A Guesstimate to convert microcovids to dollars. Microcovid.org is the best calculator of Covid-19 risk that I’m aware of, but the microcovids that it returns are unintuitive for me. Peter Hurford’s spreadsheet converts risks to dollars, which are very intuitive for me. My Guesstimate is a probabilistic version of the lower part of that spreadsheet but also includes risk from long-term damage. The Guesstimate allows you to plug in the microcovids range and get a dollar result. The article is password protected because I’m afraid I made mistakes that might kill people (not because it’s old, yet).
Introduction
You can find my Guesstimate here.
Warning: I know just about nothing about epidemiology or even medicine in general. Don’t trust my judgment on questions in those fields.
Warning: I copied a bunch of numbers from Peter Hurford without understanding them, and he’s also not an epidemiologist.
Warning: See the Limitations section for a lot of limitations of the model. In particular, the model ignores the risks you impose on others because I have no idea how to narrow those down.
Warning: I invested about 30 minutes of research into this.
Usage
The idea is that you can either live a normal life, or you get Covid-19. In the second case:
your life expectancy drops because you have a risk of dying from Covid-19,
you may suffer long-term cognitive (and other) damage, and
you suffer flu-like symptoms (or worse) for 2–3 weeks.
So you need to put in the data from microcovid.org about the risk of your scenario in question. That’s the microcovids range and the duration of the event.
Further, you need to put in your estimate of the value of your life (your whole life, including the parts that are past), the fraction that you think your life will still be worth to you after sustaining long-term damage, and the cost that you’d pay to avert a day of suffering from Covid.
Finally, the model needs some general data on you: your age, life expectancy at birth, sex, and various risk factors. Note that you can edit the Guesstimate model and use the calculator without anyone seeing what you put in.
The model then tells you that your scenario comes with a risk of $41–680 (median $190) or so.
Then you can reflect whether it’s really worth $190 to you compared to some lower-risk counterfactual. E.g., I like to go bouldering if the median Covid risk is below $20. If I didn’t have my climbing wall at home, maybe my threshold would be higher at this point. Or, if I have some other bad disease, I’d accept the higher risk of a doctor visit.
Oh, and ignore the pre-entered inputs. Those are from whatever scenario I’ve wanted to calculate most recently.
Limitations
But that brings me to the greatest limitation of the model: It’s only about personal risk. I have currently no idea how to narrow down the additional risks of:
infecting others who infect others and so on until someone dies who would otherwise have lived for much longer, and
infecting others who infect others and so on to the point where Covid manages to stick around in the population whereas it would’ve otherwise been eliminated.
My estimate of the long-term risks is based on the single study “Cognitive deficits in people who have recovered from COVID-19 relative to controls: An N=84,285 online study,” the only one I’m aware of (apart from this news article)
It makes a good first impression, but of course Covid-19 hasn’t even been around for a year, so data on long-term effects is necessarily rather short-term. My estimate is that the long-term effects make a counterfactual difference for 0.3 to 30 years, log-normally distributed.
The study compares people who’ve probably had Covid to people who didn’t, so it needs to control for all the ways in which formally Covidious people may be different from average people – they survived, they may be unable or unwilling to protect themselves, etc. Controlling for things is hard, I heard, and they corrected only for “age, gender, education level, income, racial-ethnic group and pre-existing medical disorders,” not for like 60+ factors. (Later in the text they say, “… after factoring out age, sex, handedness, first language, education level, country of residence, occupational status and earnings.”) So that may not be good enough.
Also note that the sample of people who were ill at all (and the fraction of those who had covid-19) is much smaller:
Amongst 84,285 participants, 60 reported being put on a ventilator, a further 147 were hospitalised without a ventilator, 176 required medical assistance at home for respiratory difficulties, 3466 had respiratory difficulties and received no medical assistance and 9201 reported being ill without respiratory symptoms. Amongst these 361 reported having had a positive biological test, including the majority of hospitalised cases.
I did a quick aggregation of the study results to put a simple range with a big sample size into my model. Separating the ranges by age would be a lot of work, and the cohort sizes of the people with the worst symptoms per age group are tiny.
Peter Hurford’s spreadsheet also contains a few point estimates that I don’t understand. (There’s a source, but I haven’t looked at it.) I haven’t converted those to sensible ranges because I don’t know what ranges would be sensible.
The model does not take medical costs into account.
The model is probably importantly wrong and incomplete in unknown ways.