Learning From Engineers

By Braden R. Allenby, Mikhail Chester

Like many professors, we have been trying to explain (online, of course) to our undergraduate engineering students the United States’ botched response to the COVID-19 pandemic. And they rightfully raise the question, So what would you have done? In the course of this dialogue, we have realized two things. One, COVID-19 may be a new pandemic, but it’s also a variation of a basic engineering challenge that extends across all infrastructure. Two, experience drawn from existing engineering practices and institutions may help the nation do much better than it has so far.

There are two important underlying principles involved. The first is that with all types of infrastructure, including the medical infrastructure system, efficiency and resilience are in conflict. The most efficient system is one that is carefully optimized to use the fewest possible resources to meet current conditions. So long as nothing important changes in its internal or external states, and the variability of environmental conditions stays within a narrow envelope, such a system will most efficiently meet the needs of its users. But when the situation does change—and especially if the change is anomalous, high-impact, and rapid, allowing little time for adaptation—such a system will be very fragile, since the conditions to which it has been adapted no longer prevail. As an obvious corollary, competition in a relatively stable environment will drive firms, institutions, and infrastructure toward efficiency, with an emphasis on rigid processes and assets that change slowly, and away from resilience. For example, RCA was very efficient at making vacuum tubes, and Kodak was very efficient at making film and traditional cameras; both firms failed when technology changed more rapidly than they could adapt.

The second principle is that infrastructure systems are designed for what electrical engineers call “peak load.” Urban transportation systems aren’t scaled for early Sunday morning traffic; they’re built for rush hours. Electricity infrastructure isn’t scaled for middle-of-the-night troughs in demand, but for peak daytime usage in the service area. These peaks are fairly predictable and occur within an understood range. When load on the infrastructure exceeds those predicted peaks, it can lead to failure: standstill traffic jams or rolling blackouts. Medical infrastructure, including hospitals, intensive care units, and ventilators, is analogous: it’s scaled for anticipated peaks in the service area (say, for flu season), but not for a pandemic that sickens thousands of people in the same place at the same time. Because of its complex institutional, functional, regulatory, and mixed private-public nature, medical infrastructure scale and peaks are difficult to perceive clearly in practice, even if the principle holds.

The corollary to the second principle is clear: system managers can save resources by reducing peak usage, regardless of the infrastructure involved. This is an old principle, often called “demand side management,” and has been a part of electric utility regulation since the 1970s. How this is done will depend on many things: on price structures, which might feature higher peak and lower off-peak rates, or perhaps on less mechanistic practices. With traffic, for example, demand for roads at peak times was cushioned by morning and evening traffic reports, which helped people choose less congested routes to work, and then by tools such as Google Maps and Waze, which guide drivers to the least congested routes.

These two principles open up the possibility that policy responses to COVID-19 might be improved by a closer examination of experience with other types of infrastructure, and by different ways of conceptualizing future investment in what might be called “efficient resilience” when it comes to medical infrastructure. Efficient resilience is related to the concept of “graceful extensibility,” or the ability of a system to adapt to handle surprise; it’s the opposite of brittleness.

COVID-19, of course, won’t be exactly like existing infrastructure analogs. For example, “flattening the curve”—or what an engineer would call “peak shaving”—in the case of COVID-19 is not being done only to try to match the amount of available medical infrastructure with demand but also to buy time for the development of more effective treatments and eventually a vaccine. That doesn’t happen with, say, communication infrastructure. But the mechanisms available to manage demand that are now somewhat routine in infrastructure engineering could prove helpful during this crisis.

An example is provided by AT&T’s long-standing Network Disaster Recovery program, which is activated whenever there is a local challenge—a hurricane, an earthquake, or the terrorist attack of September 11, 2001. When one of us (Allenby) was with the firm, it kept large semi-trailers loaded with fully operational electronic switches in central staging areas, to be driven to incident sites as soon as required. Today, the company’s response equipment includes mobile cell sites and mobile command centers, such as Cell on Wheels (COWs), Cell on Light Trucks (COLTs), and Flying Cell on Wings (known, of course, as Flying COWs), along with mobile base camps and the requisite hazmat equipment and supplies. Maintaining these assets everywhere there might be a hurricane or storm, earthquake, terrorist attack, or flood would be far too expensive for a competitive firm; but maintaining a mobile capability that can be shifted to meet unpredicted peak demands or unanticipated system challenges is a commercially viable balance of efficiency and resilience.

To date, the US policy response to COVID-19 has been almost exactly the opposite. States and urban centers are competing with each other, and with foreign countries, to get the medical resources required to meet their projected or actual peaks. Rather than take advantage of existing elements of an integrated federal system (such as the scientific competence of the US Centers for Disease Control and Prevention, the federal supply of medical resources in the Strategic National Stockpile, and the expertise of the largely disbanded National Security Council pandemic response team) to ensure efficient allocation of response resources as local COVID-19 infections peak, every entity has been, by deliberate policy and incompetence, left on its own. There has been no integrated policy or response at the federal level. Policy-makers are trying to shift hockey helmet companies to making face shields and craft breweries to producing hand sanitizer on an ad hoc basis, which, even if it does eventually succeed in providing some resilience, does so only at enormous economic cost and frequently too late to actually be effective in meeting peaks. This is an instance of “inefficient resilience,” and the fact that it is essentially the only way the United States is responding so far is a damning indictment of American governance.

