This transcript is taken from an interview conducted April 7, 2020.
The problem with producing ventilators in such a short time frame is that you’re ordering them from manufacturing plants that typically make products that are very different from ventilators. The factories are configured to supply parts for automobiles, and they’re being asked to turn on a dime and start producing ventilators. In a typical manufacturing environment, retooling and reconfiguring manufacturing lines is an arduous task. It takes a lot of engineering effort and testing, and it’s unclear whether enough production capacity could be built in a week or two.
Not only that, but new manufacturing processes have to go through a series of rigorous quality-control checks. Ventilators are sophisticated devices that allow you to control the flow of oxygen carefully, and I worry about whether US manufacturers can quickly produce defect-free, usable ventilators. In the worst-case scenario, you could have ventilators go out into the field and fail on patients.
We need another plan that allows us to make use of the ventilators that already exist. It might be that the ventilators we have around the country are enough to save many more lives. The problem is they may be in the wrong place at the wrong time.
If COVID-19 were to peak in all states at the same moment, we wouldn’t have enough. But in the way this pandemic is evolving, states will peak at different times. As that happens, ventilators will become less in need in some parts of the country and more in need in other parts. They could be moved.
You would need to have states agree to exchange ventilators. We have seen some of that already, for example with Oregon offering ventilators to New York. But this kind of exchange would require a massive logistical effort on the part of the federal government, and likely include the Army, which has expertise moving military equipment and tanks on the battlefield. It’s a similar problem, but you need good optimization and logistical tools.
The problem with producing ventilators in such a quick timeframe is that you’re asking manufacturing plants that make products that are very different from ventilators, in particular automobiles, and supply chains that are configured to supply parts for automobiles to suddenly turn on a dime and start producing ventilators.
In a typical manufacturing environment, retooling and reconfiguring manufacturing lines is a really arduous task. It takes a lot of engineering effort, a lot of testing, and it’s just unclear whether or not enough production capacity could be built in really just a short time, a week or two weeks or something like that.
Not only that, but when you build manufacturing plants, it’s not like you just start producing ventilators. You’ve got to go through a series of very rigorous quality-control checks to ensure quality. And these ventilators are not simple devices. They’re very sophisticated devices where you can control the flow of oxygen very carefully and whatnot, and so I would very seriously worry about relying on the manufacturers in the US to so quickly put together a production process that’s able to produce defect-free, usable ventilators. In fact, you know, what might happen in the worst case is you could have ventilators that go out into the field, and they might fail on a patient in the middle of the night because it didn’t go through all the right quality assurance processes that we would normally undertake. And so if it fails in the middle of the night, the patient dies.
So I think we need another plan other than just relying on manufacturers. It’s great that they’re doing that, but I think we need another plan to make use of the ventilators that do exist. The idea is that there’s ventilators all around the country. And it might be the case that even in aggregate, we have enough ventilators, but the problem is they’re in the wrong place at the wrong time.
If all states were peaking at the same moment and needed at the same moment the maximum number of ventilators, we’re not going to have enough. But the way this pandemic is evolving is that different states will peak at different times, and so after New York peaks, it’ll kind of move along different parts of the country. And as that happens, ventilators become less in need in some parts of the country and more in need in other parts of the country, and they could be moved.
However, in order to do that you need to have states agree to exchange ventilators. But this would require a massive logistical effort on the part of our government, the federal government, and likely the Army, which is used to moving military equipment, tanks and things on the battlefield in real time for where they’re needed. It’s a similar kind of problem where you really need to be able to move ventilators, in this case, around the country to where they’re needed. But in order to do this you need good optimization and good logistical tools to help plan it out.
To figure out what the benefit of doing this is, you’ve got to have a model of demand, and so you need to have a forecast of how the pandemic is going to evolve over time. And, in particular, what you need is a forecast of new ICU admissions. We estimate that roughly 80 percent or so of admissions to the intensive care unit will require mechanical ventilation, and based on that assumption and other assumptions like the length of time that patients need to be on ventilators and whatnot, you can then put all this together along with the supply of ventilators across the country, if you’ve got access to those numbers, and you can put them all into an optimization model that then figures out how to move the ventilators around to maximize the number of lives saved.
Now all of this is very much contingent on the assumptions you bring in. And so there are many, many forecasting models out there now, and the one that I chose to use is, in fact, the one that the federal government is using for its planning. My forecasts come from the Institute for Health Metrics and Evaluation from the University of Washington, and they have these three scenarios: there’s the expected trajectory; there’s a lower bound, which is like a best-case scenario; and then there’s an upper bound, which represents the worst-case scenario.
Starting from that, you can then figure out what effect they provide in their data forecast of new ICU admissions. And from that you can build a model, in this case of three different scenarios of what the demand will be for ventilators over time, state by state.
