The Vanderbilt University Medical Center and Duke University School of Medicine, in collaboration with the Coalition for Health AI and the University of Iowa, partnered to develop a design model aimed at improving health system monitoring of AI systems. The Gordon and Betty Moore Foundation contributed $1.25 million in funding to support this joint initiative.
SIGNIFICANCE OF THE PROJECT
Researchers from VUMC and Duke highlighted the growing utilization of algorithms within health systems but expressed concerns about gaps in organizational infrastructure and oversight within medical institutions. These deficiencies pose risks to the safety, fairness, and effectiveness of AI applications in healthcare.
Their goal is to pinpoint essential characteristics that health systems need to cultivate to ensure the reliable implementation of AI models.
Dr. Peter Emb, a project leader at VUMC, emphasized the importance of this endeavor, stating that it will equip the healthcare system with the necessary tools to select, develop, and oversee health AI applications, ultimately enhancing healthcare’s safety, efficacy, compassion, and equity.
By creating an evidence-based model for healthcare AI, the project aims to enable health systems to comprehensively document the algorithms in use, their principles, the overseeing entities, and the individuals responsible for their deployment.
Over the next year, the VUMC and Duke teams will outline the key components essential for the dependable integration of AI within health systems.
Nicoleta Economou, the director of algorithm-based clinical decision support monitoring at Duke AI Health, emphasized that establishing an age model for healthcare AI will empower health systems to evaluate their competencies and shortcomings in adopting and utilizing AI solutions, thereby catalyzing positive transformations in healthcare delivery.
ADDITIONAL INSIGHTS
To address trust issues surrounding AI and machine learning, CHAI advocates for a patient-centric approach in its healthcare AI framework released earlier this year. By aligning health AI standards with monitoring practices, CHAI aims to facilitate patients and healthcare providers in critically assessing the algorithms underpinning patient care.
Dr. Brian Anderson, the chief digital health practitioner at MITRE and a CHAI co-founder, stressed the criticality of accountability and trust in AI technologies that influence healthcare decisions. In August, Anderson underscored the importance of testability, transparency, and usability of AI in a guideline for secure and reliable AI deployment.
He further mentioned collaborating with the White House, state agencies, and various stakeholders to develop metrics, measurements, and tools for managing the lifecycle of AI models. Anderson acknowledged the rapid pace of AI advancements, expressing concerns about the lack of established safeguards necessary for these evolving technologies, particularly in the healthcare domain.
