“Artificial intelligence (AI) holds great promise in helping health care providers gain insights and improve health outcomes,” reports the American Hospital Association. The national organization further asserts that “the use of AI in clinical care is no longer in its infancy and is expected to experience exponential growth in the coming years.” However, realizing AI’s potential requires training and testing algorithms with clinical data, which hasn’t been readily available — especially for patients in perioperative settings.
Now, researchers have access to a new repository of publicly available data for 58,799 patients who have undergone surgeries at the UC Irvine Medical Center. The Medical Informatics Operating Room Vitals and Events Repository (MOVER) is the product of a team of researchers from UCI and UCLA. Leading the effort at UCI is Pierre Baldi, Distinguished Professor of UCI’s Donald Bren School of Information and Computer Sciences (ICS), and Dr. Joe Rinehart, a clinical professor of anesthesiology at UCI. They are working in close collaboration with Dr. Maxime Cannesson, professor and chair of anesthesiology and perioperative medicine at the David Geffen School of Medicine at UCLA.
Bridging the Gap
The goal is to enable effective AI training, leading to the deployment of novel predictive and diagnostic healthcare tools.
“MOVER is a data repository that contains detailed information about more than 80,000 surgeries performed at the UCI Medical Center, including information about the surgeries as well as various patient vital wave forms, such as heart rate,” says Baldi. “There are extremely few data sets of this kind that are publicly available, which hinders research.”
The repository, hosted at UCI, was created by a team of researchers that includes Muntaha Samad and Mirana Claire Angel, former graduate students in Baldi’s research group, as well as Yuzo Kanomata, a project scientist in the UCI Institute for Genomics and Bioinformatics (IGB). In building MOVER, care was taken to ensure patient privacy and HIPAA compliance, stresses Baldi, who serves as IGB director. “The data is de-identified so that the original information about patient names and the dates of the procedures is removed or scrambled.”
Sharing the Data
The seven-year effort to acquire and process the data needed to create MOVER is outlined in “Medical Informatics Operating Room Vitals and Events Repository (MOVER): A Public-Access Operating Room Database,” a paper recently published in JAMIA Open. The authors detail the various datasets in MOVER and outline potential uses — such as real-time predictions to assist anesthesiologists, or outcome predictions to shed light on postoperative complications.
“These data sets,” explains Baldi, “are essential for developing various AI/machine learning/statistical methods for predicting adverse events and outcomes, improving downstream healthcare.”
Researchers and qualified physicians can freely gain access to MOVER and its electronic health records and high-fidelity physiological waveform data by signing a data usage agreement. Leveraging this data for AI testing and training will help foster AI innovation in healthcare, improving patient care and outcomes.
— Shani Murray