Category: Covid-19

In the UK, about 4 million people swim outdoors in pools, seas, lakes and rivers. I can’t find a figure specifically for rivers, but swimming in the River Cam between Grantchester and Cambridge is certainly popular — the Newnham Riverbank Club has several hundred members, and many more people swim from other points in Grantchester Meadows — and I believe this is true at many other sites around the UK, and in many other countries.

It’s a little surprising that there has been almost no discussion of possible covid-19 transmission risks from wild swimming. Here is one interesting but inconclusive article, focussing mainly on possible risks from sewage. This also contributed to cautious advice during peak lockdown. Nonetheless, government guidance since mid-May has been that wild swimming per se is low risk so long as social distancing is maintained.

It may well be; I hope so. But we’ve learned during the pandemic that there seems to be a high transmission risk in settings ranging from choir practice to meat-packing factories. Afterwards, very plausible explanations have been given, but no one seems to have identified the risks in advance. Might river swimming at popular sites be another potentially risky setting?

I have some concerns, which may not stand up to empirical test but which I haven’t seen considered carefully. My worry isn’t about transmission from sewage: I’m going to assume that swimmers will avoid sewage-contaminated waters. The worry is that person-to-person transmission while swimming in a river might be much more effective, at much longer distances, than is generally true outdoors, or even perhaps indoors.

First, let’s look at the survival of coronaviruses in water. This study looks at other coronaviruses, but I’ll take it as applicable to the covid-19 virus in the absence of contrary evidence. In the most hostile aqueous enironment studied, primary filtered effluent at 23C, 1% of coronaviruses remained active after 1.5 days. The deactivation rate seems to be roughly time-independent, which gives us 10% still active after 18 hours, or ~90% still active after 1 hour. If someone sneezed into your river an hour ago, any coronaviruses they expelled are likely still active.

Ok, but we’re also not sure how long coronaviruses stay active in air or on surfaces. You can keep a social distance — let’s say the recommended 2m — while swimming, just as you would elsewhere, so what’s the problem?

Here’s one possible concern. In the air, droplets fall to the ground, and aerosols disperse in all directions. If someone coughs or sneezes while swimming in a river, droplets go into the water surface. What about aerosols? I don’t know — we need a fluid dynamicist and probably a range of experiments. But rivers have banks, and a direction of flow. Maybe aerosols mostly stay in a mist not far above the river, and over time a fair fraction maybe end up in the river.

And here’s another concern. Almost no one wears a mask while swimming (and it’s not obvious any standard mask would be effective). Few people wear goggles while swimming in the wild. Your eyes, nose and mouth are all close to the water surface — right in the zone where viruses are potentially concentrated. You’re often breathing hard, splashing water into your face, probably inhaling and swallowing some, and of course inhaling any aerosols or droplets above the surface.

Still, rivers are big, sneezed droplets are small. Surely they quickly dilute to irrelevance? Well, let’s try to estimate. A sneeze might emit 200 million viruses; an infectious dose might be 1000 viruses.      So you need to inhale just 1/20000 of the viruses from a single sneeze to be infected.  Let’s first think about the aerial route. Take a 10m river with 1m high banks; suppose the cough/sneeze spreads over 1m along the river, 1m above the river, and 10m across the river. Crudely idealizing, suppose that for some while it moves downriver as a 10x1x1 box.      Your breath has volume ~ 6 litres.     is 1000 litres.     If you breathe in while going through the box, you might inhale of the 200 million viruses, i.e. 12 infectious doses.     For how long is the rectangle model good enough to give roughly right answers?       It’s very hard to say without empirically modelling.      At a guess, if someone sneezes 10m upstream from you, the situation is worse than the model suggests: the viruses won’t spread out across the river by the time they reach you.     At another guess, if they sneeze 1km upstream from you, the situation is better –- maybe much better — than the model says.    But I don’t know; I wonder if anyone does.

Now suppose most of the viral load is in the water.   The River Cam is  ~6mx2m = in cross-section.     For a sneeze with 20 million viruses to dilute to 1 virus per cc, assuming equal dilutions at all depths and across the river, it would have to spread out uniformly over a length of 166m.     If it stays in the top 20cm — maybe more reasonable, for a long while, for a slow river on a calm day — it has to spread over 1.66km. If it stays in the top 2cm, the length is 16.6km. For 1000 viruses per cc, staying in a 2m x 10cm cross-section, the length is 16.6m.   Take a look at the photo above, and ask how confident you are that the river flow will quickly stretch out sneeze particles over that volume. Now 1cc is about a quarter of a teaspoon. How confident are you about not exposing your eyes, nose and mouth to ¼ teaspoon of water while swimming?

My conclusions? Swimming in an uncrowded unconfined area of sea seems a much better bet: there’s much more turbulence and no confinement except the beach boundary. My guess is that dilution is effective enough in the sea if you stay well away from others. I’m not completely confident about this, but personally, I’d take the risk.

If you’re going to swim in a river, I’d try to be upstream of, well, ideally everyone. All else being equal (don’t let a real drowning risk replace a hypothetical infection risk!), you should maybe prefer large faster-flowing rivers to small slow ones. I would keep well away from anyone not in your household if they’re upstream of you. I’ve no idea whether 100m might be a safeish distance in a small river; I don’t see any reason to think 2m is.

For me, regretfully, the unknowns deter. I love Cam swimming, but haven’t indulged this summer. The risks clearly aren’t huge — no clusters of cases have been reported among wild swimmers in Cambridge, or anywhere else as far as I’m aware. But the general infection rates in Cambridge, and most of the UK, have, thankfully, been low over the summer. There may thus have been few or no asymptomatic but infected people swimming in the Cam, just as there may be few or none in any given pub or gym in Cambridge. This leaves a niggling worry that wild swimming, like bar-hopping or gym-going, may nonetheless be a relatively risky activity.

