Ramping Up Manufacturing of Ventilators and Other Medical Supplies


In a previous post, I described a simple mechanism by which the government can encourage private spending toward some goal, such as reducing the spread of COVID-19. That mechanism is detailed in my recently completed paper, currently under submission to law reviews, Random Selection for Scaling Standards, with Applications to Climate Change. I received excellent comments and will return in a later post to answer some of the questions and objections raised by commentators. But I now want to make the previous post more narrow, by focusing on a subset of the COVID-19 spending problem, specifically the challenge of ramping up the manufacturing of medical supplies. Tyler Cowen's Emergent Venture Prizes generously announced $1 million for rewards to be given ex post for COVID-19 work.  My suggestion here is that the government should spend far more than this to ramp up manufacturing, relying primarily on ex post evaluations rather than ex ante contracts.

It is now clear that there will be a massive shortage of certain medical products. Let's just focus on one particularly important product category: ventilators. If we had enough ventilators, many more future critical patients could be saved than with current capacity. Along with many other improvements, this might increase the production capacity of the healthcare system, eventually reducing the extent to which the curve must be flattened. If we no longer needed to worry about "flattening" the curve, global economic activity could resume, reducing the devastation of the inevitable recession. Thus, increasing ventilator production ought to be one (of many) important public policy goals.

So, what is happening? Well, the U.K. (whose government has resisted the urge to flatten the curve on the theory that it is better to generate herd immunity by allowing the virus to spread rapidly in low-vulnerability populations) has urged industry to switch to making ventilators. The U.S. hasn't done much of anything at all as a policy matter. According to Forbes, ventilator firms themselves seem to think that they could increase production within a few months, but hospitals may not place orders that far into the future, given the risk that the machines will not be necessary. One imagines that ventilator firms and hospitals will increase production and orders considerably, because they want to do the right thing and be perceived as doing the right thing, but the overall level of ramping up might be considerably less than the social optimum. Rules and norms against price gouging limit the incentive to expand new production.

Vox reports that if the situation were similar to the Spanish flu pandemic, approximately 750,000 people would need ventilation. Even under worst-case scenarios, they would not need them all at once. Suppose that we need one-third of that capacity at any given time and that we accept Vox's estimate that we currently have about 160,000 altogether. That suggests that an order of 100,000 ventilators would greatly increase our ability to meet this challenge (though still possibly fall short, given non-COVID-19 patients needing ventilation). Hospital-grade ventilators typically cost around $25,000, so if we were willing to wait indefinitely for ventilators, the total bill for this country would be around $2.5 billion. But we need these ventilators to be made very quickly, and that would presumably be a lot more expensive. Not only would ventilator makers need to increase their production capacity dramatically, but so too would their parts makers (and the makers of parts for those parts, and so on to not-quite-infinite regress). Even if the government were to spend $25 billion on 100,000 ventilators, that would easily pass cost-benefit analysis if just 5,000 lives could be saved, with a valuation of life of $5,000,000. It seems plausible that the investment could save many more lives than that, especially if the private sector is able to produce a larger number of ventilators.

I don't have the knowledge to speculate how much one would need to spend to persuade ventilator makers and parts makers to greatly increase their production capacity quickly. Perhaps industry or government could study the question, but that would involve a fair amount of guesswork, requiring speculation about which parts of the supply chain will turn out to be bottlenecks and how expensive it would be to develop alternatives. The Hayekian claim would be that no one has the relevant knowledge, some of which doesn't exist yet and some of which is diffused through markets. In any event, while even preliminary estimates would help, we don't have time for a detailed study.

The conventional approach to this problem would be for the government (presumably the federal government, though one could also imagine states acting independently) to engage in procurement, letting private firms bid on price and perhaps some other variables. A benefit of this approach is that our legal institutions for conducting such procurements are greatly developed. There are even rules that relax other rules for emergencies, including "when the President has made a declaration under the Robert T. Stafford Disaster Relief and Emergency Assistance Act," as the President has now done. For example under FAR 18.125, "agency protest override procedures allow the head of the contracting activity to determine that the contracting process may continue after GAO has received a protest."

