsis patients through a variety of approaches, including identification of novel clusters of sepsis patients at discharge and use of natural language processing of a large data set to identify actionable reasons for readmissions. Finally, he seeks to determine if the application of a wearable patch to sepsis patients discharged to a long-term acute care hospital when combined with a machine-learning algorithm may reduce unanticipated 30-day sepsis readmissions. This research and career development plan affords Dr. Wardi an impressive foundation to develop into a prominent clinician-scientist working to improve care by developing and implementing novel approaches to detection and classification of sepsis patients. Dr. Wardi is fully committed to improving the care of sepsis patients by embracing innovative strategies.