Mixed commission, chai partner to develop advice on health AI

Diving brief:

• The mixed commission is associated with the Coalition for Health IA in order to develop advice for the deployment of artificial intelligence tools in health care, organizations said this week.
• The group of safety and quality standards will work with Chai, a non -profit organization which aims to define directives on the adoption of AI, on play books for the use of technology in the sector. The first guidelines should be published in the fall, and the mixed committee will later publish a certification program to assess responsible deployment of AI.
• The advice is intended to help health systems to adopt, manage and monitor AI tools in health environments – which is not an easy task, said Dr. Brian Anderson, CEO of Chai, in Healthcare Dive. “It takes real resources of a people and from a technological point of view, and being able to do so in a financially sustainable way has been a real challenge for even some of the largest health systems in the United States,” he said.

Diving insight:

Founded in 2021, Chai has 3,000 members Organizations, including university medical centers, rural health systems, technological companies and startups.

The group has developed resources for AI in the sector, such as a development and deployment guide as well as model cards describing AI tools and a register where health companies can access it.

Chai’s work will inform the game books developed with the mixed commission – an important partner given its wide range with hospitals, Anderson said. The group has more than 23,000 health care organizations and programs in the United States, according to a press release.

The partnership occurs while AI has become an increasingly exciting technology for the health care industry. AI tools could help hospitals reduce the heavy burden on service providers, such as documenting patient visits, helping previous authorization requests or whistling clinical data, according to supporters.

But there are risks, including AI hallucinations, information or inaccurate biases anchored in models that could worsen health disparities. In addition, health systems will have to continue to monitor their AI tools over time, because the assumptions underlying the models could change and worsen their performance.

It is an important company for service providers, said Anderson.

“These were systems that deploy pilot programs and AI governance processes for one or two or three models, and they quickly noticed that this type of approaches costs between $ 1 [million] And $ 2 million, “said Anderson.

The directives developed by Chai and the mixed commission will likely help to attack some of this work, such as how to set up governance processes and committees, the technical infrastructure necessary to monitor the performance and value of a solid partnership with model suppliers, he said.

Playbooks should also apply to different types of suppliers. For example, a critical access hospital will have fewer resources at its disposal, so the advice might recommend that it have a smaller governance team while working with a reference partner, Anderson said.

The first tips will be available in the fall, with the certification program to come later. The program will be the “next chapter” of responsible use of the certification of health data by the mixed committee in 2023, Dr. Jonathan Perlin, president and chief executive officer of the mixed committee, said by email.