Still, there are reasons to wonder about the government's ability to execute effectively. The absence of any call for proposals even today is not encouraging. Consider also at the government's procurement of surgical masks. There was a two-week solicitation period ending March 18—very fast by government standards, but still, a long time by the time scale of a pandemic where cases are doubling every few days. Perhaps of greater concern is that the solicitation asks for "a proposed schedule for items that can be delivered quickly without overwhelming the supply chain, with additional items delivered over a period not to exceed 18 months." Presumably, proposals that offer a greater proportion of masks sooner will be favored, but the solicitation gives no indication of how much of a bonus earlier delivery would be, and there is no statement that the goal is to greatly expand maximum supply capacity. It seems unlikely that anyone will perform even a back-of-the-envelope estimate of how much more valuable a mask would be if that mask could be produced next month than six months from now, let alone include performance incentives based on such a calculation.

Similarly, with ventilators, the ultimate question is how many QALYs will be saved by the production of a particular new ventilator using a particular technology at a particular time. I am skeptical that the government has the foresight to assess this for all types of ventilators, free from political considerations, let alone the ability to modify the contracting process over time. Yet it seems plausible that a few years from now, the government might be able to produce reasonable retrospective estimates of the utility provided by different ventilators at different times, based on statistical evidence about when patients needed ventilators and when they performed best. A system that relies more on ex post governmental assessments than on ex ante governmental projections allows the market to respond more quickly and with greater agility, changing plans and making adjustments as necessary. This is especially important in an emergency, where uncertainty is great and time is critical, but it matters at other times too.

So, how might an ex post decisionmaking approach work? The government commits to allocating $25 billion for ventilators. (Perhaps that could be limited to increased production over previous capacity, but it also sets a good precedent to reward producers for the production capacity they offered as of the beginning of a crisis.) The reason for the government to commit to payment is that the government might otherwise have incentives to lowball after the fact. Given the Winstar case, the government probably could not easily escape a commitment of this sort. An alternative approach would be for the government to set a range (say $15 billion to $60 billion), with the total amount to be based on the total number of QALYs estimated to have been saved in the country, perhaps at $100,000 per QALY, but constrained to this range.

Place aside for now the challenges involved in the ex post adjudication of QALYs, and consider the incentives of ventilator companies and providers of capital on the assumption that even if the decisionmaking process is imperfect, one's best guess of the government's evaluation of the contribution of a particular ventilator produced will be roughly proportional to the number of QALYs that it saves. What would a governmental commitment of this sort achieve? Ventilator companies will seek to make production decisions that maximize QALYs saved per dollar spent. That may affect which models they invest in. For example, some might decide that it's worth producing many simpler ventilators, on the assumption that many imperfect ventilators (perhaps even models similar to DIY designs, if they believe that regulations could be changed to allow their use) are better than a small number of perfect ones. Moreover, ventilator companies could adjust to changing information, about both the disease and parts supplier capabilities, in making decisions, without the need for cumbersome procurement change orders.

The most obvious objection to this approach is the claim that ventilator companies would not be able to handle the risk. But capital markets are vehicles for eliminating risk through diversification. Pharmaceutical companies and others take massive risks all the time. Moreover, ventilator companies could raise money specifically for this purpose. The rights to eventual governmental payments could be tradeable, so the ventilator companies could exchange some of these tradable rights for funders, or they could do the more conventional thing of creating subsidiaries and selling stock in the subsidiaries. Even if any particular ventilator investment is risky, these investments can be part of a diversified portfolio. With interest rates exceptionally low, the claim that ventilator companies would not invest is absurd. Sure, government contractors may be accustomed to greater certainty, but there is no reason that they cannot function in such an environment. The risk might be reduced by requirements that companies can only receive reimbursement for efforts that they disclose in advance or shortly after beginning, so that each company can assess what others are doing and gauge how much total investment is chasing the government's reward money.

Another objection is that the focus solely on ventilators is arbitrary. If the government procures ventilators but not other necessary life-saving equipment, the ventilators might not be used. Moreover, just as the government may not be well positioned to determine in advance exactly what type of ventilators maximized QALYs per dollar, so too might the government not be so well positioned to choose how to distribute total available government funds among alternative life-saving investments. In my original proposal, I suggested that perhaps the government might spend a trillion dollars to reward all investments, including medical supplies like protective gear and R&D on new drugs. That would give the private sector to assess what type of investments are most appropriate.

One might object that some level of generality will be too much. After all, even those who applaud ex post decisionmaking for a specific application will likely not want a government fund to be used for private contributions of any sort, whether or not they relate to COVID-19. There is some value to political accountability. Perhaps a fund for "hospital equipment and supplies" is the right level of generality, or perhaps "ventilators" is. Yet few would say that political accountability demands that members of Congress or the President choose ventilator specifications. Thus, it seems reasonable to me to have a ventilator fund, as well as maybe other funds.

