How to evaluate Covid risk in an evolving pandemic
Estimating risks to decide what safety measures are worthwhile, and how to think about changing risks with surges and variants
This is part of a collection of articles on Covid-19
Summary
Use microCOVID to estimate your risks of catching Covid and evaluate the impact of different safety measures.
Focus on the highest-risk parts of your day - time spent in crowded indoors spaces. They can easily account for the vast majority of your risk even if they're relatively short.
See below for rule-of-thumb guidelines to risks and the effectiveness of different precautions
The primary metric that you should look at to understand how your risk of Covid infection is changing over time is the current prevalence of COVID in your region (i.e. the new case count, but adjusted to reflect estimated total cases, not just confirmed cases). As that goes up, your risk goes up proportionally.
In a surge like right now, cases in some regions have gone up 5x or more in a single week, meaning the risk of the same situation is 5x higher; so your choices about what Covid precautions to take may need to look very different from one week to the next.
Even if one thing you're doing is unavoidably risky, that doesn't mean it's not worthwhile to take effective Covid precautions at other times. Think of Covid risk as linear in time (i.e. microcovids per hour) and make the best tradeoffs for any given span of time.
E.g. if you are dining at a crowded indoors restaurant and there's 20 minutes before the food arrives, it's still worthwhile to wear a mask for that duration even though you are taking it off to eat later.
Ok, we can estimate the chance of infection, but how bad is getting Covid? (This part is much less certain and my estimates are low-confidence, but this is the best I have right now.)
For the average vaccinated person, the main risk is probably the chance of a fatal case of Covid, and this is a lot higher than most everyday risks.
Around 3,000 microcovids costs you 1 hour of life expectancy. So, let's say you plug a scenario into the microcovid calculator and it estimates that a particular safety precaution (could be wearing a mask, or deciding not to dine indoors at a restaurant) will reduce your risk by 1,000 microcovids. You can ask yourself: would you rather lose 20 minutes of life expectancy, or take this safety precaution?
For young vaccinated people, the fatality rate is much lower, but still pretty high compared to other everyday risks. And long covid may be a larger risk:
A very rough estimate (which I still have very low confidence in, pending better long covid research and analysis) is that 6,000 microcovids may cost you 1 hour of life expectancy (quality-adjusted) due to long covid. You can think of the tradeoffs of safety precautions using the same reasoning as above.
Estimating your Covid risks
I highly recommend microCOVID - this is an excellent calculator to estimate the risk of getting infected by Covid in different situations. You can use this to see how risky an activity is, and how effective safety measures like masking, distancing, vaccination mandates, etc are. Their site also has extensive research and information on these topics.
I make one major adjustment: I estimate the risk reduction factor for being outdoors is about 1/100, not the 1/20 in their calculator, so I divide all outdoors risks by an additional 5. Microcovid says that they believe 1/20 is a conservative upper bound, see https://www.microcovid.org/paper/all#outdoor for details. This means I basically think there is negligible risk of covid transmission outdoors (unless you know someone is covid-positive - you should still avoid/minimize contact then.)
When thinking about your risk, focus on the highest-risk parts of your day - time spent in crowded indoors spaces. They can easily account for the vast majority of your risk even if they're short. E.g. for me looking at the risk of going to the office, half an hour eating lunch indoors in close quarters with many people dominates the risk compared to an entire workday sitting masked and distanced at my desk.
Rule-of-thumb risk estimates
Some quick rule-of-thumb estimates for reference - these are the same as microcovid's calculator, but I find it useful to have some numbers I can easily work with in my head. This is also a walkthrough of the basic principles of how microcovid’s calculator works.
Key data points we need
The first number we need is the estimated prevalence of COVID in your region from microcovid - e.g. this is about 8,000 microcovids for SFBA as of Feb 11 2022
On microcovid, you can see this number in Risk Profile for “An average person in your area” - this is the current Covid rate per 1 million people. (You can look at the overall average, or the average among those vaccinated.)
Usually other cities are in the same ballpark but right now with the Omicron surge there’s massive differences between regions, and also rates are changing rapidly from week to week.
