Photo: Greg Nash-Pool/Getty Images
LOS ANGELES – MobiHealthNews attended Future of Health’s invite-only, trailblazing event here on Tuesday, where former U.S. Centers for Disease Control and Prevention (CDC) Director Dr. Rochelle Walensky sat down with MHN for an in-person interview to discuss the promise and pitfalls of AI in healthcare.
Dr. Walensky cautioned against overburdening clinicians with unrealistic efficiency expectations and emphasized the need to rethink medical training as AI continues to reshape clinical roles.
MobiHealthNews: Do you see AI adoption in healthcare as a plus, or are you wary about its use in any capacity?
Dr. Rochelle Walensky: I mean, I can see so many different ways that it could help both patients and physicians and outcomes. We heard today about places where AI is being used. Where there's an absence of physicians, there is an absence of care, and it is filling the gaps of where there is nothing. How can AI help triage? I can see how it can help in documentation. I can see how it helps in streamlining and differential diagnoses. There are so many different ways.
I am also cautious about the expectations on the human. If the final common denominator is that all of this goes through a human and yet the human has to be the last touch of, I've synthesized a gazillion pieces of data points and now you just have to sign off on this, that seems untenable for any given human.
I'm sensitive to the fact that we are not yet training in our schools to understand what the training is that we need if AI is going to do so many of these things in the future. How should we be training the next generation of physicians if, five years from now, AI is going to be doing chart reviews and summaries and all the documentation? What are the core true skill sets that a human needs to have? And then is there some expectation that because AI is there, humans have to be even more efficient?
I've heard a lot of discussion about, like, well, now we have AI, it will deal with your charting and your documentation, so you now need to see 12 patients instead of 10 in your panel this morning. I don't know that that ultimately gets to better things for physicians and for patients.
MHN: Is there still concern around bias in AI?
Walensky: I'm very concerned about the bias in AI. I don't actually think that people are talking about the bias in AI.
MHN: It is like that conversation just stopped.
Walensky: It totally stopped.
So, you know, one could imagine a situation where you say this underserved population actually does not need as much healthcare because, in fact, they are not in our system, right? Because they, in fact, are underserved. So, I think that there are a lot of biases in AI.
We do need the human piece of this, right?
How is it that when the human says, "Actually, this is more CHF than pneumonia or this is more pneumonia than CHF," what is it that the human is doing to sort of pull that together and make that? And in fact, is the human more often right or wrong?
MHN: How can health tech companies ensure that they always have the most important person at the forefront of their technology, which is the patient?
Walensky: I think there's plenty of money to be made here, and I know there are, sort of, comments about, "You need the margin for the mission," but I think we do have to have a patient-centered approach here, and that may mean that we give up a little bit of margin.
Ultimately, healthcare is 18% of our GDP, our outcomes are poor in this country compared to other high-income countries, and so if this is about making money, there are plenty of places where money can be made. If this is about improving healthcare, you might make a little less and have a huge impact.
MHN: Or if you are really good and patients all improve, maybe you will make more money.
Walensky: Right, exactly.
MHN: Were there any health technologies that emerged during the pandemic that were unexpectedly found to be vital?
Walensky: Telehealth, I think, was huge and, unfortunately, is now sort of less impactful because of policy.
Certainly, wastewater was super helpful for us, and now we're expanding the use of wastewater.
There were the Kinsa thermometers. So, these thermometers that crowdsource essentially what your temperature is in population. We can say, "Oh, this population actually has an average temperature that is a little bit higher. We may have an outbreak here."
I went to this incredible health clinic in Vietnam on a Sunday, and in a second-grade school, they were reading chest X-rays for tuberculosis screening for elderly people, using AI to find cooling towers for Legionella outbreaks.
Or using AI to understand where you might have an outbreak based on OpenTable reservations.
MHN: How is that possible?
Walensky: Well, because people do not go out if they are not feeling well.
So, there are a lot of different ways where we can combine things like OpenTable and Kinsa thermometers, and I think we are just at the tip of the iceberg here in terms of what we could do.
MHN: And there is a lot that we do not know yet.
Walensky: Right. And how can our data systems, like our airport systems and our OpenTable systems, and thinking about using data that we do not even consider to be health data, how can we leverage that to help in health?