Another argument for a relatively narrow level of generality in construction of a fund is that it may make it easier to create an ex post decisionmaking process. It seems more plausible that a group of experts could compare the success of different ventilator producers than that they could compare ventilators against many other products, let alone make apples-to-oranges comparisons across many different types of responses to the COVID-19 epidemic. It might seem that a decisive argument against ex post decisionmaking, especially if the fund is at a relatively high level of generality, is that there will be a large number of complicated decisions to make. This can lead to one of two results. First, enormous sums might be spent on the adjudication process, and these anticipated transactions costs are moneys that otherwise could be spent on medical equipment. Second, decisionmakers might make relatively superficial decisions. Anticipation of that might distort private decisionmaking. Why make a slight improvement to a ventilator if one does not think that the eventual decisionmakers will notice it?

These concerns help explain the most unusual aspect of my proposal, the use of random selection. This aspect is not necessary if the level of generality is ventilator production, however, because there may be a relatively small number of claimants on the fund. At a broad level, my goal is to argue for distribution of government funds based on ex post evaluations, particularly in circumstances of emergency, and this argument suffices for my proposal here to spur production of ventilators (and perhaps a few other key products). It should not be difficult to craft legislation accomplishing this proposal, especially since the details of ex post adjudication can be worked out by an administrative agency later. It requires much less governmental advance planning than traditional approaches to government spending, including congressional authorizations and procurement. The government can do this, and it seems much more valuable than most of what the government is doing instead.

I recognize that the idea of using random selection may be too new to be considered for any legislation to be passed in the short term. Still, it's worth explaining when random selection would be useful in an ex post decisionmaking system, as well as how and why it would work. The case for random selection increases in importance, the greater the level of generality and the greater the number of heterogeneous claims that might be filed. The beauty of random selection is that we could use ex post decisionmaking even if the government expected literally millions of different claims for many different types of contributions. That would be more likely for the climate change applications that I address in my paper, but could apply as well to a COVID-19 fund at a level of generality considerably higher than ventilator production.

How does random selection work? The government chooses only a small number of claims, say 100, and it distributes the entire fund to the owners of these claims in proportion to the measurement of social value (whether QALYs or broader). Everyone else gets nothing. (If the fund is variable in size, instead of fixed at a number like $25 billion, the total size of the fund would be extrapolated based on social value estimates extrapolated from the cases randomly adjudicated.) The advantage of this approach is pretty simple: Many fewer cases can be adjudicated, and so there will be sufficient resources to adjudicate each case with some care.

The disadvantage is that the proposal seems like a lottery. The project of my article is to argue that this is not a problem, that capital markets can absorb the risk. I will not repeat all of the relevant points here, but I will sketch out the general argument: Intermediary companies would be expected to buy up diversified portfolios of claims based on their assessments of how the government might value them if randomly selected. All producers need concern themselves with is how much these intermediaries will pay. (Some large producers might not need intermediaries, relying instead on other capital market tools.) Assuming there is sufficient competition among intermediaries, the amount the intermediaries pay should be equal to the amount they expect to receive minus the costs incurred by the intermediaries themselves. The intermediaries provide a socially useful service of evaluating producers' plans (if the purchase is before production) and achievements (if the purchase is after production), as well as in prosecuting the few claims randomly selected for ex post evaluation.

To an intermediary or other diversified firm, it doesn't matter that 1/n claims will be worth n times socially assessed value, because the expected value of a claim remains the same. Moreover, diversification does not need to be perfect. Random selection of claims by the government is a risk that can be precisely mathematically evaluated, and thus it is insurable. Indeed, the government itself could and should offer insurance at actuarially fair rates. Thus, an intermediary should be able to largely eliminate the risk attributable to uncertainty about which claims will be randomly selected. The intermediary will still face risk associated with randomness or arbitrariness by governmental decisionmakers in their evaluations. But because relatively few claims are adjudicated, these adjudications can be performed with care by multi-member panels, using expert testimony where necessary, thus reducing this form of uncertainty as well. The knowledge that any eventual adjudication will be careful will have ex ante benefits, as intermediaries will expect that the decisionmakers will notice subtle advantages of their products.

NEXT: Adjudication Outside Article III (part one)

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. “Vox reports ”

    Using Vox for any fact is very questionable.

    1. Which is a logical fallacy. That is a big deal at a site called “Reason.”

      1. 3D printers aren’t Star Trek replicators (yet). They’re also slower than conventional manufacturing techniques.

        1. Damn….ignore that (here, at least). The page moved out from under me as I was hitting the “Reply” button.

  2. The problem is not the number of ventilators, the problem is the number of people that are qualified to run them. That is the bottleneck.

