Becker's Health IT recently published an article, "Health Systems Lack AI Governance" based on a recent report published by Center for Connected Medicine. Some key takeaways for our audience along with what Asher Informatics PBC see as Health care AI Governance needs being request by clients.
Charlotte Kalafut, CEO, Asher Informatics
As early as 2019, we began seeing #AIGovernance Committees being established at large Academic Medical Centers (AMCs). This practice accelerated as purchasing moved beyond Radiology and Cardiology to the C-suite, fueled by the excitement around generative AI applications and their increasing presence in the media.
In 2022, the discussion around #clinicalAI solutions shifted from “Will these AI solutions work?” to “How do we get them to work?”. Mid-sized healthcare systems began adopting 6-month pilots, often driven by a singular clinical champion. However, many of these pilots failed to transition to full adoption due to a lack of understanding of how these solutions support and fit into a learning healthcare system’s AI strategy and roadmap. Furthermore, the number of AI solutions for any one use-case has multiplied, making the management and oversight of many pilots unrealistic. Even pilot projects require time and effort, leading many healthcare organizations to realize the need for a better approach to AI adoption, selection, and governance. Also, without standardized success criteria and methods to evaluate ongoing efficacy, change management and confidence in AI can be challenging.
Organizations like CHAI and Health AI Partnership have begun summarizing, and sharing best practices used at AMCs and other pioneering organizations. They’ve also started to sound the alarm that AI solutions perform differently at each location. It’s not enough to test an AI solution just once. AI solution outcomes are influenced by data, how the AI is deployed and integrated into a workflow, and how the end-user uses the output. As government and regulatory bodies strive to keep up, it is likely that healthcare systems will be mandated to take-on more risk and liability considerations. This shift, along with factors like #AIAssurance infrastructure and regional assurance labs increases the complexity of healthcare leadership decisions. Systems are seeking help to navigate this ever-changing landscape. They are looking for particular help in AI strategy development, AI procurement process execution, and independent AI monitoring to support a robust, actionable AI Governance process.
“Medical artificial intelligence needs governance to ensure safety and effectiveness, not just centrally (for example, by the US Food and Drug Administration) but also locally to account for differences in care, patients, and system performance.” - W. Nicholson Price II, et al.,
Nature Machine Intelligence, August 2023
Becker's Health IT, Laura Dyrda, February 2024