Month: May 2022

CHAKRABARTI: Dr. Obermeyer, what do you think about that? Is that realistic or too Pollyannaish?

OBERMEYER: I think it’s a very astute comment, and I think it highlights the importance of doing that rigorous evaluation that we apply to any other new technology and health.

When a pharmaceutical company produces a new drug and wants to market it, we don’t just say, Sure, go ahead. We say, Well, why don’t you test it compared to some acceptable standard that we currently use. And that’s why we have big randomized trials that pharmaceutical companies do before that drug ever makes it to the market.

And I think similarly, when AI is being deployed in very high stake settings, we need to compare it to what we’re currently doing. And I think that can expose some of those data blind spots that you mentioned, which I think is a real concern.

But it can in general just tell us, are these technologies doing more good than harm? And should we be investing in them, or should we be applying a much more cautious approach, and not? It all needs to be judged on the basis of the costs and the benefits that these algorithms produce in the real world.

CHAKRABARTI: Well, you know, obviously, the far horizon of what AI could do in health care captures the mind. Helping better understand if a heart attack is actually happening. Some of the things we heard about a little earlier in the hour about pattern recognition in cancer and things like that. Very, very alluring possibilities.

But reality check, right? Dr. Obermeyer? Because those technologies are actually quite far away. What’s more probable in the near future is AI’S impact in, you know, what seems like a potentially mundane aspect of health care. Mundane, but critically important. Things like tracking when health care workers sanitize their hands before interacting with patients.

DR. ARNOLD MILSTEIN [Tape]: That tends to be about 20 to 30{cfdf3f5372635aeb15fd3e2aecc7cb5d7150695e02bd72e0a44f1581164ad809}, which is on the face of it, indefensible and crazy.

CHAKRABARTI: So that is Dr. Arnold Milstein, who was talking about the failure rate of health care professionals to actually sanitize their hands. It is about 20{cfdf3f5372635aeb15fd3e2aecc7cb5d7150695e02bd72e0a44f1581164ad809} or 30{cfdf3f5372635aeb15fd3e2aecc7cb5d7150695e02bd72e0a44f1581164ad809}. And so Dr. Milstein and his colleagues at Stanford University are developing an AI enabled system that reminds medical workers to sanitize their hands.

So algorithms are also proving to be unrivaled medical assistance, as well. Here’s another area. Natural language processing, which can crawl through patient records. Radiologist Dr. Ryan Lee at the Einstein Health Network told us that logistical AI systems can automatically send notifications to patients for follow up care.

DR. RYAN LEE [Tape]: This is a real opportunity to close the loop, so to speak, in which we’re able to directly notify and know when a patient has actually done the appropriate follow up.

CHAKRABARTI: There’s also another example. Dr. Erich Huang, chief science officer at the company Onduo, says health care has a huge paperwork problem. By some estimates, time doctors spend on clinical documentation can cause anywhere from $90 to $140 billion in lost physician productivity every year.

DR. ERICH HUANG [Tape]: Algorithms can lift some of the sort of grunt work, documentary grunt work of clinical medicine off of the physician’s shoulders. So that he or she can actually spend more time taking care of the patients.

CHAKRABARTI: Dr. Obermeyer in Berkeley, California, tell me a little bit more about these, again, mundane but actually critically important aspects of health care that AI could have a really profound impact on.

OBERMEYER: I love these examples. Because when you look at where AI has had impacts in other fields besides medicine, it’s often these very similar things that are like back office functions or, you know, routing trucks a little bit more efficiently. But those kinds of things stack on top of each other, and make the whole system much more efficient.

So I love these examples because, you know, the health care system does a lot of things besides curing cancer. And I think AI can really help with those simple tasks. I think one of the challenges is trying to make sure that the things we think of as simple tasks are indeed simple tasks. If you think about the task that a physician is doing when she’s documenting, when she’s writing a note.

Part of that is mundane grunt work. Because you have to check a lot of boxes. But part of it is you have to put a lot of thought into summarizing, Okay, what is going on with this patient? What do I think? And those are things that algorithms are going to have a much harder time doing. Because those are things that rely very heavily on human intelligence in ways that we haven’t yet figured out how to automate.

CHAKRABARTI: Okay. So that’s a really, really interesting point. And it links back to this broad range of estimates in the impact that AI could have, even in something as seemingly simple as clinical documentation, right? That $90 to $140 billion annually in lost physician productivity.

Presuming that the truth falls somewhere in that range, I mean, how much of an impact could AI have in the delivery of health care overall, say, if physicians were freed up a little bit from the burdens of clinical documentation?

OBERMEYER: I think it’s a fantastic area of study because I do think that physicians are not only wasting time on doing a lot of mundane tasks, but it’s also almost certainly one of the big causes of burnout. You sign up to be a doctor, but then you get to your job.

