Why Every Existing Approach to Covid-19 Reopening Is Wrong
Top-down approaches are doomed to costly failure
As a resident of the San Francisco Bay Area, I’ve been under mandated lockdown for 48 days--longer than nearly any other community in America. So I’ve been following efforts to reopen the country during Covid-19 with intense interest. And I’ve concluded that nearly every approach so far is wrong.
These approaches are wrong because they ignore the fundamental interconnections and network dynamics underlying modern societies. Here’s a real network map (graph) of a modern community; in this case the community of Canadian professional cyclists. Each node represents an individual. The lines represent connections between individuals.
Here is another, denser map; in this case of climate change professionals in the Marshall Islands.
Note that in both cases, the maps contains clusters of nodes with dense interconnections. These are groups of people who are all connected to each other--in daily life (rather than these specific cases), think of a family, a knitting circle, or a group of friends who lift weights together every day before work.
There are also nodes that are loosely connected to multiple networks (high betweenness centrality). These are the individuals who travel within multiple social groups (such as a person who loves to attend the symphony on Friday night with a friend, gets together with their roller derby team on Saturday morning, and goes to a bible study on Sunday). These individuals serve to connect multiple dense networks to each other.
That’s why viruses, in an open society, spread nonlinearally (often exponentially). Once they enter a dense network, they infect all nodes rapidly. Then, they infect a connector, jump to another dense network, and infect all nodes in that network, too. And so on.
It’s also why R0 is a useless metric in a pandemic like Covid-19. It reflects the average number of people one infected person will go on to infect. But being an average, it ignores the exceptionally well connected individuals that bind networks together (either through social connections, extensive travel, or both), and events which bring a large number of susceptible individuals to one location.
These “superspreaders" (people and gatherings) can have a personal/group R0 that’s an order of magnitude (or more) higher than the average. All it takes is a few individuals or events like this to get a virus spreading rapidly around a society.
It’s also why top-down reopening will always fail. Top-down approaches expose all of a society’s networks to the same risk at the same time. They make the assumption that all individuals are the same. As we see with superspreaders, that is patently not true.
They also assume that all networks are the same. This is also not true. There is a big difference between a network of vulnerable individuals in close physical proximity (a nursing home) and a group of lower-risk individuals who are physically more isolated (young parents of kids in preschool, living in separate homes). Exposing all these networks to the same risk with a top-down reopening (even one targeted at specific industry sectors) is a recipe for disaster, and nonlinear spread.
What I propose instead is a bottom-up approach to reopening societies. This begins at the level of tiny, dense family networks and works upwards.
Lockdowns have proven so successful (when done properly) because they prune back social ties, almost (but not quite) to zero. With a strict lockdown, a network map like the above ends up looking something like this:
Those tiny groups are individual families (of 1 to 6 people, let’s suppose). They remain densely connected, but are isolated from the rest of society.
Note that there are no connector nodes. If one dense network becomes infected from the environment (say, by a member visiting their grocer or podiatrist), everyone in the network will become infected. We’ve seen this with actual families during Covid-19, where one infected family member rapidly infects the rest of the family.
But because of the lack of connectors and the small size of these dense networks, the virus doesn’t spread to the broader society. New infections are kept at acceptable levels, or even reduced to zero. Lockdowns work. But they’re economically and socially unsustainable.
Reopening with a bottom-up approach would simply involve taking these isolated, dense networks of family members (which have proven reasonably safe, even with a level of contamination in the environment) and gradually expanding them.
The expansion could go as follows (in ascending order of dense network size):
- Nuclear families; small groups in lockdown together (current situation)
- Extended families
- Small community groups; small groups of friends; synagogue havourot; walking groups
- Small businesses (<10 individuals); small functional teams within larger companies (Agile software development teams, groups of accountants)
- Larger community groups; larger religious groups; full offices of <50 individuals
- Full towns or local communities; full companies, if geographically concentrated
At each stage of reopening, the size of densely connected networks would be gradually expanded. Crucially, testing for Covid-19 would need to be aggressive and continuous. A great degree of trust (or government control) would need to exist at each step between all network members to allow a dense network to expand.
Here’s how this would work in practice. We would begin with the isolated, dense family networks that exist in (intelligent) locked-down societies right now, like the ones in the map above. These groups would choose a small number of additional densely-connected networks with which to merge their network.
For example, a nuclear family of four isolating together could choose to re-engage with several of their extended family members or single friends, expanding the size of their dense network to 10-20 individuals.
