Marcio Aloisio Bezerra Cavalcanti Rockenbach, MD
Massachusetts General Hospital
Boston, Massachusetts
Disclosure information not submitted.
David Buric, n/a
Critical Care Medicine Fellow
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Jeffrey Carness, MD
Critical Care Medicine Fellow
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Caitlin Burke, MD,
Acute Care Surgery Fellow
Hartford Hospital
Farmington, Connecticut, United States
Disclosure information not submitted.
Jakob Wollborn, n/a
Clinical Fellow in Cardiothoracic Anesthesiology
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Saniya Sami, n/a
Critical Care Medicine Fellow
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Varun Buch
Director of AI Product Development
Massachusetts General Hospital, United States
Disclosure information not submitted.
Quanzheng Li
Associate Professor
Harvard Medical School, United States
Disclosure information not submitted.
Nir Neumark, n/a
Product Manager
Mass General Brigham, United States
Disclosure information not submitted.
Romane Gauriau, n/a
Senior Machine Learning Scientist
Center for Clinical Data Science, United States
Disclosure information not submitted.
Dufan Wu
Instructor in Investigation
Massachusetts General Hospital, United States
Disclosure information not submitted.
Hariharan Ravishankar
Senior Scientist
GE Healthcare, United States
Disclosure information not submitted.
Rohit Pardasani
Senior Data Scientist
GE Healthcare, United States
Disclosure information not submitted.
Abhijit Patil
Senior Director - AI & Analytics
GE Healthcare, United States
Disclosure information not submitted.
Dirk Varelmann, MD, EDIC, DESA
Anesthesiologist
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Pankaj Sarin, n/a
Anesthesiologist
Brigham & Women's Hospital, United States
Disclosure information not submitted.
Title: Comparing AI and Clinicians in Oxygenation Device Management for Suspected COVID-19 ICU Patients
INTRODUCTION: Deciding the appropriate ventilation regimen is one of the most common and challenging tasks in an Intensive Care Unit (ICU). Clinical decision support tools that continuously analyze patient data can assist in improving patient management. We compare the performance of a machine learning model in providing recommendations for oxygenation device management for patients admitted to the ICU with clinician’s assessment.
Methods: Clinical data from 10 patients admitted due to respiratory illness (suspected or confirmed COVID-19) was retrospectively collected throughout their stay at the ICU. Data included demographics, comorbidities, vital signs, lab tests results, and current oxygenation device being used (categorized into room air, low flow oxygen, high flow oxygen and mechanical ventilation). Timepoints where the current oxygenation device was mechanical ventilation were excluded. Readers (two ICU attendings and seven clinical fellows) used a web-browser application to evaluate the available clinical data at every hour and to provide a clinical recommendation for the following hour (step down, maintain or step up the oxygenation device). For each timepoint, the most frequent recommendation among the readers was considered the group consensus. The consensus recommendation was compared to the predictions of a machine learning model trained on a similar dataset of 990 patients. The ground truth was based on the actual oxygenation device utilized.
Results: 73 time points were considered in the final evaluation, including 31 transitions (step up or step down) and 42 ‘maintain’ decisions. The readers’ consensus showed 61.2% sensitivity, 95.2% specificity and 80.2% accuracy for the prediction of a transition event, while the machine learning model demonstrated, respectively, 80.6%, 64.3% and 71.2%.
Conclusions: The trained model demonstrated promising performance in using clinical data to provide recommendations for patient management. Integration of such a tool into clinical workflow could potentially benefit patients through the continuous monitoring of data recorded in electronic health records and by providing recommendations to the clinical team minutes or hours prior to a potential change in patient’s condition.