A framework for the oversight and local deployment of safe and high-quality prediction models.
Date
2022-08
Journal Title
Journal ISSN
Volume Title
Repository Usage Stats
views
downloads
Citation Stats
Attention Stats
Abstract
Artificial intelligence/machine learning models are being rapidly developed and used in clinical practice. However, many models are deployed without a clear understanding of clinical or operational impact and frequently lack monitoring plans that can detect potential safety signals. There is a lack of consensus in establishing governance to deploy, pilot, and monitor algorithms within operational healthcare delivery workflows. Here, we describe a governance framework that combines current regulatory best practices and lifecycle management of predictive models being used for clinical care. Since January 2021, we have successfully added models to our governance portfolio and are currently managing 52 models.
Type
Department
Description
Provenance
Subjects
Citation
Permalink
Published Version (Please cite this version)
Publication Info
Bedoya, Armando D, Nicoleta J Economou-Zavlanos, Benjamin A Goldstein, Allison Young, J Eric Jelovsek, Cara O'Brien, Amanda B Parrish, Scott Elengold, et al. (2022). A framework for the oversight and local deployment of safe and high-quality prediction models. Journal of the American Medical Informatics Association : JAMIA, 29(9). pp. 1631–1636. 10.1093/jamia/ocac078 Retrieved from https://hdl.handle.net/10161/33637.
This is constructed from limited available data and may be imprecise. To cite this article, please review & use the official citation provided by the journal.
Collections
Scholars@Duke
Armando Diego Bedoya
Nicoleta Economou-Zavlanos
Director of Duke Health AI Evaluation and Governance
Founding Director of Algorithm-Based Clinical Decision Support (ABCDS) Oversight
Nicoleta Economou-Zavlanos, PhD, is the Director of the Duke Health AI Evaluation & Governance Program and the founding director of the Algorithm-Based Clinical Decision Support (ABCDS) Oversight initiative. In this capacity, she leads Duke Health’s efforts to evaluate and govern health AI technologies. Dr. Economou also serves on the Executive Committee of the NIH Common Fund’s Bridge to Artificial Intelligence (Bridge2AI) Program. Additionally, she served as Scientific Advisor for the Coalition for Health AI (CHAI), driving the development of guidelines for AI assurance in healthcare, from 2024 to 2025.
A nationally recognized expert in health AI governance, Dr. Economou has been instrumental in creating frameworks and methodologies for the registration, review, and assurance of health AI systems. Her research, published in leading journals such as NPJ Digital Medicine, JAMA, JAMA Health Forum, and JAMIA, reflects her commitment to advancing the responsible development and use of AI in healthcare.
Benjamin Alan Goldstein
I study the meaningful use of Electronic Health Records data. My research interests sit at the intersection of biostatistics, biomedical informatics, machine learning and epidemiology. I collaborate with researchers both locally at Duke as well as nationally. I am interested in speaking with any students, methodologists or collaborators interested in EHR data.
Please find more information at: https://biostat.duke.edu/goldstein-lab
John E Jelovsek
Dr. Jelovsek is the F. Bayard Carter Distinguished Professor of OBGYN at Duke University and serves as Director of Data Science for Women’s Health. He is Board Certified in OBGYN by the American Board of OBGYN and in Female Pelvic Medicine & Reconstructive Surgery by the American Board of OBGYN and American Board of Urology. He has an active surgical practice in urogynecology based out of Duke Raleigh. He has expertise as a clinician-scientist in developing and evaluating clinical prediction models using traditional biostatistics and machine learning approaches. These “individualized” patient-centered prediction tools aim to improve decision-making regarding the prevention of lower urinary tract symptoms (LUTS) and other pelvic floor disorders after childbirth (PMID:29056536), de novo stress urinary incontinence and other patient-perceived outcomes after pelvic organ prolapse surgery, risk of transfusion during gynecologic surgery, and urinary outcomes after mid-urethral sling surgery (PMID: 26942362). He also has significant expertise in leading trans-disciplinary teams through NIH-funded multi-center research networks and international settings. As alternate-PI for the Cleveland Clinic site in the NICHD Pelvic Floor Disorders Network, he was principal investigator on the CAPABLe trial (PMID: 31320277), one of the largest multi-center trials for fecal incontinence studying anal exercises with biofeedback and loperamide for the treatment of fecal incontinence. He was the principal investigator of the E-OPTIMAL study (PMID: 29677302), describing the long-term follow up sacrospinous ligament fixation compared to uterosacral ligament suspension for apical vaginal prolapse. He was also primary author on research establishing the minimum important clinical difference for commonly used measures of fecal incontinence. Currently, he serves as co-PI in the NIDDK Symptoms of Lower Urinary Tract Dysfunction Research Network (LURN) (U01DK097780-05) where he has been involved in studies in the development of Symptoms of Lower Urinary Tract Dysfunction Research Network Symptom Index-29 (LURN SI-29) and LURN SI-10 questionnaires for men and women with LUTS. He is also the site-PI for the PREMIER trial (1R01HD105892): Patient-Centered Outcomes of Sacrocolpopexy versus Uterosacral Ligament Suspension for the Treatment of Uterovaginal Prolapse.
