Although machine learning applications are now pervasive to every industry, adoption into healthcare remains a challenging and arduous process. Barriers to implementation include clinician trust, algorithm credibility and actionability, promoting clinician literacy in machine learning methods, and mitigating unintended consequences.
In the high-risk operating room setting, anesthesiologists are recognized leaders in patient safety, and manage uncertainty through careful considerations of risk and benefit based upon a thorough understanding of disease processes and treatment mechanisms. In this talk, the speaker highlights how obstacles to implementation of machine-learning based healthcare applications can be mitigated, and how an understanding of such applications can be promoted among clinically-minded anesthesiologists who may not necessarily be expert data scientists.
Dr. Mathis has research interests in improving perioperative care for patients with advanced cardiovascular disease, particularly for patients with heart failure. As part of the Multicenter Perioperative Outcomes Group (MPOG), an international consortium of perioperative databases for which U-M serves as the coordinating center, he serves as Associate Research Director and plays a lead role in integration of MPOG data with data from national cardiac and thoracic surgery registries. He also has interests in leveraging novel data science methods to understand patterns within highly granular intraoperative physiologic data, studying hemodynamic responses to surgical and anesthetic stimuli as a means for early detection of cardiovascular diseases such as heart failure.