This is what percentage of the population is estimated to be Covid-positive right now, and in the highly transmissible stage of infection. Microcovid estimates this based on current testing data and so it is automatically calculating your current risk as the caseload varies.
For more details, expand the Details section below the location dropdown - this is the adjusted prevalence.
Estimate the number of nearby people - "how many people are usually within 15 feet (5 meters) of you, at any given time?" - see microcovid's calculator for details
Duration: how long is the activity?
Baseline risk
For indoors, without precautions. These numbers are assuming you are fully vaccinated with a booster, and at least one of those shots was Pfizer/Moderna.
If in close contact with someone with Covid, you have a baseline 3.5% chance of infection per hour (this includes the risk reduction from you being vaccinated)
Updated Jan 6 with more Omicron data
So an indoors interaction with people of unknown vaccination status, with no precautions, is: 300 microcovids per hour, per other person (this is 3.5% * Covid prevalence, adjust your numbers based on the current prevalence in your area)
Indoors interaction with everyone vaccinated, no other precautions: about half as much
Risk adjustments of precautions
(1/5x risk reduction means you divide the risk by 5, e.g. 500 microcovids becomes 100. Multiple precautions stack together by simply multiplying the risk multipliers together.)
Outdoors: 1/100x risk (Here, outdoors means being in the open air - being in an enclosed tent is indoors.)
This is different than microcovid as discussed above
How much does being indoors but near an open door/window help? What about in a crowded 3-sided tent? I don't know, let me know if you do have studies, air flow models, etc.
Good air filtration/ventilation: 1/4x risk
Masking:
Common examples:
1/2x risk if both you and they are wearing cloth masks
1/16x risk if you are wearing a seal-checked N95 and they are wearing cloth masks.
1/8x risk if both you and they are wearing surgical masks
To calculate it for specific mask types:
For your mask: 1x for cloth, 1/2x for surgical, 1/3x for KN95 or poorly-sealed N95, 1/8x for seal-checked N95, and 1/20x for P100 (see my masks guide)
For their mask, basically multiply those by an additional 1/2x.
Then multiply the "your mask" and "their mask" factors together.
Distancing: 1/2x risk for staying 6 ft apart instead of 3 ft apart
Talking volume: 1/5x risk if everyone is silent, 5x risk increase if everyone is talking loudly/singing.
Testing: if the people you're interacting with have a negative rapid antigen test, estimate that as 1/6x risk, based on an estimated 85% sensitivity, sources: 1 2 3. (Low confidence in this estimate, let me know if you have better research)
Vaccination (their vaccine status, we are already assuming you are vaccinated): vaccinated and boosted vs unvaccinated is ~1/4x risk (reduced efficacy due to Omicron, probably 1/3x-1/5x). Vaccinated vs unknown vaccination status is ~1/2x risk reduction (depends on vaccination rates in your area). So a situation with a vaccine mandate is ~1/2x risk reduction vs a situation with random average people.
Their risk profile: you can also try to adjust for how cautious or risky their behavior is compared to the average (how effective are the safety precautions they usually take, how much do or don’t they participate in crowded indoors activities). E.g. if you believe their behavior is 1/2x as risky as average, the risk to you also drops by 1/2x. You would want to estimate the average riskiness of the group. I usually just use the population average, but for small groups this can make a noticeable difference.
How to make intuitive sense of these risks
For many people, working with small probabilities is unintuitive and confusing. Here’s one way to think about it that may be helpful.
Let's say you spend an hour visiting someone with Omicron. If you are vaccinated+boosted, indoors, within ~3 feet, and you’re both maskless, then you have about a 3.5% chance of getting infected. In other words, if you made a one-hour visit to someone with Omicron every week, you can (very roughly) expect to be infected on average once every 28 visits.
Now if you put on a well-sealed N95+ mask, that chance is reduced by a factor of about 1/8, to 0.45%, or (very roughly) on average once every 220 visits.
You can do a similar analysis with other precautions, and other situations. In this example I explained it with someone who is known Covid-positive, to make the explanation clearer, but the same analysis works with regular situations where you don’t know who has Covid or not. Just take 1 million divided by the microcovid estimate, and that is the expected number of times you can repeat the situation and expect to get infected roughly once, on average.