Cara Louise O'Brien
Erich Senin Huang
Former Chief Data Officer for Quality, Duke Health
Former Director of Duke Forge
Former Director of Duke Crucible
Former Assistant Dean for Biomedical Informatics
Dr. Huang is currently Associate Chief Clinical Officer for Informatics & Technology at Verily (Google's life sciences subsidiary), and is now adjunct faculty at Duke. Dr. Huang’s research interests span applied machine learning, research provenance and data infrastructure. Projects include building data provenance tools funded by the NIH’s Big Data to Knowledge program, regulatory science funded by the Burroughs Wellcome Foundation. Applied machine learning applications include “Deep Care Management” a highly interdisciplinary project with Duke Connected Care, Duke’s Accountable Care Organization, that integrates claims and EHR data for predicting unplanned admissions and risk stratifying patients for case management; CALYPSO, a collaboration with the Department of Surgery for utilizing machine learning to predict surgical complications. My team is also building the data platform for the Department of Surgery's "1000 Patients Project" an intensive biospecimen and biomarker study based around patients undergoing the controlled injury of surgery.
As Director of Duke Forge, Dr. Huang built a data science culture and infrastructure across Duke University that focused on actionable health data science. The Forge emphasized scientific rigor, awareness that technology does not supersede clinicians’ responsibilities and human relationship with their patients, and the role of data science in society.
Eric Gon-Chee Poon
I currently serve as the Chief Health Information Officer for Duke Medicine. I also practice primary care internal medicine at the Durham Medical Center as part of Duke Primary Care. In my capacity as CHIO, I am responsible for the visioning and strategic planning of clinical and analytic information systems that impact patient care, research and education. I work with the Duke Medicine leadership to ensure technology solutions are well aligned with our overall organizational objectives. I oversee the optimization of the Maestro Care (Epic) electronic health record, and partner with physicians, patients and operational leaders to effectively leverage innovative IT in support of the Duke mission. I also have a keen interest in IT innovation, and work with investigators across Duke to pursue new and innovative ways to efficiently deliver high quality care to our patients.
My research interests have revolved around the use of health information technology to improve the quality of care and patient safety in both the ambulatory and hospital settings. My work in the ambulatory setting has focused on the efficient delivery of decision support to clinicians to prevent errors of omission and commission during diagnostic test ordering and review of test results. I have also worked to use information technology, including secure on-line patient portals, to improve the communication between clinicians and patients around health maintenance and the follow-up of abnormal test results. In the inpatient setting, I have conducted several studies to delineate the barriers to and facilitators of the wide-spread diffusion of computerized physician order entry and have led many studies evaluating the safety, financial and socio-technical impact of barcode technology in the hospital pharmacy and nursing units.
Michael J Pencina
Michael J. Pencina, PhD
Chief Data Scientist, Duke Health
Vice Dean for Data Science
Director, Duke AI Health
Professor, Biostatistics & Bioinformatics
Duke University School of Medicine
Michael J. Pencina, PhD, is Duke Health's chief data scientist and serves as vice dean for data science, director of Duke AI Health, and professor of biostatistics and bioinformatics at the Duke University School of Medicine. His work bridges the fields of data science, health care, and AI, contributing to Duke’s national leadership in responsible health AI.
Dr. Pencina partners with key leaders to develop data science strategies for Duke Health that span and connect academic research and clinical care. As vice dean for data science, he develops and implements quantitative science strategies to support the School of Medicine’s missions in education and training, laboratory and clinical science, and data science.
He co-founded and co-leads the national Coalition for Health AI (CHAI), a multi-stakeholder effort whose mission is to increase trustworthiness of AI by developing guidelines to drive high-quality health care through the implementation of innovative, credible, and transparent health AI systems. He serves in a leadership capacity for the Trustworthy & Responsible AI Network (TRAIN), a new organization Duke co-founded with leading health care and technology organizations to develop tools and technologies that promote the adoption of high-quality, novel, and safe health AI solutions for patient care and research. He also spearheaded establishing and co-chairs Duke Health’s Algorithm-Based Clinical Decision Support (ABCDS) Oversight Committee.
Dr. Pencina is an internationally recognized authority in the evaluation of AI algorithms. Guideline groups rely on his work to advance best practices for the application of clinical decision support tools in health delivery. He interacts frequently with investigators from academic and industry institutions as well as government officials. Since 2014, Thomson Reuters/Clarivate Analytics has regularly recognized Dr. Pencina as one of the world’s "highly cited researchers" in clinical medicine and social sciences, with more than 400 publications cited over 135,000 times. He serves as a deputy editor for statistics at JAMA-Cardiology.
Dr. Pencina joined the Duke University faculty in 2013, and served as director of biostatistics for the Duke Clinical Research Institute until 2018. Previously, he was an associate professor in the Department of Biostatistics at Boston University and the Framingham Heart Study, and director of statistical consulting at the Harvard Clinical Research Institute. He received his PhD in Mathematics and Statistics from Boston University in 2003 and holds master’s degrees from the University of Warsaw in actuarial mathematics and business culture.
Email: michael.pencina@duke.edu
Web Sites: medschool.duke.edu; aihealth.duke.edu; https://scholars.duke.edu/person/michael.pencina
Phone: 919.613.9066
Address: Duke University School of Medicine; 2424 Erwin Road, Suite 903; Durham, NC 27705
Unless otherwise indicated, scholarly articles published by Duke faculty members are made available here with a CC-BY-NC (Creative Commons Attribution Non-Commercial) license, as enabled by the Duke Open Access Policy. If you wish to use the materials in ways not already permitted under CC-BY-NC, please consult the copyright owner. Other materials are made available here through the author’s grant of a non-exclusive license to make their work openly accessible.