It’s not too late to move toward a more efficient and resilient medical infrastructure in response to COVID-19. In fact, it is both timely and critical, because although the initial response has been terribly wasteful in terms of infrastructure—and consequently in human health and people’s lives—virtually all epidemiological experts expect a second wave of infection later this year, which, depending on whether and to what extent initial infection confers immunity, could be equally difficult.

To understand how to build efficient resilience into the country’s medical infrastructure, the first step is to examine what should have been done before the current crisis, and in so doing draw on historical experience across engineered systems of many kinds. It’s crucial to note that the United States does not face one large peak of COVID-19 incidence, but a multitude of local peaks, and that medical infrastructure, as with most complex infrastructure, consists of a number of components, some of which are mobile, and some of which are not. If the United States had approached COVID-19 as an engineering problem at scale—that is, acknowledging that it has a limited and complex medical infrastructure system, and that the primary tasks are to a) shave peak load, and b) create efficient resilience across the entire system—the nation wouldn’t have had such a confused and inefficient response.

For example, in addition to beefing up the federal stockpiles, which are necessarily static, national policy-makers could have implemented an organized system to create a mobile reserve supply of ventilators, and perhaps field hospitals, that could flow rationally around the country as peak demand shifts. Mobile intensive care units are common at the local level to complement first responders in serious accidents, and at least some ICU capacity could also be routed around the country as local conditions warrant. In engineering terms, this would mean not just managing the stock of medical infrastructure rationally, but having a process that would manage the flow of medical infrastructure rationally in near real time.

Such response capabilities require coordination, management, and logistics competence that encompasses the entire medical infrastructure system. That’s why AT&T’s Network Disaster Recovery program works; it is managed by a single entity. For COVID-19, efficient resilience—and minimizing unnecessary sickness and death—implies that the system be national in scale (it should also be international, to expand the system to be optimized, but let’s walk before we run). So, for example, if the federal government had put an experienced senior military officer with logistics experience in charge of a dynamic national medical infrastructure system in place in January, the United States would not have nearly the chaos and unnecessary loss of life that it is now looking at. Had that been done, a lot of unnecessary damage to the economy and the public could have been avoided; even if it were done at this late date, much unnecessary damage could be reduced.

The American medical infrastructure is dauntingly complex: it includes thousands of public and private hospitals and clinics; more than a million doctors, nearly four million nurses, and uncounted administrative professionals; and many thousands of manufacturers of pharmaceuticals, medical equipment, and devices. Policy-makers have attempted to build in some resilience, as with the Strategic National Stockpile, which in this crisis has been badly mismanaged. Given this reality, any attempt at long-term central control is doomed to failure. But that doesn’t mean that dynamic resource allocations can’t be developed that do a far better job than current systems at balancing efficiency and resilience. Even physical infrastructure systems involving expensive and long-lasting equipment can be made more agile and adaptive, but it requires:

• Modularity. Design the system to be modular, with certain modules adapted for mobile inventory. For example, conceive of different elements of medical infrastructure—general purpose pharmaceuticals, ventilators, personal protective equipment, field hospitals, and so on—as modules of medical infrastructure to be managed both in themselves and as part of the overall infrastructure.
• Intelligence. Understand the system involved, including having good data on peaks and assets. Even with high levels of uncertainty, the United States could have done this with COVID-19, but hasn’t so far, thanks to botching the testing function.
• Coordination. Be aware of the status of critical infrastructure resources across the system as a whole. The daily pageant of different jurisdictions chasing different assets, from masks to ventilators to skilled personnel, is not only testament to incompetent management but also highly confusing to an already quite reasonably frightened public.
• Clear objectives. Manage the system with clear goals. For instance, in AT&T’s case, the goal was to keep phone service working at all costs; with COVID-19, it is to minimize health and economic damage.
• Financial support. Pay for resilience measures; even mobile assets aren’t in use all the time, so the efficiency versus resilience tension doesn’t go away. Building in efficient resilience is more expensive than not, but the country is discovering that it’s far cheaper than the economic consequences of a broken system.

Efficient resilience for medical infrastructure is not just a matter of physical assets, but of organizational capability. Moreover, it is not responsive only to pandemics, but would be useful for other extreme phenomena that pose peak demands as well. Not all such events place the same demands on medical infrastructure: COVID-19 is causing a lot of local peaks, a major Midwest flood may cause a number of regional peaks, and a hurricane may be a single but massive peak. But the principles remain the same: reduce or spread out peak demand to match available medical infrastructure resources; size the national medical infrastructure to be both efficient and resilient, to the extent allowed by politics and culture; centralize leadership and focus on clear objectives; and build dynamic response systems that enable efficient resilience. Engineers know how to do it. And we know that not doing it is costing lives, health, and trillions of dollars.