AI May Drive Health Costs Up, Doc-Economist Says

In this series, MedPage Today is asking healthcare economists and policy experts the same questions about the high costs of U.S. healthcare. They’ll reveal what they believe is working, what’s not working, and what else can be done to bring costs down.

In this interview, Ben Sommers, MD, PhD, a primary care physician and professor of healthcare economics at the Harvard T.H. Chan School of Public Health in Boston, talks about the artificial intelligence (AI) “arms race” in medicine and how it may end up driving healthcare costs up.

An edited transcript of the conversation follows.

What has been the greatest contributor to high healthcare costs in the U.S.?

Ben Sommers: Why do we have the highest costs? Why is our total national expenditure on healthcare so big compared to other countries? Primarily, it’s because our healthcare prices are very high. That’s been known for decades. Some of the original papers on this topic were written by an economist, Uwe Reinhardt, PhD, who used to say, “It’s the prices, stupid.”

He showed that on a lot of measures, the U.S. uses similar amounts of healthcare as other countries. Some countries use a little bit more of prescription drugs, a little bit less of advanced imaging or vice versa. But in general, the U.S. wasn’t really that different in terms of the overall patterns of use of how much care we were getting. It’s that we were spending a lot more for everything we got because the prices were high.

And when we say healthcare prices, it’s important to clarify what that means. It’s the price that we pay doctors. It’s the prices we pay hospitals. It’s how much we pay for our CT scans and MRIs. It’s how much we pay for our prescription drugs. It’s all of the different things that go into healthcare. Each has a price and the U.S. is paying higher prices than most other places and that really has led to us having much higher costs.

There’s a separate question of why are our costs rising year to year. That’s not only about prices. There is some price inflation that we see in the U.S., but we do also see, year-over-year, some changes in the kinds of services that are getting used. We see new technologies and new prescription drugs that are coming out that are quite costly. We see some changes in the ways that billing is being done by providers and health insurers related to how severe the codes are.

But at the end of the day … it’s that our healthcare prices are so high.

Has physician pay contributed to high costs?

Sommers: When you’re talking to a clinician, and I myself am one, we have to remind ourselves that when I say healthcare prices in the U.S. are high, if I translate that into my selfish interest as a doctor, what I’m saying is my pay is high, my income is high, compared to what doctors in other countries are getting for similar services. … Yes, U.S. clinicians generally get paid more than their peers in other countries, and that’s a contributor along with the higher prices for prescription drugs, for hospital care, for scans, and the like.

Now, there’s another piece of this related to clinical pay and physician pay, which is U.S. healthcare practices have to spend more money on administration than they do in other countries because the billing complexity in the U.S. is so much more.

We have this patchwork system of all these different insurers. We have Medicare, we have Medicaid, we have all the private insurers that are competing, we have the marketplaces. And as anyone who practices clinically knows, you need a whole team to help you navigate that and make sure you get reimbursed and to make sure the bills are being paid and to deal with the denials and all that sort of thing. And those jobs are real resources. We spend money to hire folks. Hospitals spend money, practices hire support staff and that drives up costs as well because in other countries they just don’t have that level of complexity and they don’t have to spend that much money on administration.

I was going to ask about administrative costs, but you’ve answered that. Instead, I’ll ask, what is the best solution you’ve heard for lowering the cost of healthcare?

Sommers: That is a very tough question because it’s easier to say the solutions I’ve heard that I don’t think will bring down the cost of healthcare. We hear a lot about if only we did more cost sharing. When people have more cost sharing, typically they will use fewer resources, they will spend less. But we’ve already shifted a lot of our system towards higher deductibles and higher cost sharing. There’s only so much room you can continue to push on that before we have real adverse effects on people’s ability to get care and their health outcomes. That as a proposal I think is probably not a panacea.

There was a lot of optimism around healthcare delivery reform, things like accountable care organizations, primary care, medical home. Unfortunately, the savings from these programs in many cases is small or even zero — maybe 1% or 2% savings from some of these programs that have been more successful than others. That’s not nothing in the scheme of the U.S. healthcare system. That’s still hundreds of millions and billions of dollars when you look at the system as a whole. But it doesn’t get us out of the challenge that we spend so much more on healthcare than other places.

So at the end of the day, what solution is most effective here? I think it’s largely tackling the underlying problem of prices.

The U.S. generally has not had a lot of political interest in regulating healthcare prices. Regulating healthcare prices means setting prices lower than they currently exist. Clinical groups, hospitals, doctors groups are always up in arms anytime there’s a suggestion that that’s gonna happen. We hear this essentially almost annually in Medicare, like some slated cuts will be catastrophic and we can’t have that happen. But if you want our healthcare costs to look more like other countries, we have to start paying our providers, our clinicians, our device manufacturers, drugmakers, more like what other countries are paying.

And we are seeing a little bit of a move in that direction. The Inflation Reduction Act introduced Medicare drug price negotiation for the first time. That was not allowed before this law. Medicare Part D could not negotiate prices. Now Medicare is using its purchasing power to bring down prices for a handful of drugs each year that are high-cost drugs that take up a fair share of the prescription drug market in Medicare. And this is making an impact.

Would having a single payer help control costs?

Sommers: I think most reasonable assessments of single payer that it would save money overall in terms of the total amount spent on healthcare, and it does that through a few pathways. The first is it would eliminate a lot of the administrative costs we were talking about earlier. If you have one insurance plan for everybody, the government is paying all those bills, that’s a lot simpler. You wouldn’t really need a billing apparatus in your office. It would become very simple, and that would bring down costs from the administrative side.

The other reason it probably would reduce costs is because in general, when the government is responsible for a greater share of spending, the budget strings tighten a bit. The government says, we can’t afford to pay some of the rates that the private insurers used to pay. There’s just no way we can do that. Now that we’re covering everybody, we’re going to have to negotiate drug prices even more aggressively. And so those changes likely would bring down prices.

They would reduce profitability and income for clinicians and for those who work in the healthcare sector, which is why we’ve seen a lot of vested opposition to any kind of single-payer model in the past from physicians groups, from hospitals, from drug and device manufacturers. But would it reduce costs? I think it absolutely would.

Would greater transparency in pricing help bring costs down?

Sommers: I think based on the evidence we have so far, price transparency is another one that I would put in the category of, we wished it was going to save us a lot of money, but it doesn’t look like it does.

The idea here is if we made it easier for folks to see the cost of services at different providers, they could go to the more efficient lower-cost providers. There’s not a lot of evidence that people do that sort of shopping, or at least they don’t do it consistently. Most patients are not navigating the system with those sorts of considerations in mind, often because it’s complicated. And you’re catching people in a time of illness and they often, this is the last thing that they have the capacity to do right now is to try to figure out all the ins and outs of the healthcare system. Where do I get that data? Do I understand what conditions I’m comparing? It’s not like getting two estimates on your car or for a house repair, which I think folks are a little more familiar with and it happens in a different phase of your life.

There’s actually evidence that price transparency can be counterproductive. For instance, there’s some evidence that if organizations can see the prices that they’re being paid by different insurers, they might see, oh, those providers actually are getting paid a lot more than us. We should be negotiating a harder bargain with the insurers. And in doing that it could actually increase prices. So it’s not always the case that transparency is going to lead folks to gravitate towards cheaper options and lead to a good price competition. It may be the opposite. In some cases, we may be undercutting the ability of insurers to negotiate discounts because that cat’s out of the bag.

Will AI help the U.S. control healthcare costs?

Sommers: Given that we already went through a laundry list of the things that we hoped were going to save us money and turned out not to really be that effective, my hunch is AI is probably going to be in that category, but let’s not just be reactionary. Let me actually give you a more analytical answer to that.

What is AI potentially going to do well in healthcare? It may allow us to outsource some of what currently you have more expensive clinicians doing. Maybe some of the letter writing for test results. We’re already seeing AI scribes that are used to simplify note writing for doctors. You know, what about reading EKGs and radiology? Maybe you don’t need a doctor. Maybe one doctor can supervise an AI system and there’s a total number of reports they can write in an hour can proliferate rapidly.

All of those things could, in theory, reduce the cost of each of those services. And so that would be a way of reducing spending. But I think there’s some countervailing factors we have to think about.

The first is it may be that because now they’re more efficient that they’re just gonna make up for that change with more volume. Now it’s easy for me to read scans. We can just do a whole lot more. Or EKGs don’t take as much time. Like we cut the time for reading it from a minute to 2 seconds, we should just be getting these on everybody because it’s super fast. So we increase the volume in a way that counteracts the fact that we think we’re more efficient in the individual unit.

The other thing to recognize is AI also has a potentially very large use in revenue maximization, which is if you’re a hospital and you want to make sure you’re getting the most back for your hospital stay on a particular patient, put AI into the chart, come up with which diagnosis codes that could plausibly fit this, the notes that are written here and the testing that’s been done, what maximizes revenue?

You could very easily imagine AI as a way of actually increasing the coding intensity and would lead to higher reimbursement rates. So it’s a bit of an arms race because AI could be used to reduce costs in some settings, but it could be used to maximize revenue in other settings. As we started talking at right at the outset, providers maximizing revenue, that’s increasing costs to the system. We think of it as, I’m making my practice more profitable. That drives up healthcare costs across the board. So I would be very cautious in thinking AI is going to fix our problems. And I think in some settings, it could very well make them worse.

Other Interviews in This Series:

Price Controls ‘Inevitable’ in U.S. Healthcare, Economist Says

Hospital Prices Drive High Healthcare Costs, Economist Says

Single Payer ‘Hands Down’ the Best Way to Solve High Healthcare Costs, Advocate Says