Tech guy rants about epidemic control

April 22, 2020

Tech guy here, not a real doctor, officially out of my area. That said, I wish to rant.

For this pandemic, there are a lot of things we can do to control it, and a common criticism is “but X won’t handle Y” (and this is true), and the criticism is posed as if it is a deal-breaker, we should just abandon X, it won’t work. This is wrong. Pile on enough imperfect measures, and you can shut down the epidemic, and as long as those half measures continue to be employed we might be able to “open up” many parts of the economy once the infection rate is low enough. One thing that motivates this criticism is a misplaced focus on what things like masks are for — the mask on your face is not to protect you, that mask is intended to protect other people from you. If the only thing protecting you is the single mask on your face, then it has to be supremely good, and nope, a cloth mask does not meet that standard, those good masks are relatively scarce and doctors and nurses need them because they are exposed to nearly 100% actual sick people. If, on the other hand, you are protected by the dozens to thousands of masks on the faces of all the people you might encounter in a given day, those masks don’t have to be nearly so good, just good enough (in combination with other measures) to stop the epidemic. 

So, first, a discussion about how an infection apparently progresses and spreads. To start, you get some virus in your mouth or eye or up your nose or whatever. A few days go by, and you become infectious. A few more days go by, and you may start to feel generic symptoms, maybe a fever, and are infectious. A few more days go by, and you may start to feel terrible symptoms, check with a doctor, whatever, you’re really darn sick, and infectious. Some people never show symptoms, and they may or may not be as infectious as the sick people, but we’ll assume they are. We’ll also assume that the number of people who don’t show symptoms is “about half”, meaning anywhere from 25% to 75% (this, from randomized sampling in Iceland, and from cruise ships full of old-ish people where everyone was tested and from naval ships full of healthy young-ish people where everyone was tested). There’s guaranteed to be a few days when you’re infectious but don’t feel sick, or think it might just be allergies, and it’s possible that you might be infectious for more than a week without ever feeling sick at all.

The fact that so many infected people are asymptomatic is not good news; it means that it’s much harder to rely on people feeling sick and staying home. It also means that nobody should assume that they are uninfected, which can be hard for some people to understand.

So, now, consider “half measures”:

  • Cloth masks. Cloth masks are far from perfect at stopping inbound, atomized virus (for example, measles). Atomized virus particles are tiny, and require an N95 mask to get someone to the point where they might feel “safe”. Depending on the cloth, cloth masks may not be that good at stopping even bacteria. Or they may be worn too loose to seal properly, or worn in a way that doesn’t cover the nose. But: cloth masks are pretty good at stopping you from projecting infection; if you happen to spray it instead of saying it, it stays in the mask. If you cough or talk loudly, the airflow is thwarted, the larger drops are captured, the medium-sized drops fall to the ground closer to you, the smallest ones are at least not propelled away from you in a puff of air. Perfect? No. But better. Suppose it cuts the rate of outbound infections by half per mask-wearer, and 2/3 of the people wear masks well enough to do this. That means the revised infection rate R’ is R0 * 2/3 * 1/2 + R0 * 1/3 * 1 = 2/3 R0. If R0 was 3, R’ is 2. The virus is still spreading, but not as fast, despite half-effective masks only mostly-worn correctly.
  • Contact tracing.
    The old way: Suppose, crudely, 1/2 the people with the virus have symptoms, and when this causes them to seek medical help, they are asked about their contacts. Suppose that they remember 1/2 of their contacts, who are then re-contacted, and quarantined. I.e., 1/4 of all cases, that would ultimately be both symptomatic and asymptomatic, are now intercepted and quarantined before they can spread the virus. I.e, R’ = 3/4 R0.
    With an app: Assume half the population uses the Google/Apple tracing app (this is actually very optimistic). Now, symptomatic cases, half (1/4 of total) do it the old way and half (1/4 of total) use the app. Naively, the app will yield perfect results for the half of the contacts that also use the app, and the other half have to be done the old way. It’s messy arithmetic, but R’ = 11/16 R0. We’d want more, but this is “free” — one reason to use the app is that it is far less costly, and quick. If we were lucky enough to get 70% of the population using the app, the math is even messier, but the factor is now:
        0.5 [asymptomatic, never traced] + 0.15 * 0.5 [no app, missed half of contacts] + 0.35 * 0.3 [contacts not using app] * 0.5 [half found through old-way tracing]
    or about 5/8. Suppose we couldn’t afford to do detailed person-by-person contact tracing — if 70% of the population used the contact-tracing app and we only relied on the app, the factor is still slightly below 11/16. What this means is that the app can still help when the number of infections is more than can be managed with person-by-person tracing. But if you don’t have at least half the population using the app, it isn’t helping that much.
  • Partial work-from-home. Not everyone can work from home, but many people can. If 1/8 of the workforce that would normally show up at work can stay home, that reduces the infection rate; everything is less crowded. Crudely, if 1/8 stays home, the infection rate is reduced by a factor of 7/8.
  • Partial testing. Testing every person every week or every two weeks is an enormous undertaking, and it doesn’t look like that’s happening soon in the US (I’d love to be wrong). Some jobs require lots of interaction with other people, which creates a high risk of infection for the person working that job and a risk of spreading that infection widely. Assume for the moment that this is the source of 1/4 of infections, mostly on the outbound side. To reduce the risk of infections, test these people frequently (for example, every Monday and every Thursday), and if they show positive, they are quarantined to prevent further spread.