How to evaluate your risk as the pandemic evolves
The primary metric that you should look at to understand your risk of Covid infection (and how that risk is changing over time) is the prevalence of COVID in your region - what percentage of the population is estimated to be Covid-positive right now (and in the highly transmissible stage of infection). Microcovid estimates this and so it is automatically calculating your current risk as the caseload varies. You can see the estimated prevalence by expanding the Details section below the location dropdown - the adjusted prevalence is the number you want.
So you should adjust your safety measures based on the current prevalence (and of course based on the safety measures of the others you're interacting with too). Other factors, such as vaccination rate, help simply by reducing the Covid prevalence, so you don't need to factor them into your current risk calculations. Similarly, lockdowns drive reduced covid rates, but they don't change your risk calculation for the same scenario aside from looking at the current covid rates.
During a surge, the Covid prevalence can change rapidly from week to week, so it's important to adjust your behavior quickly to match. For example, during the Delta surge, the case count in SFBA went up 7x in a single week, meaning every activity was essentially 7x riskier. Which means that if you're using a microcovid risk budget, the same activity went from being a once-a-week activity to a once-per-2-months activity.
(It’s hard to accurately estimate the true case count from the confirmed case count, especially because changes in testing patterns can be a huge confounder, but I think most of the time it still gives you a reasonable picture to work off of.)
Variants
If/when a new variant develops, it can change the calculus in a couple ways. If it's more transmissible (including via vaccine escape), it will both cause the Covid prevalence to go up (which is factored into risk estimates on microcovid automatically), and also increase the risk of you getting infected for the same prevalence. Of course, those two together are the feedback loop that causes a surge of exponential growth, but when you're looking at your own current risk you just need those two estimates separately. If it's more lethal or causes more severe disease, that increases the danger as well.
It requires a lot of time and research to accurately estimate these additional effects, but they are second-order adjustments in the risk model compared to the primary thing, which is the caseload. When Delta hit, the transmissibility to an unvaccinated person went up by a factor of 1.5, and the transmissibility to a 2-dose mRNA vaccinated person went up by a factor of 2.6, according to microcovid's research. Yes, 2.6x is a big difference. But the case counts went up ~15x - which makes any given scenario 15x riskier because the people around you are 15x more likely to be infected. The fact that a 2x increase in transmissibility causes a 15x increase in case counts is the nature of exponential growth, and the point is that while transmissibility determines future case count, today's case count is still the most important determinant for today's risk of infection.
Severity and availability of treatment
There are some other factors that can be relevant to your risk of serious complications from Covid, such as ICU capacity, which may affect quality of care in case you get severe covid, and availability of treatments such as Paxlovid. Microcovid only models chance of infection, not chance of severe disease. During a surge, these factors could make a fairly big difference to the expected harm you suffer from severe acute covid, and that's important for the average person. If your ICUs are starting to get full and hospital staffing and resources are strained, that can increase your risk a lot.
However, if you're a young vaccinated person (which is most of my current readers), you are at relatively low risk of severe acute covid, and long covid is likely the bigger expected source of harm (see below) - so I don't think considerations such as ICU capacity make a lot of difference for very-low-risk groups.
Linearity is how you should think about risk and safety measures
I think of Covid risks as basically linear in time and number of people. So I find a good way to think about risk is often microcovids per hour, not necessarily microcovids per event. One reason is that the costs of e.g. masking or distancing in terms of discomfort or reduced quality of social interaction are often best measured on a per hour basis as well.
Linearity means spending 40 minutes in the same setting is twice as risky as 20 minutes, and being with 10 other people is twice as risky as 5 other people. Additionally, the risks of different events simply add up.
Maybe that sounds obvious, but for a practical example where this leads to counterintuitive recommendations contrary to how people typically behave, let's say you're dining at an indoors restaurant. It's true that wearing a mask for 30 seconds while walking to your table and then immediately taking it off is essentially pointless, but it is perfectly sensible to wear a mask for the 20 minutes before food arrives - sure, you will be taking it off to eat, but wearing a mask for those 20 minutes in a crowded restaurant where other diners are unmasked is still protecting you for those 20 minutes - which would reduce my covid risk by 600 microcovids as of today. If you would have worn a mask for 20 minutes in any other similar setting, then you should wear a mask here too. And if you think about e.g. spending 20 minutes in a less-crowded grocery store where everyone is masked, that's much less risky than the restaurant, but everyone still wears masks there.