And most of your job is doing paperwork, and making phone calls and being on hold with an insurance company trying to make sure that your patient is getting what they want.

And so I think that these kinds of technologies, by freeing up doctors to do the work that we’re trained to do, have huge potential. Just in the same way that the historical example of the ATM machine was very transformative, it freed up the bank teller to engage in much more sophisticated work with clients, rather than just dispensing cash.

CHAKRABARTI: It seems to me that one of the takeaways here is that however we want to judge the potential of AI in health care, that potential is proportional to the problem that any particular algorithm is asked to solve, or analyze. And the risks that come with applying an AI or machine learning tool to that problem. What do you think about that?

OBERMEYER: Absolutely. And I think, you know, clearly, the benefit is going to be proportional to the size of the problem. I do think that the examples you just mentioned also have this nice illustrative feel, that we also need to make sure we’re targeting the problems that machine learning can solve, the data problems.

Many problems in medicine are problems for which we don’t yet have data. And we need to be very careful to only aim AI at those questions where we have data that can help answer them.

CHAKRABARTI: Well, when we come back, we’re going to talk in detail about the tradeoffs. With all that potential that could come with artificial intelligence in American health care, what are the tradeoffs and what are the particular areas of concern?

CHAKRABARTI: Welcome back to the first episode of On Point’s special series ‘Smarter health.’ And today, in episode one, we are taking a look at the potential for artificial intelligence and machine learning to change, even transform medicine. Here’s Dr. Kedar Mate, CEO of the nonprofit Institute for Health Care Improvement.

DR. KEDAR MATE [Tape]: There is tremendous, tremendous potential in AI, machine learning that goes along with that AI, to augment and improve our capacity as clinicians and as humans, frankly, to be able to do the mountain of diagnostic work that we have to do to manage the information flow that’s coming at us at all times as clinicians.

And to be able to provide just in time absolutely critical, precise, personalized care to the people that we’re taking care of. But there’s also, like any technology, considerable risk. Unless we mitigate those risks with deliberate design, we won’t necessarily solve for those problems.

CHAKRABARTI: I’m joined today by Dr. Ziad Obermeyer. He is a distinguished associate professor of health policy and management at the University of California, Berkeley School of Public Health, also an ER physician as well. And Dr. Obermeyer, one of the areas of concern — and there are several which we will be exploring over the course of this four-part series here.

But one of them is, you know, how much do people actually understand right now between accurately regarding the state of AI in health care? Do you think patient perception matches the current reality?

OBERMEYER: I think one of the things that’s probably underappreciated is how widespread these algorithms already are. In some work that we published a couple of years ago, we studied a set of algorithms that are used for what’s called population health management.

So this is the function of health systems where they try to get an overview of all of their patients and figure out which ones need help today so that we can prevent deteriorations in their health tomorrow.

So we studied one commercial product that was being used to make decisions for about 70 million people, every year. If you look at the industry estimates, those algorithms are being used for between 150 and 200 million people per year in the U.S. So essentially most of the population.

CHAKRABARTI: Already?

OBERMEYER: Already. And so the scale of these things already has gotten huge, and I don’t think that’s very well appreciated. Unfortunately, that study that we did also showed that these algorithms suffered from a large degree of racial bias. So I think that’s another thing that’s not very well appreciated. Is that there are both reasons to be incredibly optimistic about AI, as all of the examples you already mentioned convey. But there are also reasons to be very, very careful.

CHAKRABARTI: Can you just describe briefly what kind of decisions the algorithms that you just talked about were making or assisting with?

OBERMEYER: So what health systems have to decide is, well, you’ve got a bunch of patients in your population that you’re responsible for. Some of them are going to get sick tomorrow from things that we could have prevented, had we known about it today. So what algorithms are being used for, which is a very good use of algorithms, is looking into the future and trying to predict, OK which patients are going to get sick?

Which patients are going to have an exacerbation of some chronic condition that I can help them with today? And so the patients that are identified as high priority get a bunch of extra help from the health care system, extra primary care visits, extra visits from a nurse practitioner, a special phone number that they can call for help any time. So it’s very, very helpful. But we can’t do it for everybody. We have to prioritize. And that’s where the algorithms come in.

CHAKRABARTI: And those algorithms already, as you said, are being used on hundreds of millions of people.

OBERMEYER: Yes.

CHAKRABARTI: Amazing. Okay. So I have to tell you that the next episode of our series, really goes in true depth to these ethical considerations. The concern about bias in the data that’s being used to train algorithms in health care. That’s the whole hour next week. So we will examine that closely.