In the upper-bound scenario, the worst-case scenario, I estimate that without the national stockpile ventilators being used, we could lose about 80,000 to 83,000 lives. With the best-case scenario, without the national stockpile, just where the states are all acting independently, we could lose about 8,000 lives due to there not being a mechanical ventilator.
From there, you then have to consider different scenarios. The most likely scenario, based on the way the federal government’s operating now, is that they’re basically just giving their national stockpile allocated to locations. It’s unclear whether or not there’s plans for them to actually be moved once they’re in position.
If you allocate the national stockpile ventilators as a one-time allocation to states and then they’re not moved around, there’s obviously substantial benefit from doing that based on an estimate of 8,900 ventilators in the stockpile. And it’s unclear exactly how many there are. I estimate that you could save about 5,500 lives in the best-case scenario, and you could save up to 14,000 lives just by doing a one-time allocation of the national stockpile.
But once you allow the ventilators to circulate in exchanges between states, you can get substantially more benefits. So in the best-case scenario, in the lower bound of the forecast for demand, you can save approximately another 1,500 to 2,000 lives just by allowing these exchanges of the ventilators across the states. And in this worst-case scenario, the upper bound of the forecast, you could save as many as about 14,000 additional lives by just allowing the ventilators to circulate. So that’s a pretty wide range.
Now, of course, all this is contingent on the forecast that you’re using, but it’s also contingent on the estimate of supply that you have, and for that there’s not really great information about how many ventilators each state has. There was a study done in 2010 that’s in the published literature that breaks it down based on a survey of how many ventilators each state has in its possession, so I’m using those estimates for the number of ventilators. These are full-featured ventilators, and I’m subtracting off the ventilators that need to be used for other patients in the ICU. It’s not like during this time other patients don’t need them, and so in reality, in terms of the number of ventilators in the country, there may be quite a bit more.
There are other kinds of ventilators. I’m only looking at full-featured mechanical ventilators, which would represent sort of the American standard of care for how we would treat these patients. But there may be other kinds of ventilators that could be retrofitted, older ventilators that can be brought back into use, and it’s unclear what those numbers are, how many of them we have. Presumably, hopefully, the federal government would have that. Certainly the states, hopefully, would have that information.
The other thing, which is really important, is how are you going to get the states to actually exchange ventilators? When I’ve talked to some people, the response that I’ve gotten is that no one’s going to let you move their ventilators. States are kind of in a position where they’re hoarding their ventilators in anticipation of this peak coming. There’s uncertainty around how big the peak is going to be, and so it’s not a real easy matter for states to actually release these ventilators right now.
I also consider the situation where states freely exchange ventilators even before the peak, and this is logistically a bit more challenging because the idea is you would have to have a state predict, with reasonably good accuracy, what its peak is going to look like, ship out their ventilators, and if they ship out ventilators they’ll need later at their peak, they’re going to need to make sure they get them back. And that’s problematic.
But there also might be a situation where even in the worst-case scenario, some states have far more ventilators than they need anyway. They have things to spare. In either case, in this second run of the model, I actually allow the ventilators to move anytime, and once you do that, the number of lives you can save in addition to the original estimates is many, many more thousands of people.
To figure out the benefit of doing this, you’ve got to have a model of demand. You need to forecast how the pandemic is going to evolve. In particular, you need a forecast of new intensive-care-unit admissions. Recent medical evidence suggests that roughly 80 percent of people admitted to the ICU with COVID-19 will require mechanical ventilation. We can put this assumption and others, such as the length of time that patients need to be on ventilators, and mortality, along with the supply of ventilators across the country, into an optimization model to figure out how to move the ventilators around to maximize the number of lives saved.
There are many forecasting models out there. My forecast comes from the Institute for Health Metrics and Evaluation (IHME) at the University of Washington, which has three scenarios of the ICU-admissions trajectory. It is also the one that the federal government is using for its planning. Therefore, whether or not this model produces an accurate forecast, our results will be applicable to the government’s plan.
Plotting the trajectory is like forecasting hurricanes. There’s an expected path, and there are uncertainty bounds around that path. The lower bound is a best-case scenario. The upper bound represents the worst-case scenario. Starting from these, you can forecast new ICU admissions, and from that you can build a model—in this case, of three scenarios— that plans the exchanges of ventilators across states over time so as to maximize the number of lives saved.
I estimate that in the worst-case scenario, without the national stockpile of mechanical ventilators being used, you could lose over 83,000 lives. Even in the best-case scenario, without the national stockpile used, where the states all act independently, you could lose about 8,000 lives due to the lack of readily available ventilators.
In my model, I assume that states would only allow ventilators to be shared after the virus peaks, when they might not need them anymore, but while other states are in need.
But if you allocate the national stockpile of an estimated 8,900 ventilators to the states, and don’t move them further, I estimate that you could save about 5,500 lives in the best-case scenario.