The justification for treating risk as approximately linear in time is:
Risk of infection and severity of infection are highly dependent on the amount of Covid virus you are exposed to. The longer you are exposed, the higher the expected viral dose you are exposed to, and I think it's reasonable to model this as roughly linear.
In many settings, such as restaurants and grocery stores, you are exposed to different people over time, so if you spend twice as long you'll have roughly twice the chance of being exposed to someone Covid-positive.
This only holds up to a point - there probably isn't much difference in risk between spending 3 vs 10 hours in close indoors unmasked interaction with the same person (e.g. people you live with). But I think for most typical scenarios, modeling risks as linear is close enough.
Estimating risk of illness and death, from acute and long Covid
Confidence level: low (back-of-the-envelope calculations based on a few large-sample but hard-to-interpret studies)
The above risk analysis is focused on chance of infection. But how much does being infected actually harm you? This is important to decide how to act - what's worth doing and what's not, and what safety precautions are worth taking. How to evaluate the tradeoff between the risk you incur of future illness vs the cost (in effort, time, money, sacrifices, opportunity cost) of taking some safety precaution or avoiding doing something that's valuable to you.
There's two main components:
The chance of death due to severe acute illness. This is the main source of expected harm for the average vaccinated person (and even moreso for unvaccinated people).
The chance of chronic illness, i.e. long covid. This is the main source of expected harm for people at very low risk of severe acute illness, like young healthy vaccinated people.
This spreadsheet on Covid mortality and long covid estimates has my back-of-the-envelope calculations for both of these components, and my attempt to convert these risks into more intuitively useful numbers.
My analysis of long covid is based on these articles and the studies discussed in them. (To date, there have been a few large studies but the results are hard to interpret because you have to be very careful about comparing to controls.)
Scott Alexander: Long COVID: Much More Than You Wanted To Know
Matt Bell: If you're vaccinated, your main risk from the Delta variant is probably long-haul COVID
AcesoUnderGlass: Long Covid Is Not Necessarily Your Biggest Problem
Here are some rule-of-thumb risk estimates for the average vaccinated person:
The main thing to be worried about is probably severe acute illness, although long covid may not be far off in importance
The average person in this population loses ~4 days of life expectancy to covid, per year
That's a lot! For comparison, this is roughly 2x the life expectancy lost due to fatalities from car crashes.
~3,000 microcovids of covid exposure loses you 1 hour of life expectancy
So, let's say you plug a scenario into the microcovid calculator and it estimates that a particular safety precaution (could be wearing a mask, or deciding not to dine indoors at a restaurant) will reduce your risk by this many microcovids. You can ask yourself: would you rather lose 1 hour of life expectancy, or take this safety precaution? (In general, calculate the hours equivalence by dividing the microcovids by this figure.)
Note, for young people, 1 micromort and 1 hour of life expectancy lost are in the same ballpark. See also microlife.
For young, vaccinated people:
The average person in this population loses ~1 day of quality-adjusted life expectancy to long covid, per year
For comparison, this is similar to the life expectancy lost due to fatalities from car crashes
~6,000 microcovids of covid exposure loses you 1 hour of quality-adjusted life expectancy.
Same analysis as above
~1,000 microcovids of covid exposure causes you an average of 1 hour suffering from long covid (chronic fatigue syndrome or chronic respiratory illness).
You can use the same analysis as above, but now you're considering: would you rather take this safety precaution, or suffer 1 additional hour of chronic illness?
Covid compared to other everyday risks
Some back of the envelope calculations show that for vaccinated young adults in the US, the typical annual risk of death from Covid about as high as flu, and 1/10 as much as driving. You can take that as either reassuring about Covid, or scary about driving/flu, or probably a mix of both. Note that for higher-risk groups, e.g. unvaccinated or older people, Covid is much riskier.
