After methodically assembling more than 1,300 members, the Coalition for Health AI formally launched on March 5 with the ambitious promise by its chief executive to be “the trusted source and curator of best practices of AI in health.”
To earn that trust in the “AI revolution,” added CHAI CEO Brian Anderson during a webinar, “we have to be a big tent” inclusive of all voices.
As I listened to the group’s grand plans, it struck me that CHAI has set itself up to be the “goo-goo” of health care AI. If that label doesn’t ring a bell, let me explain.
The term “goo-goo” was coined in the late 19th-century as shorthand for the “good government” groups that sprang up in New York, Chicago and elsewhere with the aim of countering the harm caused by corrupt city governments. The phrase was revived in late 20th-century Chicago by columnist Mike Royko to refer to inveterate do-gooders. While CHAI is a partner of government rather than an adversary, like the traditional “goo-goo’s” it is similarly seeking to institute good governance practices to protect the public from harm, in this case from the potential damage to health of an uncontrolled private-sector stampede for AI wealth.
The Civic Federation of Chicago, a pioneer “goo-goo,” sought “to awaken in the minds of all citizens…a sense of the obligations they carry” by enlisting them in a campaign to improve health through better sanitation standards. In a parallel manner, CHAI is seeking to enlist every sector of the health care citizenry to “improve lives” through standards, in this case for “credible, fair and transparent health AI systems.”
The Civic Federation knew it needed clout and credibility. Initial members included a famous “consumer advocate” (social worker and future Nobel Peace Prize winner Jane Addams) and powerful business leaders. CHAI’s initial partners include a respected consumer group, the National Health Council, and powerful private-sector stakeholders such as Google, Microsoft and the Mayo Clinic.
Moreover, just as the Civic Federation developed legislative and regulatory proposals to ensure its reforms forced change, CHAI plans to develop “guidelines and guardrails” in conjunction with the federal government. CHAI lists seven federal agencies as “observers,” with the director of the Digital Health Center of Excellence at the Food and Drug Administration and the National Coordinator for Health Information Technology serving on the nine-person board.
That level of involvement, wrote Casey Ross of STAT is “an experiment that will test whether industry and government can effectively partner in the regulation of a fast-moving technology.”
Yet for all the do-gooder vibe, I was left pondering two large, unanswered questions.
The first is whether CHAI’s stated commitment to “the responsible use of AI in health care” will give it the courage to take on a dubious health AI use that already affects tens of millions of Americans. As I’ve written in health policy publications and elsewhere, major health plans and medical centers are secretly collecting from data brokers and elsewhere individually identified information that includes credit reports, shopping habits and even website logins. This trove of data is then fed into AI algorithms in a process of “data fusion” meant to identify individuals who might need special assistance due to social determinants of health.
This process is much the same as that secretly used by the National Security Agency – like in health care, ostensibly for citizens’ own good – that was recently exposed by Sen. Ron Wyden (D-OR). In health care, the algorithms’ predictive accuracy has never been subjected to transparent, third-party scrutiny for either bias or effectiveness.
(When I mentioned this lack of public evidence to an executive at one organization using the data, he smiled and said, “We know it works.” Perhaps the NSA felt the same way.)
CHAI’s purpose statement reminds us that “lives are at stake in health care.” But also at stake here is big money: the government pays MA plans hundreds of billions of dollars annually, while many large health systems have “value-based” payment contracts. Will CHAI stand up for its ideals even if it irritates prominent supporters?
My second concern, a broader one, is whether underneath the soaring rhetoric about “creating an ecosystem where we fully realize the promise of AI to transform patient care,” CHAI really intends to do more than serve as a technical solution for industry. Right now, the group’s action plan is technician-tidy: setting forth principles, evaluation criteria, standards, evaluation, monitoring.
All of that is vitally necessary and certainly benefits the public. However, emerging AI tools are starting to provide patients with the kind of personalized, timely and reliable health information previously reserved for professionals. That kind of AI use, which threatens to profoundly disrupt the role of the doctor, is, as I’ve written, a lot messier.
Will CHAI be an ally of patient empowerment that disturbs the status of those who dominate health care today? CHAI’s current “focus” on the needs of patients, families and communities seems mostly to entail soliciting their input on how the health system should use AI tools and then using that input to persuade the public to trust them. Or, in Anderson’s words, to “empower the public to better understand how these tools are being developed and used.”
Indeed, it’s only when you dig deeply into the list of CHAI’s 13 different workstreams that, at number ten, you find a workstream involving “AI solutions intended to be used directly by patients without a health care provider involved.” Notably, although each board member spoke briefly at the webinar, none made mention of this kind of independence.
So are patients only objects – of secret, AI-powered analytics purportedly designed to help them and of the open attention of clinicians, researchers and AI experts who also intend to help them – or are they also subjects, capable if they wish of using AI to independently make their own decisions? Though CHAI’s leaders regularly raise the banner of revolution, they have yet to show a readiness to embrace radical health information democratization.
CHAI’s leader needs to choose. Do they see themselves like the business executives leading the “goo-goo’s” of old, mostly focused on fixing a list of specific, albeit important, problems? Or do they also possess the spirit of Jane Addams, a fiery activist who fought tirelessly for true transformation?
I know many of CHAI’s leaders personally and others by reputation; I have great respect for their abilities, and I don’t doubt their hearts are in the right place. How CHAI actually acts, however, remains to be seen.