But I wanted to just stick for a moment with, again, patient perception of what’s really going on in health care right now. So we spoke with Dr. Richard Sharp. He’s the director of the bioethics program at the Mayo Clinic. And he and his research team conducted 15 focus groups to try to understand current patient perceptions of AI in health care.

DR. RICHARD SHARP [Tape]: When most people hear about artificial intelligence, things that come to mind for them, are, you know, science fiction movies where computers somehow take on an aspect of our lives. The machines become sentient and rebel against humanity and those sorts of scenarios. In health care, though, those sorts of tools are a lot more mundane.

CHAKRABARTI: So Dr. Sharp says right now he sees a perception gap. The research team found, though, that they could narrow that gap by giving patients real world scenarios, using very neutral language about specific applications of AI in health care. And that did indeed help, but it didn’t completely allay patient concerns.

SHARP: The folks that we talked to mentioned self-driving cars multiple times. And what they told us again and again was that they were uncomfortable with a self-driving car, but they definitely did not want a self-driving clinician. They did not want a self-driving doctor. They wanted to be sure that they had the ability to talk to the real deal and make sure that there were appropriate safety checks in place.

CHAKRABARTI: So what patients really wanted? Transparency. Everything from how algorithms were being deployed, to who had access to the information used by the algorithm, to maintaining the ability to make decisions with their doctors, even if that decision defied an algorithms recommendation.

SHARP: They were worried that an AI algorithm might recommend a particular treatment or drug that would be more expensive than maybe a drug that they’re currently on. That’s really the promise of AI, is to be able to identify early on in the course of the disease, those treatments that are likely to be most effective.

With that capacity, though, it can create a situation where maybe that ideal treatment is simply too expensive for an individual patient, or not covered by a particular insurer. And patients were quick to point out that they saw that as one of the major downsides of these tools.

CHAKRABARTI: So Dr. Sharp says that successful treatment really hinges on patient compliance. But the patients in his focus groups were clearly saying that compliance hinges on having confidence in the new technologies used to treat them. So that leads Dr. Sharp to a clear conclusion. Patient education about AI, and addressing the concerns they have must be rolled out in parallel with the tools themselves.

SHARP: I think it would be a mistake for the future of health care if patients discovered after the fact that the care they were receiving had been influenced by AI algorithms.

CHAKRABARTI: That was Dr. Richard Sharp, director of the bioethics program at the Mayo Clinic. Dr. Obermeyer, what do you think about that? Do you think that what Dr. Sharp said there is actually happening?Concurrent patient education, along with the development of the tools used to treat them?

OBERMEYER: I love the idea that Dr. Sharp proposed a concrete example. So let me try one from a completely different field, which is that I’ve been traveling a lot now that lockdowns are over.

And I was reflecting on the fact that when I get on an airplane, I actually have no idea how the autopilot was trained, evaluated, deployed. And I think that, you know, if I think about everything that happens inside the hospital today, there are algorithms that have been operating for decades that help MRI machines process the image, that help laboratory analyzers process the single cell measurements that they do.

So algorithms are actually being used all around us, and either we have no idea, or we don’t care. But I think that that’s because we have confidence in a set of practices, and procedures and regulations that guide the deployment of all of those algorithms in high stakes settings.

And so I think that a useful complement to the things that Dr. Sharp was proposing is developing that regulatory structure from the government, but also developing the procedures and practices that the health care system uses before it ever deploys an algorithm to test it and make sure that it’s safe.

CHAKRABARTI: Yeah, so the regulatory structure is going to be episode three of our series here. Now, in the last few minutes that I have with you, Dr. Obermeyer, look, we have to acknowledge that one of the screamingly unique things about anything regarding American health care is the fact that we are living in the country that spends the most money on health care than any other nation in the world. I started off the hour by highlighting that.

And the numbers are actually just like jaw dropping, right? That the Centers for Medicaid and Medicare Services says in the next couple of years, next five years, the U.S. is going to be spending $6 trillion on health care. So it’s still going to be 20{cfdf3f5372635aeb15fd3e2aecc7cb5d7150695e02bd72e0a44f1581164ad809} of our economy. And that’s, I think, one of the things where, you know, the technology evangelists are really excited about the possibility of AI because they say it could bring down costs.

You know, bringing in those algorithmically driven efficiencies into health care could bring down costs. But here’s what Dr. Kedar Mate, again, CEO of the nonprofit Institute for Health Care Improvement, says about whether we know anything at all about … AI [reducing] the cost of health care in America.

DR. KEDAR MATE [Tape]: Virtual care, just as an example, virtual care has likely done little to reduce total cost of care. In fact, during the pandemic, you’ll probably recall that we collectively argued for pay parity between virtual care and in-person care. And you can just imagine if we’re arguing for pay parity, then even if we have all of our care being virtual, it’s going to cost exactly the same.

