In a previous column on how artificial intelligence is on the verge of dominating large workflows in healthcare without a human being “in the loop,” I warned that the very nature of healthcare is now up for cultural grabs.
asserted
nature → dominate → grabs
The bean counters and efficiency mongers are demanding the use of AI in ways that threaten the profession itself.
asserted
that → demand → profession
I detailed several recent examples, but here’s another one: the health system giant Kaiser Permanente is using AI to evaluate everything from a nurse’s productivity to her empathy and tone of voice.
asserted
Permanente → detail → voice
The result, some nurses said, was a fundamental change in the nature of the care they offered.
asserted
they → say → care
One nurse nervously watched the clock as she stayed on the phone to care for someone experiencing suicidal ideation.
asserted
she → watch → ideation
Another cut short a conversation with a newly diagnosed terminal patient.
asserted
Another → cut → patient
Happily, nurses are pushing back in defense of their vocation to care for a fellow human being.
asserted
nurses → push → being
They circulated petitions, went on strike, and picketed using the tagline, “Trust nurses, not AI.”
asserted
They → circulate → nurses
The state of California, too, is resisting.
asserted
state → resist → California
Yesterday, for instance, Gov. Gavin Newsom (D-CA) signed AB 1979 into law.
asserted
Newsom → sign → law
This bill insists that healthcare professionals must maintain independent professional judgment and that autonomous AI may not perform licensed medical work apart from a human being.
uncertain
AI → insist → being
It also places health chatbots under California’s medical confidentiality law and requires that patient messages written by AI be identified as such.
asserted
messages → place → AI
Pushing in quite a different direction, however, is a federal administration easing oversight of AI in healthcare.
asserted
administration → push → healthcare
One might think that time saved by super-efficient AI would be given back to the human relationship between healthcare practitioners and patients in the form of more time for interaction.
uncertain
time → think → interaction
But that is not what is happening.
asserted
what → happen → ?
Most often, the bean counters put physicians at the service of making more money by seeing more patients.
asserted
counters → put → patients
Christina Farr recently pointed out that “about 75% of health-system IT and informatics leaders say the revenue cycle is their top area for AI.”
asserted
cycle → point → AI
The New York Times recently reported that AI is driving healthcare costs up, not down, at least in part because AI helps describe patients with “more complicated conditions and adding a secondary diagnosis.”
asserted
AI → report → diagnosis
In turn, hospitals “were paid an average of nearly $12,000 more per case.”
asserted
hospitals → pay → case
As Bishop Conley of Lincoln, Nebraska, mentioned at his homily at the opening Mass of the annual meeting of the Catholic Medical Association a few weeks ago, the use of AI is accelerating the growth of “structures of sin” in our healthcare system.
asserted
use → mention → system
In my previous column on this topic, however, I did promise to follow-up my criticism with a vision of healthcare that incorporates clinical uses of AI.
asserted
that → promise → AI
The most powerful case for doing so comes from a focus on justice for the most vulnerable, who have limited access to healthcare.
asserted
who → do → healthcare
The American Association of Medical Colleges predicts a shortage of 86,000 physicians by 2036.
asserted
Association → predict → 2036
If we want Americans to have access to primary care, mental healthcare, and geriatric care, there is a case to be made that — if current trends continue — it will be delivered by AI, or it will not be delivered at all.
asserted
it → want → AI
Thus, the controversial plan that Dr. Ezekiel Emanuel laid out recently in the Journal of the American Medical Association for incorporating autonomous AI into healthcare workflows (again, without a human in the loop) seems like it may be the only thing that makes it possible for our most vulnerable neighbors to get what they need.
uncertain
they → lay → what
But is such an outcome worth it if we must fundamentally change the nature of medicine itself by turning it into something that isn’t human?
asserted
that → change → something
In a powerful talk on AI in medicine given at Benedictine College recently, Dr. Kristin Collier of the University of Michigan School of Medicine offered story after story demonstrating the power of human connection and relationship in the healthcare encounter.
asserted
Collier → give → encounter
In one case, a patient of hers with heart failure gets hospitalized, almost like clockwork, several times a year.
asserted
patient → hospitalize → clockwork
Every workup comes back negative, but she, as this man’s primary care physician who got to know him so well over many years, knows why: his wife leaves to visit her sister and, with no one around to cook, he eats processed food and his sodium spikes.
asserted
sodium → come → food
Collier worries that clinical workflows run by autonomous AI would erode this essentially human part of the experience.
asserted
workflows → worry → experience
Indeed, it would accelerate the process of dehumanization that has been underway for some time.
asserted
that → accelerate → time
She notes that an intern comes into the job able to describe her vocation by focusing on a human relationship, saying something like, “I want to take care of people with diabetes.”
asserted
I → note → diabetes
When she becomes a fellow, however, the language changes: “I want to take care of diabetics.”
asserted
I → become → diabetics
Finally, as an attending physician, she describes her work by saying, “I manage diabetes.”
asserted
I → attend → diabetes
The AI revolution has provided many “it is a time for choosing” moments about who we want to become as a culture.
asserted
we → provide → culture
In healthcare, the following choice lies before us quite directly: Will we push back against a system already tempted to erase human relationships?
asserted
we → follow → relationships
Collier notes those called to medicine almost always want to care for people: someone’s mother, their sister, their best friend.
asserted
those → note → people
Those pushing for a shift toward autonomous AI often have a good goal in mind, such as expanding access to medical services for those who do not have them, but we simply must not undermine the very nature of healthcare itself to produce good outcomes.
asserted
we → push → outcomes
But what do we do with massive questions of lack of access for the vulnerable?
asserted
we → do → vulnerable
The solution must be to build new healthcare institutions, which train more physicians and other healthcare practitioners in an art that focuses on in-person, human encounters.
asserted
that → build → encounters
…and 17 more, not listed.
The Aporia