On day 0, all members of the prospective expanded network would be tested, before connecting. This would set a baseline of zero infection within the network — including asymptomatic carriers. The members of the network would then begin to associate with each other freely.
Close, frequent ties within the network would be encouraged. The goal would be to derive as much social and economic benefit from the expanded network as possible, since all members would be assuming the risk of densely connecting themselves with the others, and would want the maximum benefit from that risk.
Crucially, all members would continue to maintain lockdown-level isolation from those outside the dense network. This would include:
- Engaging physically with society only for essentials
- Universal use of masks when not engaging with members of the network
- Hand sanitizer, hand hygiene, physical barriers
- No or limited travel, or strict 14 day quarantine from the group following travel
The members of the network would need to trust every other member to adhere to these guidelines strictly. That’s why it makes sense to begin with networks which are already accustomed to dense interconnection (families, small communities) rather than functional groups (one doctor, one hairdresser, one driver, and so on).
Each node in the expanded network would face a nonzero risk of being infected by the environment. It is assumed that absent intervention, they would then spread the infection to all other members of the dense network. So in a default state, the network’s risk of environmental infection would rise linearly with the number of nodes--a problem.
To manage this risk, all members of the network would be tested frequently--ideally weekly, or more often. The goal would be to detect an environmental infection of one member and isolate the infected individual (for 14 days) before they were able to infect the entire dense network. Frequent testing is the crucial piece that would keep dense networks safer as their node counts grew.
Networks are (probably, more below) fractally self similar and scale-free. This means that if a bottom-up approach worked for the move between nuclear families and extended families, the same approach would likely ultimately work for the move between regions and states, states and countries, and so on (assuming ungodly amounts of testing, on which more below).
At a societal level, measures would need to be taken to avoid creating superspreaders as dense networks expanded. These would likely involve:
- Continued cessation of most travel; entry restrictions
- Elimination of high-risk events (concerts, sports with audiences, conferences)
- Mandatory physical distancing measures in all public spaces
This would keep each expanding dense network as isolated from other expanding dense networks as possible, reducing the potential for the virus to jump between dense networks and restart nonlinear spread.
The most crucial part in all this is rapid, cheap, continuous testing. Ongoing, continuous testing and isolation of infected individuals is the only way to stop an infection within a dense network. It’s also the only way to manage the risk of asymptomatic carriers.
Without it, again, the risk to a dense network scales linearly with its size. As networks reach a certain size, the risk becomes unacceptable. Only testing — continuous and aggressive — can manage this.
Some countries claim to have the virus under control. Don’t believe them. Island nations have it comparatively easy. They can physically close their borders, trace infections internally, and enforce quarantines to keep the virus in check. In effect, they are one large (but not giant) dense network.
Ditto for authoritarian states (who can completely cease travel and effectively turn themselves into islands). Or states so small and isolated already that they don’t have much interconnection to begin with. None of these are America — none have our mix of size, diversity, and personal liberties. None of them are instructive when planning for our own re-opening.
Further, none of their success will last, at least in their present form. Absent enough testing to continually monitor individuals across a dense network (whether that’s a church group or an island nation), risk scales linearly with n.
As soon as the virus is reintroduced into these closed societies (through reopening trade to foreign nations, etc.), it will resume its spread, unless these societies have enough testing capacity for continuous monitoring of every node in their dense networks — which they do not, yet. We saw this with Singapore — the virus initially seemed under control, but a missed well of infection (in migrant communities) rapidly restarted nonlinear spread.
The idea of a bottom-up approach (relying on massive testing as networks scale) to reopening has several advantages. For one, it allows the immediate resumption of some of the most sustaining social and economic ties--those of family and close friends — not the ties one has to their golf course attendant or bowling alley manager.
In effect, it creates a new tribalism which is probably compatible with our innate desires, Dunbar’s number, and the like.
It also allows the reintegration of the most at-risk individuals into a network, at almost no cost to the network. A family could choose, for example, to reintegrate an elderly, institutionalized parent into their dense network. This would raise the network’s risk negligibly (since these individuals rarely venture out and thus have low risk of environmental infection), while greatly reducing the risk to the vulnerable person.
A bottom up approach could also be enacted without the cooperation of government. Even in a place which has chosen to rapidly (recklessly) reopen top-down, individual families could choose to remain in isolation and slowly adopt a bottom-up approach to re-engaging with the broader society. It is comparatively easy for governments to order you into your home--it’s much more difficult to order you back out again.