Covid, among vaccinated US young adults (during Delta and before Omicron - data as of 11/20/21, CDC with adjustments for official vs true case count)
20% annual case rate (94.41 * 365 / 7 * 4 [estimated true Covid cases per reported])
57 per 100,000 annual hospitalization rate (CDC - data table at bottom, 1.1 * 365 / 7)
1.1 per 100,000 annual death rate (0.02 * 365 / 7)
2019 flu season, among US young adults (from CDC)
0.7% symptomatic illness rate
0.3% medical visit rate
40 per 100,000 hospitalization rate
1.2 per 100,000 death rate
So in November (tail end of the Delta period), Covid was about as risky as a typical flu season for vaccinated US young adults, in terms of hospitalizations and deaths (on an annualized basis).
Driving stats, in 2019 in the US (from NSC):
US population: 328.3 million
Deaths in motor-vehicle crashes: 39,107 = 11.9 per 100,000 population
Medically consulted injuries in motor-vehicle incidents: 4.5 million = 1.4% per population
This means "an injury serious enough that a medical professional was consulted".
People involved in motor-vehicle crashes with property damage or injury: 27 million = 8% per population
The NSC statistic is the number of drivers involved, but does not count passengers, so I multiplied by the average occupancy per vehicle which is 1.5 people (source).
So Covid is about a tenth as risky as driving for US young adults in terms of deaths (driving is a very large everyday risk!) - but that's not counting the long covid aspect (see discussion above).
Open questions
I haven't yet had time to look into these deeply and write about them, but I'd love to hear your thoughts.
What does all of this mean for how you should act about other everyday risks?
Covid does appear to be a much larger risk than most other everyday risks we take, and it's also a novel problem so we've had less time to develop robust safety measures against it, so it does make sense that Covid is a good target to focus on. But there may be other large risks in your life that you can address even more easily than Covid. One example I learned about in the process of researching Covid was air filtration and how you can easily save 2 months of life expectancy by using air purifiers. That's huge compared to the Covid numbers above!
If Omicron is so transmissible that we're all going to get it, doesn't that change the calculus for whether it's worth even spending time thinking about precautions?
See here for an updated version of this section
Perhaps. I think this depends on how much immunity infection provides from later reinfection, especially long-term and with new variants.
It does seem true that if the vast majority of people are likely to get Omicron before the next big variant arrives, then Covid precautions would be of little value aside from flattening the curve considerations (hospital and ICU capacity, etc). This assumes that Omicron provides strong long-lasting protection from Omicron reinfection - if not, then you might expect to get Omicron multiple times, so you should still try to minimize the number of times you get it.
But in our experience so far, the next big variant will probably arrive after about half a year on average. For Delta, about 20% of the vaccinated population was getting infected with Covid per year, so only about 10% in half a year (based on CDC). Omicron case rates are far higher, but the bulk of the surge has always been among the unvaccinated population, and overall it seems like there's a good chance that you will never get Omicron before Omicron is displaced by another variant.
The next consideration that goes along with that is, how much immunity does one variant grant from infection from a future variant? And how does that compare to vaccine efficacy? It seems to me that they're going to follow similar trends, most of the time. In other words, I wouldn't bet on Omicron granting me hugely more immunity from the next big variant than my vaccine does, and therefore I'd rather not get Omicron if I have the choice.
That said, I don't think we will (or should) be going back to early pandemic lockdown-style life. One important thing to remember is that that despite the variants, vaccination (especially with a booster) massively reduces your risk compared to back then. Lockdowns are a very costly, short-term intervention. Masks and testing are important too, especially during a surge, but I don’t think we should expect to keep that up for years and decades. But for the long-term, I put most of my stock in lasting countermeasures, namely vaccines and air filtration/ventilation. I think that's the only way we as a society are going to effectively handle this and future pandemics.
Changelog of major updates
Jan 3 2022: corrected an arithmetic error in “Covid compared to other everyday risks” section which gave a 5x higher death rate for young vaccinated Americans than correct (this did not affect the numbers in other sections)
This article is an evergreen, living document that I will continue updating over time.
Like with all my posts, feedback/questions/comments are very welcome - feel free to send me an email or message, or comment on this post.