If you can produce more reassuring data or better arguments, I and (at least) one or two other cautious swimmers would be very grateful.

[Disclaimer: these are my own informal calculations based on my inexpert impression of the science and data, which themselves seem still quite uncertain. They’re meant to encourage you to look at the current data and do your own.

Updates September 14, 2020.

1. After writing the original post, I found a nice article by Tim Harford that spells out the basic risk estimate calculation.
2. The figures below have been updated with the most recent ONS and ZOE estimates (as of 14th September). ]

It’s difficult to know what’s worth doing, or not doing, to reduce the risk of covid-19 infection. Here’s a way of cutting through all the uncertainties and getting an estimate of your actual risk levels. I’ll give current figures for England; obviously the method works in any region where reasonable infection rate estimates are available.

First, find the current estimated daily infection rate for covid-19. Early September estimates for England from the ONS are ~3200 per day for people living in households (i.e. not care homes or hospitals). The ZOE Covid-19 Symptom Study give ~4220 per day; this is for the entire UK and appears to be for all settings (including care homes and hospitals). All of the figures come with error bars; for example, the ONS 95% confidence upper bound is ~4600 per day. The figures currently seem to be increasing; the highest current estimate (that I’m aware of) is that they’re doubling every 7-8 days at present; other estimates suggest a lower rate of increase.

Assuming you’re not in a hospital or care home, you might thus conservatively go with twice the latest figures, i.e. roughly 8000 per day, for today. The population of England is 55 million, and the numbers in care homes (<500000) and hospitals (smaller) are small fractions of that. Call the residential population 50 million, rounding down.

So, if you’re a typical resident of England exposed to typical covid-19 risk, your risk of infection is ~ 8000/50000000 = 1/6250 per day. Annualized — if the risk were the same every day for the next year — this gives a risk of about 6% of infection.

Your risk of dying from covid-19 if you’re infected depends on your age, sex, and health. For most people it’s not that high. Estimates of the overall infection fatality rate vary; if we take it to be 0.6% then the average English resident’s annualized risk of death from covid-19 is the product, 6% x 0.6 %, about 1 in 2700. That’s about 1/25 of the overall mortality rate. Your covid-19 death risk might be higher, if you’re old or have existing conditions — but so will your overall death risk, and they roughly scale together. If you die in the next month, unless you’re very atypically at risk, it’s pretty unlikely it’ll be of a covid-19 infection you contracted today.

Of course, the risk won’t be the same every day for the next year. There may be a very serious second wave. The current annualized risk is only a good guide for decisions right now. If you’re behaving like a typical English resident, in a typical environment, it looks as if your risk today is quite low. If you’re following government guidance, keeping social distance, wearing masks if you go shopping or on public transport, and don’t have a job that exposes you to atypical risk, your risk is very likely lower than the figures above. Most non-mandated risk avoidance measures — such as disinfecting groceries or wearing masks outdoors when not socializing and not in crowded areas– will probably not greatly reduce your already low risk. Most experts suggest that the large majority of infections come from indoor exposure to airborne droplets or aerosols. The data aren’t solid, but my sense is it’d be surprising if 10% of infections, even now that we’re aware of the indoor risks and taking countermeasures, come from other sources. So it’d be surprising if disinfecting groceries and non-social outdoor mask-wearing reduced the average risk of infection by more than an annualized 0.6%, or the average annualised risk of death by more than 1 in 25000.

Are these extra countermeasures nonetheless worthwhile today? The emerging consensus on mask-wearing seems to be that it’s more for social good than for individual benefit: my mask protects you much more than me. If a large enough proportion of people wear masks when near others, then, models suggest, the transmission rate can be significantly reduced — plausibly by enough to mitigate a second wave. The inconvenience isn’t that great, and it’s probably a habit we should get used to when we’re anywhere around people. Wear-masks-in-public is an easy rule to communicate, follow and enforce; wear-masks-in-shops-transport-sufficiently-dense-crowds-and-less-than-fleeting-conversations, not so much.

The case is much less clear for grocery disinfection and the like. Rationally, we should put finite prices on our lives and on our time. A quick short-cut (ignoring future discounting and quality of life weighting) is to convert everything into time. A minute a week disinfecting groceries means investing about 1/10000 of the year to avoid at most a 1/25000 risk of death. If your remaining life expectancy is more than 2.5 years, that’s perhaps worth it — but unless you expect to live for 75 more years a more realistic 30 minutes a week perhaps isn’t. (If you’re very young, you might expect to live for 75 more years, but your death risk will be much lower. So you need to be optimistic, not just young, to make it worthwhile on this estimate.) But there are other costs: covid-19 also carries risks of serious and prolonged illness and perhaps lifelong loss of quality or life and lower life expectancy. Maybe the true cost of those risks is several times that of the death risk: my impression is that we just don’t know at present. Still, that might tip the balance towards grocery disinfection. There’s also some social benefit in personal risk minimisation: if you lower your risk of infection, you lower your risk of spreading. On the other hand, assigning 10% of total infections to infected groceries may well be far too high. If grocery disinfection brings you peace of mind, perhaps it’s worth doing, but I’d try not to worry about occasional lapses.

Tl;dr right now, in England (and the rest of the UK), unless you’re especially vulnerable or exposed, or live in a hotspot region where additional lockdown measures are in force, I think your personal covid-19 risk is still low today. Protect others with masks by all means; follow government guidelines, and you’ll be very unlucky to get ill today. But everything depends on the infection rate; I would follow it and reevaluate weekly.