Candidate: Dr. Michael Mathis is a cardiothoracic anesthesiologist with board certification in anesthesiology and advanced perioperative echocardiography at the University of Michigan. Through completion of a T32 Research Training Grant, Dr. Mathis has developed expertise in perioperative outcomes research for patients with advanced cardiovascular disease. His long-term career goal is to improve care for patients with heart failure (HF) through harnessing perioperative electronic healthcare record (EHR) data for early diagnosis and management. This proposal builds on Dr. Mathis's expertise, providing protected time for training in data science methods necessary to drive forward the analytic techniques proposed for improving HF diagnosis. Environment: The University of Michigan is the coordinating center for the Multicenter Perioperative Outcomes Group (MPOG), an international consortium of over 50 anesthesiology and surgical departments with perioperative information systems. Dr. Sachin Kheterpal, MD, MBA is the primary mentor for Dr. Mathis, and is the Director for MPOG and member of the NIH Precision Medicine Initiative Advisory Panel. The proposed research will be completed under the guidance of Dr. Kheterpal, as well as co-mentors Milo Engoren, MD, Daniel Clauw, MD, and Kayvan Najarian, PhD. An advisory panel of experts in HF diagnosis and data science methodologies will provide Dr. Mathis with additional guidance. Background: HF is among the most common chronic conditions requiring hospitalization and carries high rates of mortality. In the perioperative period, HF is a risk factor for major cardiac complications. Despite advances in care, little progress has been made to reduce HF healthcare burden, with difficulties attributable to a lack of inexpensive, reliable diagnostic measures. Consequently, patients with HF can go unrecognized in early stages and do not receive treatments to reduce mortality. The perioperative period is an underutilized opportunity to improve HF diagnosis. Beyond the wealth of preoperative data available, the intraoperative period serves as a cardiac stress test through which hemodynamic responses to surgical and anesthetic stimuli are recorded with high resolution. Yet, this data remains an untapped resource for HF evaluation. Research: The goal of the proposed research is to incorporate the perioperative period as an opportunity for early diagnosis of HF. The two specific Aims are to develop a data-driven diagnostic algorithm for HF using preoperative EHR data (Aim 1) as well as intraoperative EHR data (Aim 2).
Both aims will use automated techniques to extract features of HF from the perioperative EHR, developed at UM and scalable to multiple centers via the MPOG infrastructure. This work represents a paradigm shift in perioperative evaluation, using perioperative data as a diagnostic tool rather than a risk-assessment tool. The proposed research and training will provide Dr. Mathis with necessary data science computational experience to become an independent physician-investigator focused on improving perioperative management strategies for patients with HF.
Public Health Relevance Statement: Heart failure is a common chronic condition affecting over 550,000 patients each year, with a 45?60% mortality rate at 5 years. Despite advances in treatment, little progress has been made to improve heart failure diagnosis, and thus patients in early heart failure stages do not receive appropriate life-extending therapies. Through robust data science methodologies, this study harnesses critically underutilized, high-fidelity data available during the perioperative period to improve diagnosis of heart failure .
|Mathis, Michael R; Kheterpal, Sachin; Najarian, Kayvan (2018) Artificial Intelligence for Anesthesia: What the Practicing Clinician Needs to Know: More than Black Magic for the Art of the Dark. Anesthesiology 129:619-622|