Nothing here is perfect, but enough layers attacking infection spread from different directions can win. Take all these factors and multiply them: 2/3 * 11/16 * 7/8 * 0.75 = 0.3. Together, these are adequate to stop an epidemic that has a reproduction rate of 3, though it would take months (3 * 0.3 = 0.9 which is less than 1, but it would take 0.9 to the 11th power to get reduce the infection rate by 90%). Any improvement — more mask wearing, better mask wearing, wider use of a tracing app, more work from home, more widespread testing — means it can shut down the same epidemic more quickly, or stop a bug with more spread. Small changes matter a lot; if the reduced R0 is 0.8 instead of 0.9, an epidemic is stopped twice as fast (0.9 * 0.9 = 0.81).

Because small changes matter enormously, it’s helpful to examine why people might not do what they need to do, and think of ways to improve. Realistically, some people won’t stay quarantined. Some people might not quarantine immediately, which reduces its effectiveness. Suppose, you’re on your way to pick up your kid from day-care (will we have day care while we’re doing this? Pretend we will) and your phone buzzes and says “you were in close contact with someone who had COVID. Please quarantine immediately, then call this number.” So, do you pick up your kid? If you don’t, who does? If you arrange for someone else to do it, who pays the late pickup penalty? If the “quarantine now” demand comes with some assistance for unwinding your immediate obligations, you’ll be much more likely to do it. If you’re quarantined, what if you don’t have two weeks of food in your pantry? (Or, you don’t have a pantry.) Who does the shopping? This would be a good thing for a public health department to backstop (in some countries, they do just that). And what if someone has a job that cannot be done from home — do they lose their job if they stay home? Who covers their costs for the two weeks they’re not working? If there aren’t good answers to these questions, more people will either cheat on their quarantine, or try to avoid contact with public health officials altogether. That’s not how we stop an epidemic.

Or consider masks. Depending on how they are made and used, they filter inbound and outbound infections with varying efficiency. Workshop N-95 masks do a great job of protecting the wearer, but usually have a valve that allows them to bypass the filter when they exhale, so they hardly stop outbound infections at all. Surgical masks (not medical N-95 masks, which are another step up in quality) are quite good in both directions. On the other hand, surgical and N-95 masks don’t tolerate washing, so they’re harder to reuse. Cloth masks appear to be better at reducing outbound infection than they are at reducing inbound infection, but they can be washed.  If people can get good masks, they will, but if they have to make their own, the materials and fit will be somewhat random. For all kinds of masks the fit matters; if it’s not properly seated to your face, or if the nose piece isn’t snug around your nose, or if someone has the wrong arrangement of facial hair, they can leak around the edges pretty easily, and both transmit and receive infection.