    1. There may be two bottlenecks here, but I agree with you that is an issue. I considered discussing this issue and may yet. The government also needs to move quickly to encourage training. Of course, I’d rather have a technician with years of training, but I’d rather have a ventilator plus an operator with one month of training (ideally, operating under a more experienced technician) than no treatment at all.

      1. That is like sending someone with a couple of weeks of ground school up in a plane: get the wrong ventilation mode or settings, and there is a dead patient. Even worse, if you don’t know exactly what you are doing there is a huge risk of aerosolization which puts everyone else at risk.

        1. How long to train a reasonably intelligent person? Think someone who has basic training as a medic = people that are qualified to run them. Best guess. Is that a week? A month? A year?

  3. Are ventilators good candidates for 3-D printing?

  4. The hypocrisy of banning despicable price gouging by those who actually provide emergency supplies, while tossing prize money around like candy with little regard for merit, is delicious.

    What’s that you say, the prizes will be based on merit? Oh come sir, this is the government! Your merit is everybody else’s corruption.

    1. That’s some very strong question begging right there.

  5. 3D printers aren’t Star Trek replicators (yet). They’re also slower than conventional manufacturing techniques.


    A growing number of medical commentators on national television are now suggesting that the shortfall of hospital ventilators – not the shortfall of hospital beds, ICU beds, personal protective equipment [PPE], or even medical personnel – will be the most critical factor in mortality from the COVID-19 disease, and the most likely to force doctors to choose which patients live or die.

    But there may be a way to temporary alleviate, at least part of the problem, using a “poor man’s ventilator”

    Millions of Americans already use CPAPs – machines which force air under pressure into the lungs of users – to limit snoring and/or to treat sleep apnea, so they are currently available, and there are many more in reserve.

    Moreover, since they are far less complex and much less expensive than hospital-type ventilators, production could more quickly and more easily be ramped up.

    So it is at least possible that CPAP machines might be used as a temporary stopgap for those whose ventilator needs are not as severe and as complicated, and certainly such use would be better for patients with respiratory problems who otherwise might not be provided with any breathing assistance at all if the need at a specific hospital exceeds its current supply of ventilators so that some patients may be left to die.

    Although some have called CPAP machines “poor man’s ventilators,” perhaps they should more properly be termed life-saving stop-gap ventilators during the current crisis.

    To deal with this life-threatening shortage of hospital ventilators, CNN’s Dr. Sanjay Gupta suggested bringing old ventilators out of storage and trying to make them operable again, but this may not be sufficient if demands for life-saving breathing assistance continues to dramatically expand.

    Also, trying to suddenly expand manufacturing capacity for these hospital machines – e.g., by creating various incentives – also may not work, because they are both difficult and very expensive to construct.

    So, why not at least consider the possibility of using existing CPAP (Continuous Positive Airway Pressure) machines, now already in very widespread use in millions of homes to combat sleep apnea, as devices to help persons with respiratory problems due to the coronavirus, especially in less serious cases which do not require the full power and sophistication of expensive hospital-type ventilator machines.

    Here’s why:

    First, the number of existing CPAP machines greatly exceeds – by an order of magnitude – the number of hospital ventilation machines, and, because they are simpler and less expensive, manufacturing capability can be increased far more quickly, very easily, and at much lower cost.

    Second, CPAP machines, especially those with full-face masks, can provide a very significant increase in the amount of air (and therefore oxygen) a user can consume, and many can easily be adjusted to provide even higher air pressures than would ordinarily be required to overcome mild sleep apnea.

    Third, persons most at risk of respiratory problems from the coronavirus virus – i.e., those who are elderly and/or have other medical problems – are also the population most likely to already have and use CPAP machines. So they and their family members are therefore familiar with and comfortable with them.

    Therefore, in the event of an emergency shortage, these existing users could be asked to bring their units to the hospital where they might be used in all but the most serious respiratory distress situations requiring a sophisticated hospital ventilator machine and trained ventilation technicians.

    Fourth, many CPAP units are used largely to limit snoring, and in some cases users might be able to do without them in an emergency without risk of serious health problems such as cardiovascular events. Thus, some current CPAP users might be induced to permit their units to be used by hospitals or by other centers treating coronavirus patients if there is a severe shortage of hospital breathing machines.

    The threat posed by the coronavirus is so serious and unusual that all possible approaches – including simple ones such as using CPAP machines – should at least be considered and evaluated, if not actively tested in practice.