Once you allow ventilators to circulate between states, you can get substantially more benefits. In the best-case scenario, using the lower bound of the forecast for demand, you can save another 1,500–2,000 lives, just by allowing this exchange of ventilators across the states. And in the worst-case scenario, you could save as many as 14,000 additional lives by allowing the ventilators to circulate. In total, by putting to optimal use all of the existing mechanical ventilators in the country, over 28,000 lives could be saved, according to my model and assumptions.
This forecast is contingent on an estimate of how many ventilators each state has. There’s not great information for these numbers. A 2010 study broke down supply on the basis of a survey that asked how many full-featured ventilators each state had in its possession. I used those estimates and subtracted the number of ventilators that would need to be used for other patients in the ICU. There may be other kinds of ventilators that could be retrofitted—perhaps older ventilators that could be brought back into use. It’s unclear how many of those there are.
If you assume that the national stockpile of ventilators is allocated in an optimal way (though it may or may not be), you can ask the questions: What’s the incremental benefit of allowing just the national stockpile of ventilators to move? Just the hospital-owned ventilators to move? Or any of them to move? My model indicates that as you progress from one scenario to the next, the benefits increase.
If you allow just the national stockpile of ventilators to be exchanged between states, you can save an additional 5,000 lives. If, instead, you allow the hospital-owned ventilators to circulate, you also save around 5,000 lives. And if you allow any ventilator to move, you still save about 5,000 lives.
If you look at the worst-case scenario, when you allow the national stockpile of ventilators to move around, you can save approximately an additional 8,000 lives. If you just move the hospital-owned ones, you can save on the order of 11,000 additional lives; and if you allow all ventilators in the country to circulate, my estimates are that you can save an additional 14,000 lives on top of a one-shot allocation of the national stockpile. That still would leave the country with a severe shortage of ventilators, with nearly 55,000 lives lost due to an insufficient number of ventilators being available. It could be lower than this if the IHME worst-case forecasts turn out to have been too high.
States have to know where ventilators are needed in their states, and be able to conduct their own ventilator-sharing operation—or they have to develop this capability, with integration and supply-chain optimization that would anticipate demand and move ventilators accordingly.
The federal government needs real-time information on the status of ventilators all around the country—an information pipeline, which I suspect it has for some states and not for others, as some states likely have this information, and others likely don’t.
And you need good calculations to conclude when and which states have more than they can use. Apparently, Oregon has made that calculation, or it has some assurance from the federal government that if ventilators are moved before the state peaks, they can come back. But states are peaking within a matter of weeks of each other. There would have to be confidence that the US government could manage in that situation.
Another thing is really important: How are you going to get the states to actually exchange ventilators? When I’ve talked with some hospital executives, the response has been, “Well, no one’s going to let you move their ventilators.” There’s uncertainty among states around how big their peak is going to be. Politically, most state governors aren’t going to give away their ventilators and risk the public health of their own citizens.
Everyone is so focused on the ramp-up to the peak, and not many people are thinking about what happens on the other side, when they’ll be faced with the situation of having ventilators and wondering where to send them.
Could the federal government, using its war-powers authority, commandeer ventilators around the country? I don’t know the answer. Even if it could, without knowing exactly where the ventilators need to go, the federal government would be carrying out a dangerous exercise. The best way to do this would be to have states voluntarily offer up ventilators when the time comes. The American Hospital Association has recently partnered with the Federal Emergency Management Agency to start such a voluntary-sharing program, but a limited number of ventilators have been offered up as of yet.
For these reasons, in my model, I assume that states would only allow ventilators to be shared after the virus peaks, when they might not need them anymore, but while other states are in need. I would hope that the climate and the political incentives to share ventilators would be significantly greater at that point, making sharing a viable alternative and easier to manage.
(For completeness, I also considered a situation where states freely exchange ventilators even before they peak. This is logistically more challenging, because you would need a state to predict with reasonably good accuracy what its peak will look like. If one state shipped out ventilators it would need later, it would have to be able to get them back. Or some states could have far more ventilators than they need and would have some to spare. In either case, allowing the ventilators to move anytime resulted in many more thousands of lives saved.)
Everyone is so focused on the ramp-up to the peak, and not many people are thinking about what happens on the other side, when they’ll be faced with the situation of having ventilators and wondering where to send them. The federal government needs to be in a position to move those around in the right way, and it needs the right optimization tools to figure out where to send them.
Research takes a broad view of what happened when Chicago tore down distressed public housing.
Line of Inquiry: Milena Almagro on Public Housing Demolition and Urban Inequality
Policies marginalize poorer consumers and consumers with niche tastes, disadvantaging small businesses.
Data Privacy Is for the Privileged
Booth’s Raghuram G. Rajan and Martin Wolf of the Financial Times discuss the pressures faced by democratic systems.
How Can We Restore Trust in Democracy?Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.