Dive Brief:
A private-public partnership formed to create standards for artificial intelligence adoption in healthcare is launching a new registry for hospitals and developers to access and share information on AI tools.
The hub, known as a model card registry, was unveiled by the Coalition for Health AI (CHAI) and health system members of the group, including Providence, Cleveland Clinic and Kaiser Permanente, on Friday. The goal of the registry is to make it easier for healthcare companies to evaluate and shop between validated AI products, CHAI said.
AI companies and other developers can input their tools into the registry regardless of whether they’re a member of CHAI, according to a spokesperson. There’s no cost to upload model cards or access the registry, though it could include paywalled features in the future, they said.
Dive Insight:
AI is slowly entrenching itself in the healthcare sector, as providers increasingly turn to algorithms to automate clinical notetaking, chat with patients, streamline staff scheduling and more.
According to a survey from HIMSS conducted in the fall, 86% of respondents already leverage AI in their medical organizations, mostly for administrative tasks. But there’s growing interest in more clinical use cases for AI, like tailoring treatment plans or helping clinicians arrive at a diagnosis.
However, adoption is still nascent, with one in three medical organizations using AI saying they only started in the last six months, according to the HIMSS survey.
One issue is that it’s hard for hospitals and other companies interested in deploying the tech to keep up with the rapid rate of change in the industry and choose between the bevy of tools clamoring for their business. Procurement can be difficult, given health systems on the hunt for AI products have to analyze technical materials or marketing presentations making a variety of different claims.
Providers also want to ensure they can implement AI ethically and without patient harm, amid concerns the healthcare industry is moving too quickly to adopt AI without adequate internal or external oversight.
Enter CHAI, one of a number of groups looking to create standards around responsible AI in healthcare.
The group’s new registry is meant to standardize how information about AI tools is presented by putting model cards — documentation of an AI model’s development, capabilities and limits — in the same place.
Model cards are meant to serve the same purpose as a nutrition label for food, according to CHAI.
With the registry — which CHAI developed with Avanade, a joint venture between Microsoft and Accenture — hospitals can access model cards in one location, and makers of AI tools can get information on their products in front of potential buyers.
The model cards “play a crucial role in the AI governance process by consolidating information in an easily digestible format, facilitating product comparisons, and providing standardized data that meets the needs of various stakeholders, from radiologists to legal teams,” Elisabeth Garwood, associate chief medical informatics officer at UMass Memorial Health in Massachusetts, said in a statement.
CHAI released its model card template for public comment last year. The feedback was varied, with stakeholders wanting to delete or add information to the card depending on their industry, the CHAI spokesperson said.
CHAI hasn’t made any changes to the draft yet, but it’s still not final — the group is working to make sure the model card has appropriate fields for generative AI products, which are more difficult to oversee than predictive models.
CHAI is also working to figure out how to morph the model registry into a tool that will give users the right information at the right time, according to the spokesperson.
“This will take in-depth testing with a wide variety of stakeholders and organizations. This means testing in the actual workflow,” the spokesperson said.
Since CHAI was founded in 2021, the consortium has grown to 1,300 member organizations, including tech giants Microsoft, Google and Amazon. CHAI, which also included members of the federal government during the Biden administration, says its goal is to create a network of quality assurance labs to evaluate AI models and to develop best practices for deployment.
CHAI is one consortium spearhead by massive tech companies, AI developers and hospitals looking to fill a gap left by the federal government, which has largely pursued a hands-off approach to regulating emerging technologies.
President Donald Trump in January overturned a Biden-era order calling on agencies to create schema for overseeing AI, including in healthcare. Instead, the Trump administration has shifted to a framework of deregulation meant to prioritize U.S. innovation — especially following the shocking debut of DeepSeek, an inexpensively produced and open-source Chinese foundational model that rivals the performance of the best models created in the U.S.
Other standards groups include the Trustworthy & Responsible AI Network (TRAIN), which is also backed by Microsoft. Meanwhile Nvidia, a tech juggernaut that builds hardware essential for AI applications, is also working with clinical AI company Aidoc to create its own standards blueprint set to be released this year.