  7. There are some fascinating Randy-Quaid-against-the-aliens crowdsourcing ventilator engineering projects springing up around the Internet, and if movies have taught us anything, it seems that it would be wise to have a policy that favors both Will Smith (institutions) and Randy Quaid (initiative) models.

    The difficulty with using an ex post funding mechanism for something like this, though, is that (assuming no one is being paid on contingency) an intermediary market has an incentive to (1) assemble as many claims / development expenditures as possible, and (2) have as many “rich” claims as possible in their portfolio to be assured of a substantial recoupment on selection. Bypassing, for the moment, the problem of how you get funding to buy up exclusive assignments of claims prior to having sufficient portfolio size to have a reasonable shot at selection, which would seem to favor a very small and institutional intermediate market, the logic of the broad portfolio seems to preclude the moon-shot that’s going to have to go up to the alien spacecraft. Intermediary funding would seem to favor a very large portfolio of ferry businesses without actual boats, but with the most convincing appearance of boats that’s humanly possible. Or, more specifically a broad base of manufacturers each of whom has an incentive to run up their own proper costs rather than collaborate, resulting in duplication of R&D and manufacturing setup outlay across the sector. Once someone’s in the portfolio, the intermediary would seem to profit with every increased expense of each company.

    (Of course, this sort of funding is different from fixed injury claims, where the model might conceivably have a very different effect.)

    Or perhaps I’m missing something. Or everything.

    –Mr. Dove

  8. “I received excellent comments and will return in a later post to answer some of the questions and objections raised by commentators. ”

    It’s commenters, not commentators.

    From The Grammarist:

    “A commenter is someone who makes isolated comments. These days, the word most often refers to people who post comments on blogs and news websites. A commentator is someone who provides commentary. The term usually applies to professionals in sports broadcasting or television news. Commentators don’t just make one comment; commenting is what they do.”

  9. The shortage will be of respiratory therapists, and ventilators to a lesser degree. The CDC and some states have stockpiles of ventilators, ‘though likely not enough for this.

    I’m a bit puzzled at the numbers, too. At one point you mention $15B to $60B. A ventilator is about $25,000. $15B would buy 600,000 of them. Surely that’s way too many.

    I have some experience with product development and manufacturing. While there is certainly spare capacity among manufacturers, none are likely to significantly increase manufacturing capacity outside of altruism. This is because there are capital costs to do so, including plant, tooling (for molded parts, for example,) personnel, distribution; set-up and training. And once this crisis wanes, they will be stuck with this capacity which will either be retired (prematurely) or serve to lower the market price of ventilators, potentially wrecking the business. In addition, with quarantine conditions in the US now, it will be nearly impossible to increase the workforce for this. I am afraid that the ship of ventilator supply increase has sailed. Our best bet is to try to train existing health care professionals in less critically needed specialties as “associate” respiratory therapists, to work under the direction of respiratory therapists.

    One commenter wrote an excellent post on repurposing of CPAP machines, which I had thought of myself, and which idea I support. But I confess I’m coming at it from a position of ignorance and have no idea if it would really work.

  10. There are two issues I see here. First, that running the lottery per claim creates a bias toward simpler cheaper claims. You should deal with this by selecting claims proportional to their claimed dollar expenditure (with some substantial penalty for dishonesty).

    Also it is very unlikely such a process could produce approved or cleared machines in time. Not only does the fairly precise specs limit flexibility, approval limit speed and likely prevent choice of optimal functionality/numbers.

    As such it seems the critical part here isn’t rewarding good solutions but replacing the usual authorization and regulatory process with a promise to punish ec-post if someone’s cheap ass ventilator causes more harm than good. This seems like a separate proposal with its own problems.

  11. Ohh one more concern. Suppose the maximum number of ventilators used at one time is 300k and the program brings total ventilator capacity from 100k to 330k. What is the ex-post contribution to QALYs if someone who delivered a ventilator?

    It seems the natural answer is 0. If it’s not 0 you encourage overproduction. But if it is 0 you have a huge problem as if the market gets the number of ventilators needed slightly wrong suddenly no one gets anything. If you make it proportionate to fraction you contributed of needed number you encourage people to over produce.

    Ok so maybe you do it on first come first serve or whether or not they are actually used. But this then penalized the smart manufacturers who realize they can build better units if they take a bit longer (but still deliver before peak). Or if based on usage we get overproduction and a bunch of wasteful add spending trying to convince health workers to use these machines.

    In general there is just no way to cleanly and fairly divy

Please to post comments