Sometimes it’s hard to keep up with everything that is going on in healthcare IT. Regardless of how many unread newsletters and blog notifications are in my inbox, I know I can count on HIStalk to deliver the curated content that helps me identify the topics that I need to dig into and those that I can let slide for a while.
This week, I appreciated Mr. H posting a link to an article about the strategy that the Department of Health and Human Services (HHS) plans to use as it expands the role of AI in healthcare. The document is 20 pages long and reads like an ode to the wonders of AI, with less attention to the documented risks and benefits.
As someone who has spent a good chunk of her career doing process improvement work, where evidence and outcomes are key, and who is squarely under the influence of evidence-based practice where patients are concerned, I’m all about the details. It isn’t enough to just say that you have a cool technology that’s going to be revolutionary. We had enough of that with Theranos and the pharma bros. Now we are in an era where people want to see results and understand fully how care might be impacted and how patients will be protected.
There are five key pillars in the document: creating a governance structure that manages risk, designing a suite of AI resources for use across the department, empowering employees to use AI tools, funding programs to set standards for the use of AI in research and development, and incorporating AI in public health and patient care.
It sounds like the “empowering employees to use AI tools” piece is well underway since HHS has made ChatGPT available to all its employees. Based on my own experiences, I initially hoped that they were not using it to look up health-related content, because I’ve seen some wild inaccuracies over the last year even with non-controversial queries, such as asking it to summarize a movie plot.
Unfortunately, plenty of media reports say that HHS leaders are planning to use it to “deliver personalized, context-aware health guidance to patients by securely accessing and interpreting their medical records in real time.” Unsurprisingly, that idea raises concerns about having third-party vendors accessing patient medical records and how that data might be protected.
HHS has already given the protected health information, including birth dates and Social Security numbers, of Medicaid enrollees to the Immigration and Customs Enforcement department, which is cringeworthy for those of us who have had to sit through decades of HIPAA training courses. Although it appears that HHS will prioritize risk mitigation, the clinical experts who I have spoken with have serious concerns about the organization’s ability to prioritize patient protection over political requests.
Those of us who are following the evolution of vaccine policy in the US have seen a disregard for the scientific method and the removal of world-renowned experts from the process. We have no reason to think that things will be different with AI. Given that we have decades of experience with vaccine efficacy and little experience with the impacts of AI, clinicians are understandably concerned.
A comment on the document noted that although safety measures are in place for individual patient information, no similar safeguards are listed for aggregated information that is being used by AI tools.
As I began to dig into it, I was surprised at how it differed from previous HHS publications over the last few decades. A glossy cover page was followed by a full-page photo of the secretary of Health and Human Services with a superimposed quote saying, “We are making HHS the template for the Utilization of AI.” When I’ve seen splashy graphics pages like that in the past, it’s been in the context of a major discovery or a noteworthy quote, but this just felt weird, for lack of a better word.
The document continues with introductory letters from the deputy secretary and the HHS chief AI officer. In the first letter, HHS Deputy Secretary Jim O’Neill notes that “By guiding innovation toward patient-focused outcomes, this administration has the potential to deliver historic wins for the public – wins that lead to longer healthier lives.”
What does he think that all of us healthcare and health tech people have been doing for the last two decades? We’ve been patient-focused and outcomes-driven for a long time. Maybe he thinks it’s something new or unique to this leadership.
My favorite statement is in the second letter. HHS Chief Information Officer and Acting Chief AI Officer Clark Minor, states, “This paradigm shift will unleash a new era of well-being for a healthier America.” I was reading this in a room with a dozen family physicians, so I asked them, “What one thing do you think will unleash a new era of well-being for a healthier America?” None of the answers included AI.
What they did include were concepts such as universal healthcare, eliminating healthcare inequity, increasing social services that directly impact health, mitigating the impact of food deserts, investing in preschool and early childhood education, strengthening nutrition education in the public schools, and increasing the primary care workforce through additional residency training spots and low-interest loans for those who pursue careers in primary care.
The ensuing discussion made me wonder how much the folks at HHS are actually talking to those who are on the front lines of public health and primary care. What do they need to help promote health and prevent disease? What are their pain points? Which solutions have they tried, and can they share an inventory of what worked and what failed?
I’m certainly not part of the policy-making apparatus in the US, but I know how I solve workflow hospitals in a hospital. It doesn’t involve putting all of my eggs in the AI basket. We use a rigorous methodology to analyze dysfunction and to propose solutions, and it actually works.
This idea of assuming that AI will solve all our problems and then taking action based on that hypothesis makes me feel like we’re all part of a giant unregulated experiment that wouldn’t pass the basic rigors of a middle school science fair, let alone the Institutional Review Board of a research institution.
I have to admit that I haven’t finished reading the document yet, largely because the level of rhetoric present was giving me a headache. I also have a time-consuming personal project that I’m trying to complete, so I decided to switch gears. I’m eager to hear from anyone who has read the whole thing.
What are your thoughts on how expanded AI at HHS will impact the greater US healthcare ecosystem? Do you think AI is going to be a major driver of change, or is it just another distraction from the difficult and often messy work that needs to be done to improve the health of a large and diverse population? Leave a comment or email me.
Email Dr. Jayne.
Re: Dr. Craig Joseph's piece on "physical robots, not chatbots, may hold the key to effective emotional and behavioral support…