This approach also does not rely on a vaccine, or the assumption of any degree of immunity. If a vaccine is discovered, or immunity is found to exist, then terrific; dense networks could grow much more rapidly, as certain immune nodes could be added to the network with zero additional risk. If a vaccine is never discovered or immunity doesn’t happen, a bottom-up approach can still allow societies to safely reopen.
My strategy suggests a strategy for prioritizing vaccination, too, if a vaccine is developed. Individuals who connect multiple dense networks should be vaccinated first. They could then resume their crucial work of binding different dense networks together socially and economically, without additional risk to the networks.
The bottom-up strategy also has several disadvantages. It risks exacerbating the class-based fractures that already exist in American society. The rich would likely expand their dense networks to incorporate the rich, further accelerating the uneven distribution of societal wealth.
The bottom-up approach also relies on a great deal of trust between members of an expanding dense network. Each network member must trust all other members to continue ongoing testing, avoid close ties outside the network, observe proper precautions when interacting with the environment, etc. — in short, to obey the social norms around avoidance of infection that the group collectively establishes.
This may limit the size of dense networks to the number of individuals that one can trust completely--in short, to small family and community “tribes", and not whole societies. Government could assist here, for example by creating frameworks for verifying ongoing testing, much as they verify vaccination status of school-age children today.
The assumption implicit in all this is that large sets of separated, dense networks are different from the sum of their parts. Ten separated dense networks of ten individuals are not the same as a single dense network of 100 individuals — primarily because of the lack of connection between networks and thus the reduction of superspreaders.
As dense networks scale beyond a certain size, though, this dynamic may no longer hold true. A set of 10 networks of 100 people might behave much like a single network of 1,000 people (with similar risk dynamics), because a network of 100 people might be large enough to begin to have its own connector nodes and mini-superspreaders. For this reason, there may be other upper limits to how much dense networks can safely scale.
Another implicit assumption is that casual contact and close contact are different. If this proves not to be the case--and one’s risk of contracting the virus from one’s grocer and one’s brother are the same--then anyone with whom an individual interacts would be in their dense network. This would necessitate severely limiting the number of interactions each individual has--even casual ones.
Options for full immunity — such as vaccines or the discovery of meaningful immunity after infection — would help tremendously in growing dense networks between small, trusted groups. Having risk-free individuals to add to a network greatly increases its potential scale, even without trust among all members/nodes.
So, will I be starting to expand my family’s own dense network, beyond a few trusted individuals?
No, for one simple reason: testing is not there yet.
For a bottom-up approach to work, testing needs to be truly ubiquitous, accurate, cheap and easy. We need the equivalent of a diabetic’s blood glucose meter, but for Covid-19; something which can test continuously and cheaply on a weekly, daily, or intra-day basis, at home, for minimal cost.
Until that exists, the risk of scaling up dense networks is too great. Scaling beyond 10 or so individuals makes the risk that infection will spread from the environment to all members of the network unacceptably high.
To allow for bottom-up reopening in the near-term, leaders and governments should do the following:
- Immediately cancel all re-opening measures that re-establish dense connections in a top-down manner
- Aggressively expand testing, opening it to all individuals, regardless of risk profile or symptoms, at little to no cost, and with individuals encouraged to test weekly or more often; expand options for testing from home, or instant testing
- Continue to explore means to create risk-free nodes: vaccines, confirmation of the existence of immunity after infection and ways to measure this, etc.
- Maintain existing lockdown measures until testing is widely enough available to allow dense networks of 10–20 individuals with at least weekly testing
- Maintain physical distancing, mask use and other regulations until enough testing capacity is available (or vaccines exist) to allow safe, sustainable dense networks as the level of the entire society
- Put in place measures for mediated trust as networks expand beyond tribes, as as (optional) registries of testing and Covid-19 status
- Lay the legislative groundwork for bottom-up reopening, including allowing small groups to begin creating dense connections and interacting when testing capacity is sufficient to allow this to happen safely
With proper expansion of testing and good legislation, we can begin this in weeks or months, rather than the years or more it will take to develop vaccines. Through a bottom-up approach, properly implemented, we can begin to re-open societies in a truly safe and sustainable way.
I am indebted to Chris C for acting as a sounding board as I developed this. Beyond the specific citations above, I am indebted to the overall work of Nassim Taleb, from which much of the conceptual basis of the risk analysis and nonlinear dynamics here flows.