One problem with masks is that many places actually have laws against wearing masks in public, and if people (in particular, not-white people) believe that mask-wearing will create a pretext for harassing them, they might not wear masks. Or, instead, if mask-wearing is the law, how is it enforced? If the first step in enforcement is arrest instead of handing out free masks, that’s not necessarily helpful. If the arrested person is actually infected, the steps to arrest them are quite likely to spread the infection, whereas handing them a mask reduces the spread right then. The traditional enforcement approach does not directly fix the problem, it instead (at outsized expense) “creates incentives” for other people to fix the problem, and just assumes that they have the means and ability to comply. What if they don’t?

Contact-tracing apps don’t work if they’re not widely used, and if they are opt-in they won’t be widely used, especially if people have fears about being tracked or spied on. It’s possible to design a protocol that is resistant to spying and hacking, but how many people will believe that claim? One “solution” to this is to make the app opt-out, or even mandatory; this will cause all sorts of loud objections from a tiny number of well-meaning people who care a lot about privacy, but in practice mandatory or opt-out will save lives.  The traditional contact-tracing process is not exactly privacy-preserving, either. On-the-other hand, a larger fraction of asymptomatic infections will reduce the effectiveness of contact tracing in general, because without frequent and widespread testing, nobody will ever check the contacts of asymptomatic infected people.

Another attack on the transmission problem is to try to remove some of the high-risk infection points altogether. Amazon has experimented with checkout-less shopping; the cart knows what you put in it and know who you are, and you are billed as you walk out the door. There is no cashier, either to be infected, or to spread infection. Bus drivers are at high risk; do they actually need to collect fares? Could collection be automated? Do there need to be fares? (No fares might result in crowding; perhaps we declare that a “full” bus need not pick up new passengers.) Jails and prisons are terrible for transmitting this infection; one way to fix this is to only jail people who really, truly need to be jailed — so for example, unless someone is arrested for a violent crime, maybe they don’t spend time in jail before trial, period. I.e., change it so the default is not to arrest — take the entire process, and make it as if you were arrested, had a bail hearing, posted bail, and were released. The current process is actually bullshit, because it pretends to merely be necessary process and not punishment all by itself, when obviously it is also punishment administered by police with plenty of bias and not much oversight. I’m sure there’s tons more of these, where “we’ve always done it this way” or “we have a process” get in the way of less infection and fewer dead people (I mean, seriously, what about traffic stops? “License and registration, please?” All the information’s on line in law enforcement databases, why should we be tasked with carrying around stupid little pieces of paper?)

It was useful to push some plausible (I hope they are plausible) numbers through and see what popped out. One result of this exercise is that I am much more gung-ho about mask-wearing than I was, and much more worried that contact-tracing will not be nearly as useful as people seem to think. I understand why people like contact tracing; they assume that they’ll never get infected, so it won’t be them having their privacy invaded by a public health official taking notes on exactly where they spent the previous week, and therefore they don’t have to do anything different at all, life will be great. “And why should I let Google and Apple spy on me, I’m sure they’ll just try to sell me more stuff,” so they won’t want to run any app, either. But if half the cases are asymptomatic and we’re not testing very thoroughly, contact tracing cannot reduce the reproduction rate by a better factor than 0.5. In addition, I think any work-from-home that can easily be continued after the economy opens up more is a good thing; it gives us a slight edge in preventing disease spread, where a “slight edge” might end outbreaks twice as quickly.

So, that’s my rant. Don’t dismiss infection control half-measures, combine a few of them and you get some real results. And don’t quit with the half-measure you’re doing just because some new control has been added, the old and the new can work together.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: