Skip to main content
Community-Focused HealthInsiderOrganizational TransformationPayment Redesign

COVID-19 is testing the Dartmouth Atlas hypothesis: An interview with Peter Orszag

By September 8, 2021August 2nd, 2022No Comments

Peter Orszag, CEO of Financial Advisory at Lazard and former Director of the U.S. Office of Management and Budget, said at the Health Evolution Summit that the reduction in non-COVID care during 2020-2021 will yield more insight into the realities of care costs and outcomes.  

“We are living through a time in which a lot of care was not delivered and there is preliminary evidence that the excess mortality we saw in 2020 and into 2021 is not just linked to the direct effects of COVID. Whether it is linked to averted care and, therefore, speaks to a revision of the Dartmouth hypothesis will provide a lot of fodder for academics and for practitioners going forward,” Orszag said at the Summit. “This is an absolutely essential question to the future of health care in the United States and we should learn something from what is occurring.”  

Health Evolution Editor-in-Chief Tom Sullivan interviewed Orszag about how COVID-19 is testing the Dartmouth Atlas hypothesis, what he’s expecting to come out of this and how health care CEOs should be thinking about the way it will change their business.  

You said that America has gone through an absolutely massive experiment to test the Dartmouth Atlas hypothesis. Can you expand on that?  
Orszag: Let’s start with what we know so far and then what the pandemic represents in terms of being able to teach us something additional. It’s well understood that care tends to vary in quality and costs of care vary significantly across the country, especially with Medicare. It’s also true within employer sponsored insurance but for different reasons. One of the outstanding questions has always been “okay, we know that care variation occurs but does it reflect differences in outcomes?” In other words, does more care intensity at a higher cost produce better outcomes? The key finding of the Dartmouth Atlas is that there was no such correlation.  

The challenge becomes whether there are false correlations because of other factors driving what’s going on, which is to say maybe the people in Florida are just unhealthy or less healthy than people in Minnesota in ways that the risk adjusters don’t capture. So, yes, you spend more on them and the outcomes are no better, and maybe worse, but it’s because there is a hidden variable that makes it so they are just not as healthy as the folks in Minnesota. The problem is you cannot actually conclude anything from this kind of pattern across the country. 

In the past five or ten years, some clever studies have looked at people who moved from one area of the country to another. Let’s say someone moves from Florida to Minnesota. If the spending that is actually occurring is just about the individual and not about local practice norms, when you move to a higher spending area then nothing should happen. But if a person moves from Minnesota to Florida, the amount spent on that individual goes up significantly. And if one moves from Florida to Minnesota the amount spent goes down significantly. So, that’s really interesting because it helps to identify that this is a variation in the way we practice, at least roughly speaking. Of course, the academics can quibble with even that by basically saying “well, maybe someone who’s in Minnesota and is not well decides to move to Florida for the sunshine and so there’s still this hidden variable.” But I’d say it’s pretty compelling evidence. Especially because there have been other studies that look at military families who are moved from one area to another for reasons that are clearly not about their health and we see similar spending patterns.  

 The Dartmouth Atlas investigators and authors broke ground with that research and today the finding has been broadly accepted among health care executives and policymakers. How has COVID-19 tested that hypothesis?  
Orszag: We’ve just seen a massive decline in non-COVID care, especially the care that was considered nonessential or not urgent. Discretionary care plummeted, especially early on, from the spring to late summer of 2020. The degree in which care plummeted varies from state to state and county to county in interesting ways because it depends a little bit on how bad the COVID situation is in each area, how much concern people had about going outside or visiting a doctor, or into the hospital. As a result, we have experienced this very significant pattern of not only big declines but also the rate of decline differs significantly in, for example, Connecticut versus Massachusetts from city to city and so that’s a factor. And we have been seeing increased mortality in 2020 and 2021 beyond just the effects of COVID. Mortality rates are up, life expectancy has gone down too and not all of that is because of COVID. There’s a non-COVID excess death rate and the question becomes: Why is that? Is that mismeasurement such that it’s actually COVID but wasn’t counted as COVID? Or a mental health issue and other problems because people are cooped up in their homes and that causes all sorts of other issues? Or lack of social interaction?  

We know that there was a whole bunch of health care that was not delivered in 2020 and 2021 and it caused bad outcomes for people including potentially actual death. That’s really kind of right at the heart of the Dartmouth findings. There’s a considerable amount of health care delivered in the U.S. that does not help people. What we may be on the verge of being able to find out is which categories of care that were forgone actually lead to people’s health deteriorating. That’s the question and since the types of care that were cut back and the degree to which they were cut back varied across different parts of the U.S., we have the possibility for some promising academic to get tenure in the future by really nailing down what we learn from the great health care cut back of 2020 and 2021. 

