The rise of Artificial Intelligence (AI) advancement and adoption is prompting health care leaders, stakeholders, and policymakers to examine its potential to be leveraged to improve health outcomes in a fair, equitable, and scalable way. Connecting for Better Health sees the critical intersection of data sharing and health AI to be an important, nuanced partnership that will shape policy and technology decisions.
Grounding our Understanding of Health AI
Connecting for Better Health (C4BH) hosted a Health AI workshop on October 24th to assess current federal and state actions on AI and health care, examine linkages between health and social service data sharing and AI, and inform C4BH’s engagement in health AI moving forward. Current deployments of AI in health care exist along a spectrum: from simple AI, or training computers to recognize patterns and solve complex problems akin to humans, to Generative AI, or advanced models trained on vast data sets that are able to detect patterns and create outputs accordingly without much human guidance. While they hold great potential, AI models in health care can carry a great risk of exacerbating existing biases, depending on the data that models are trained on.
Limiting bias and increasing transparency in AI are high priorities for California’s legislature, as indicated by the 17 bills signed by Governor Gavin Newsom this year regarding Generative AI deployment and regulation. At the federal level, an executive order was set in place in late 2023 to govern safe, secure, and trustworthy development and use of AI. Looking ahead, the U.S. Department of Health and Human Services is anticipated to release their AI Strategic Plan in early 2025.
Highlighting AI and Health Care Research
C4BH welcomed Dr. Kamal Jethwani, managing partner and CEO of Decimal.health and participant in the VALID AI initiative, to speak upon their work in creating data-driven solutions for social determinants of health (SDOH). In addition, Dr. Ahmad Alkasir, of the Ellison Institute of Technology, and Jolie Ritzo, of Civitas Networks for Health, co-presented on the role of using existing health information exchanges (HIEs) to support AI advancements.
VALID AI is a project led by UC Davis and hosts over 50 health care organizations nationwide to work on advancing Generative AI validation and governance, specifically for social services and data. Dr. Jethwani gave an overview of one of the initiative’s workstreams: the SDOH Accelerator Program. The program aims to incorporate social data to be used as vital signs in clinical decision-making. By capitalizing on Generative AI, the program is piloting how to aggregate data that is formatted differently, contains non-standardized information, or is sent in various ways to be effective in decision making and ensure that social data is incorporated during patient care. While the variance in social data formats and content presents challenges, leveraging social data is crucial to whole-person care. If AI can be utilized to facilitate better data collection and integration of both health and social services, providers should have a more complete picture of the individuals in front of them, improving patient care and health outcomes.
Published in September 2024, “The Role of Health Data Utilities in Supporting Health AI,” is co-authored by the Ellison Institute for Technology and Civitas Networks for Health. The article argues that in order to use AI as a tool to improve health outcomes, we must improve and modernize our health systems, regulations, and governance structures to build a robust, equitable AI model. Currently, Health Information Exchanges (HIEs) are uniquely positioned to facilitate the use of equitable and reliable AI due to their existing infrastructure that is able to aggregate copious amounts of multi-sourced, representative, and local health data. Dr. Ahmad Alkasir and Jolie Ritzo discussed how HIEs can play a critical role for local data governance of AI. By integrating health and social data, HIEs could serve as the stewards of AI models and ensure that datasets are securely and appropriately provided to developers to foster an inclusive and competitive AI landscape. By creating a level playing field, small and medium-sized organizations have the capacity and opportunity to develop AI models. In turn, AI models can be more equitable and inclusive of the unique health and social needs across regions.
Transparency and trust are at the center of all considerations for safe AI in health care. The work to ensure appropriate use of the cutting-edge models requires stakeholders across HIEs, health systems, and state and federal agencies to foster a fair and transparent AI framework. As health care continues to advance and adopt new technology, it is critical to advocate for responsible data stewardship and focus on creating robust governance frameworks that advance health equity and better outcomes.
