One of the most promising announcements in healthcare and emerging tech is the recent collaboration between Duke University, UC Berkeley, Mayo Clinic, DLA Piper, and other key stakeholders in the industry.
With the abundance of healthcare software that relies on artificial intelligence (AI) and machine-learning algorithms, this collaboration – Health AI Partnership – is designed to meet the changing paradigm of healthcare technology. Due to the “wild west” nature of the healthcare software market and some of the inherent challenges that come with designing bias-free algorithms, the collaboration is intended to create a standardized curriculum for healthcare providers.
Goals of the Health AI Partnership
Unveiled at the recent HIMSS Machine Learning & AI for Healthcare, the new Health AI Partnership is designed as an innovation and learning network to address big challenges of new algorithms for all use cases in a rapidly-accelerating space.
Ranging from software procurement, deployment, and lifecycle management, some of the major goals of this partnership include:
- Create an open dialogue on how stakeholders can take bias into consideration when developing AI and machine-learning algorithms
- Reduce systemic bias into care delivery, such as structural inequalities and racism
- Setting standards for software vendors to adhere to regarding performance claims versus real-world performance
- Discuss workable approaches for how to develop and integrate AI software in clinical practice
- Reduce the cost of healthcare to at-risk patients and lessen the financial burden placed on providers
- Decentralize the high concentration of technology to develop guidelines to help other organizations make smarter decisions
- Equip IT decision-makers in healthcare with academic research and vetted guidelines to investment in dependable technologies.
- Aid physicians and other decision-makers to better serve patients
- Set standards for how healthcare staff should be trained
- Establish guidelines for legal accountability of AI and machine-learning regarding regulatory agencies and state law enforcement officials
To help proliferate the findings to healthcare and related stakeholders, the guidance and curriculum developed by the Health AI Partnership will be available online and open-source.
Timeline of the Collaboration
The scope of the project is expected to evolve over the course of 12 months in 2022.
Several phases of the collaboration have been planned:
First Phase – Information Gathering
The first phase will encompass a thorough examination of decision-making in healthcare settings, particularly with a focus on how lifecycle management programs are already being used in the pharmaceutical industry. The team will analyze how teams are staffed, decisions are made, and other processes that have applications throughout healthcare.
Additionally, there will be a series of interviews and observations within Duke and within Mayo Clinic. Once these are completed, more information will be gathered across academic medical centers, community hospitals and payer organizations.
Second Phase – Defining a Curriculum
By midyear, the collaboration is expected to define priorities for how the curriculum should be developed. With material available online, stakeholders can request for comments and participate to address all conceivable use cases. This second phase will work with all relevant stakeholders, including target users (ie. physicians), commercial payers, policymakers, regulators, and more.
Third Phase – Testing the Curriculum
At nine months into the collaboration, a curriculum will be created to facilitate the first round of user testing. Through analyzing data and feedback, the Health AI Partnership will understand the next steps needed to shape the practices of technology developers to start supporting the needs of the health system that are making these decisions.
Who’s Involved in the Health AI Partnership
Spearheading this new collaboration is Dr. Mark Sendak, population health & data science Lead at Duke Institute for Health Innovation. Dr. Sendak is joined by other leaders that offer to bring a different set of expertise. Other key participants include:
- Suresh Balu, associate dean for innovation and partnerships at Duke Health
- Michael Gao, senior data scientist at Duke Health
- Deirdre Mulligan, professor at UC Berkeley’s School of Information
- Deb Raji, a Ph.D. candidate at UC Berkeley
- Dr. Ziad Obermeyer, associate professor at UC Berkeley
- David Vidal, director of regulation at the Mayo’s Center for Digital Health
- Mark Lifson, SaMD systems engineering manager at Mayo’s Center for Digital Health
- Mohammad Ghassemi, assistant professor at Michigan State University,
- Shannon Harris, assistant professor at Virginia Commonwealth University
- Dr. Danny Tobey, global lead for AI practice at law firm DLA Piper
- Karen Silverman, HIMSS Outside Counsel
- And many others
To create a more comprehensive implementation, other groups and key stakeholders are encouraged to provide input to create an industry-wide consensus on the future of AI and machine-learning in healthcare. Groups such as the Joint Commission, Centers for Medicare & Medicaid Services (CMS), and the FDA are expected to collaborate throughout this process.