At this point, the manner of analyses you described have yet to be completed. But what are you expecting to come out of this?  
Orszag: We’ve never had this sort of dramatic experiment and I suspect that what we will learn is that there’s a lot of care that can be cut back without any adverse consequences and that foregoing some care did harm people’s lives. Data relative to those could show us what the promising pathway forward is because one of the other complaints has been “well, it’s great that we have all these kinds of high-level correlations but unless we get much more specific about what kind of care matters and when spending more matters, it’s not that helpful.” For example, there is some evidence suggesting that higher cost emergency rooms produce better outcomes. That means spending more on the ER can be a good thing and that stands in contrast to some of the evidence about post-acute care where it appears the correlation may be negative. In other words, higher cost post-acute care is not better and I think it’s that kind of a nuance that we might learn a lot more about. I don’t know exactly what the findings will be  if I did that would be remarkable  but I would be surprised if we don’t learn something important from the set of changes that were this significant. The size of the change and also the fact that it varied across different parts of the country in such significant ways, really hold up some hope that we might be able to examine the outcomes and learn from that. The big promise here is that the more we learn which parts of health care spending we can cut back on without harming people, the better. 

The potential for that type of change is significant enough that many health care leaders may want to start thinking about or even conducting their own internal analysis into how the pandemic impacts care delivery, operations and revenue now rather than waiting for someone else to do so. What from your perspective should CEOs be thinking about today, strategy-wise, to prepare for the future? 
Orszag: Looking at this within the context of what’s going to happen over the next five or ten years, we have heard a little bit from the administration about a renewed push away from fee for service and toward value-based payments. Some of those may turn out to be mandatory for providers, and payers are clearly pushing in the same direction. So if you’re a health care executive or a health insurance CEO, it’s time to evaluate what could be a different set of financial incentives and increasing value. How do I do that? Where do I look for the savings in a way that doesn’t hurt people and may benefit the organization? CEOs will learn a lot from the great health care experiment of 2020-2021 that can help guide response and strategy in terms of the new financial incentives that will be more pervasive in the future than they were in the past. 

Wright Lassiter [CEO, Henry Ford Health System] said at the Summit that this data could impact care delivery as well as potentially be used to address health equity. At the very least it could expose greater inequities than are widely understood today …  
Orszag: There are a couple of dimensions to that. The first is that within COVID, we are aware of the fact that it is quite disproportionate and that there are health equity concerns around who gets affected the most. So that’s the first point. Outside of COVID, I also think we’re going to learn, because I don’t know this for a fact since I haven’t seen the sort of detailed data, but it wouldn’t be surprising if there was also variation in who cut back on care and where and how. That variation is very informative and useful in studying discretionary care. It could be more important to a diverse population than in other settings or it may be more important to low-income Americans and higher income Americans for various reasons. Again, we’ve undergone this math experiment as an unfortunate side effect of the pandemic. We might as well get something in return for the significant shifts that were happening in the pattern of who receives health care. There’s no doubt that varied significantly by income and socioeconomic status and race and gender and all the other dimensions of health equity. So, I would have to agree with that comment, I think you’re going to see massive differences in who cut back, what kind of care and potentially what the outcome implications are from different kinds of reductions. I would be shocked if that’s not the case.  

You mentioned earlier that this could data could help inform strategies for a path forward. What shape do you envision that path taking?  
Orszag: We’re going to have the potential for a lot better information in terms of how to steer the health care system to a more promising future. The way forward is a health care system that learns from this great experiment, that is much more digitized in the future, that is much more value-based in terms of the incentive schemes, and that can adapt to whatever we learn about the care that was cut back significantly and turned out not to harm anyone. We also need to be super careful as we move forward to focus on the trade-offs between spending and outcomes and be really careful not to cut back on that kind of care that we just showed cutting back on could potentially even kill people when you get right down to it. 

Tom Sullivan

Tom Sullivan brings more than two decades in editing and journalism experience to Health Evolution. Sullivan most recently served as Editor-in-Chief at HIMSS, leading Healthcare IT News, Health Finance, MobiHealthNews. Prior to HIMSS Media, Sullivan was News Editor of IDG’s InfoWorld, directing a dozen reporters’ coverage for the weekly print publication and daily website.