The Coalition for Health AI (CHAI) is working with the National Association of Community Health Centers (NACHC) to help bring artificial intelligence to the nation’s largest primary care system.
CHAI’s efforts with the NACHC will arm AI developers and front-line clinicians with the knowledge to deploy and govern AI in safety net settings, where 90% of the population served is 200% below the federal poverty level.
The partnership will give community health centers (CHCs) a seat at the table as AI vendors seek to create responsible and trustworthy AI. CHAI CEO Brian Anderson, M.D., also said the organization has discussed offering discounts to safety net clinics with its member companies.
“Part of the partnership with CHAI is going to look at this priority of making sure that the AI that's being built is not only safe and responsible and ethical and transparent, but also accessible to all, especially the populations that we serve,” Kyu Rhee, M.D., president and CEO of the NACHC, said in an interview.
As part of the collaboration, CHAI and the NACHC will conduct a survey of CHCs to better understand current AI adoption. Based on the survey, the groups will develop playbooks CHCs can use to implement AI and manage AI systems over time.
The groups plan to include front-line clinicians—like family medicine physicians and OB-GYNs—and health system administrators in the creation of AI playbooks for safety net clinics. Because CHCs often offer additional services like social work and nutrition support, Anderson said the perspectives of these practitioners will also be included in the playbook creation.
The organizations will co-develop an AI literacy curriculum for CHC leaders, providers and safety net organizations to be delivered in virtual, hybrid and in-person formats.
For vendors, CHAI will develop a certification program for those that want to work with CHCs. Additionally, CHAI wants to validate AI models that will be used in safety net settings through an expansion to its assurance resource provider network.
“We don't want products being created without us being at the table,” Rhee said. “And the beauty of when you're at the table and you're not on the menu is that you have an opportunity to articulate the needs of that community.” Rhee has served on the board of CHAI since February.
Health center controlled networks (HCCNs) and primary care associations (PCAs) will join CHAI’s assurance resource provider ecosystem. In line with its other assurance resource providers, CHAI aims to build an ecosystem of vendors that community clinics can trust to, for example, monitor and maintain their AI systems. The vendors will be vetted by CHAI and align with their principles of responsible AI.
CHAI will leverage the HCCNs and PCAs to serve as hubs for AI technology and oversight. Through the groups, CHAI hopes CHCs can come together under a common infrastructure to work with technology vendors to implement AI.
Many safety net clinics follow a similar model for their electronic health records, often coming together to purchase a single instance of an EHR on a cloud network, like OCHIN Epic.
“There are many others out there where they have a common infrastructure that could potentially bring the data assets together to service a different use case,” Anderson said. “In this case, it's not necessarily EHR-based. It's about model training, model validation, model monitoring and governance.”
CHAI has already engaged with several major PCAs and clinically integrated networks “to get them started using AI within the next six months,” Anderson said, though he admitted he’s not sure when the majority of CHCs will be using AI.
Vendors are also incentivized to work with CHCs so they can access new types of data. With that in mind, Anderson said he’s had conversations with CHAI member organizations to offer discounted pricing for safety net providers.
“I think it's fair to say that the majority of the health data that is currently already available is usually data, for the most part, that comes from very urban, highly educated, highly affluent areas, areas traditionally serviced by [academic medical centers],” Anderson said. “Now we're talking about making new kinds of data available for new models to be trained on to service new communities or new addressable markets.”
While academic medical centers are more likely to have troves of so-called “engineered data”—data that are ready to be used by AI—CHCs often lack the staff and resources to prepare their data for AI.
CHAI aims to “[connect] them to the AI engineering teams that will be able to build and label and engineer that data such that it's ready for training and validation" and "[find and build] a set of technology partners that can execute on that work,” Anderson said.
CHAI also announced that the NACHC will be leading its tiger team on the use of AI to meet new Medicaid requirements. The NACHC will help bring between five and 10 CHCs to the workgroup that want to develop frameworks for using AI for Medicaid enrollment.
Rhee, who is a member of the CHAI board, said he is excited to ensure developers understand the needs of CHCs when creating and deploying AI that will determine whether individuals meet the new requirements for Medicaid.
The NACHC has estimated that the Trump tax package will cause CHCs to lose 35,000 employees and close 1,800 sites. The NACHC estimated that CHCs will see 4 million more patients that don’t have insurance.
Because the clinics are required to provide services regardless of patients’ ability to pay, the clinics will have to cover the costs of those 400 million people—about $7.3 billion in lost revenue. With margins already in the negative—often between -2% and -3% a year—the tax package stands to weaken CHC infrastructure in the country.
CHCs often have dedicated staff to help individuals enroll in services like Medicaid or food stamps. Rhee said the addition of trustworthy technology will help keep CHC patients eligible for Medicaid through the documentation of work, volunteering or education commitments, as required by the tax package.
“There's a whole set of new workflows and data that need to be gathered as part of those application processes, working with the clinics, with the people literally at the front line understanding what are the challenges, what are the obstacles," Anderson said. "How does it actually, really play out?”
Editor's note: The NACHC predicted that CHCs will see 4 million more patients without insurance as a result of the tax package.