This doesn’t necessarily lower the cost of care. I think a lot of AI enthusiasts, tech enthusiasts, more broadly believe that all of this will reduce the cost of care. But we haven’t seen substitution for in-person care. We haven’t seen reduced frequency. In fact, in some ways, technology enables increasing frequency of interaction with people, and it hasn’t lowered the cost basis necessarily of providing that care. So for all those reasons, I’m not sure yet. I don’t think anyone is sure yet whether or not AI and attending technologies will lower the cost basis of care.

CHAKRABARTI: That’s Dr. Kedar Mate at the Institute for Health Care Improvement. So, Dr. Obermeyer, I mean, even just increasing touch points in health care. Well, you know, it might feel good because you have more information, more access to the health care system. Every touch point is a billable moment. And in overall, the United States in a for profit health care system. Is there any possibility that the end result of AI in health care would be anything other than costs continuing to rise?

OBERMEYER: I think I’m more optimistic about this particular question. Because I think we’re we’re just incredibly early in the curve of AI being applied to health. And so I don’t think we can generalize from anything that we’re seeing today.

Ultimately, you know, if you look at our paper on testing for heart attack, the potential of AI there is to take all of these tests that we do on people who come back negative, who didn’t need the test after all, and eliminate those. And take a portion of those tests and reassign them to people who are genuinely high risk, who should have been tested, but that currently aren’t.

And I think that’s a good general principle for AI, is we do a lot of things that don’t make sense today and that becomes very wasteful. So we can reallocate some of that waste to the people who are losing out today. And everyone does better. We spend less money on testing, and we get tests of people who need them more.

And I think that that’s going to be the playbook for AI in medicine over the next few decades. So I’m very optimistic that we’re going to be reducing costs for all of the things that we are doing today that we shouldn’t be doing.

CHAKRABARTI: But haven’t we heard something similar for other technologies that have been introduced into health care? You know, electronic health records are supposed to make information sharing more efficient. Any other sort of big system that was talked about as a revolution in health care. And yet the costs still keep rising. We still keep spending more and more.

OBERMEYER: I think that’s right. But I think that’s because electronic health records haven’t fundamentally changed anything that anyone is doing in health. In many ways, it’s a lot like how the power plants that were electrified, but that were still fundamentally organized, like steam powered power plants, actually had no productivity gains from electricity.

And it was only the new factories that were reorganized around electric power. So I think medicine’s very similar. Once we have all of this electronic data, it doesn’t actually do as much good if we’re stuck in an old system. But now that we have the tools to build up a new system, I think things are going to get a lot better.

CHAKRABARTI: Well, Dr. Obermeyer, we have 30 seconds left here and just send our listeners off today with a thought or or a tool that you would add to their toolkit to understanding how AI might have an impact on their health care. What do you want them to know?

OBERMEYER: I would like them to know that AI is not the solution for all problems in medicine, because so much of this in human enterprise, where human doctors are doing really, really good things for patients. But there are some parts of medicine that are incredibly complicated from a data and statistical point of view. And I think for those parts of medicine, AI is going to be transformative.

CHAKRABARTI: Well, Dr. Ziad Obermeyer is an emergency medicine physician and Blue Cross of California, distinguished associate professor of health policy and management at the University of California, Berkeley School of Public Health.

He also helped launch Nightingale Open Science, which is taking a look at how to provide high quality data to AI systems. And again, we’re going to talk about data later on in the series. But Dr. Obermeyer, it’s been a great pleasure to have you on the show. Thank you so very much.

OBERMEYER: Thank you. It was such a pleasure.

DR. STEVEN LIN: As exciting as AI and machine learning are, there are many ethical and also health equity implications of artificial intelligence that we are now beginning to realize.

CHAKRABARTI: That’s Dr. Steven Lin, primary care physician and head of the Stanford Health Care Applied Research Team. So next week, we’re going to talk about AI, health care and ethics. And we’re going to do it through the story of what Lin calls the advance care planning model. But you and I might better understand it as the death predictor.

LIN: AI can actually pretty accurately predict when people are actually going to die. It raises the question of how accurate are these predictions? How do patients react when they are flagged by the model as being high risk of X, Y and Z, or being diagnosed with X, Y and Z?

How do human clinicians handle that? And then very, very importantly, what are the equity implications of data driven tools like artificial intelligence when we know that the data that we have is biased and discriminatory. Because our health care systems are biased and discriminatory.

CHAKRABARTI: That’s next Friday in episode two of our special series ‘Smarter health.’

We want to hear from you

Got a question about how AI will impact how you receive health care? Or maybe you’re a scientist, doctor or patient with an AI story to share? Leave us a voicemail at 617-353-0683.  

This series is supported in part by Vertex, The Science of Possibility.