experience has been a good teacher as covid-19 case-fatality rates improve.

Last week I wrote about the use of corticosteroids in treating COVID-19; the drugs, as a class, were shown to perform well in severe cases and were given the all- clear for use on those patients by the World Health Organization’s evidence- monitoring team. This is really the first treatment of any kind for COVID-19 that has been firmly validated by the strongest form of scientific evidence. ( WHO approves the use of the Gilead antiviral remdesivir in early treatment, but the basis for that is relatively slender.)

I’m sure some readers had the same thought I did: strong evidence sure is annoyingly slow to happen along. We’re six months into this business, and if you believe the hype, a vaccine — once thought to be not worth imagining until at least the one-year mark — is right around the corner. This is good news, not bad. The COVID experience has left me less sure than I once was that there will be no further viral pandemics of this scale — it could hardly have the effect of making a person less misanthropic or more confident in the authorities, could it? But it is a hell of a dress rehearsal, and we are finding out how fast Big Pharma can set to work under pressure.

Over the long weekend this left me with the question of how well doctors are learning to treat COVID using only preprint papers, the world medical grapevine and their own experiences. All of which, compared to evidence from randomized controlled trials, amounts to folklore. But folklore is not to be sneezed at (pardon the infectious metaphor). Are case-fatality rates improving?

We now have enough raw data, even in Canada, to make a guess. I looked at the files for Alberta and Ontario, which publish patient data with suspected acquisition dates, and waved my hands over them, chanting powerful statistical spells. As far as I can tell, experience is a good teacher.

You don’t need to be an Escoffier School data chef to see the effect. Imagine we took all of Ontario’s recorded COVID-19 cases that have the epidemiologist’s R marker: they’ve been resolved, either through death or recovery. In other words, we set aside still-active cases. We put the resolved cases into bins according to the official “episode date,” the best guess as to when the virus was acquired by the patient, and we also bin by age group, which will allow us to hopefully compare apples to apples. We’re mostly concerned, of course, with patients over the age of 60; there may be a serious morbidity burden on younger patients, but for this exercise we’re concerned with the people in immediate danger.

In the higher age groups, Ontario doctors and hospitals performed least well in April, when almost literally nothing was known about how to treat COVID-19 and everybody had to improvise. The observed case- fatality rate (CFR) in April was about eight per cent for patients in their 60s; 22 per cent for patients in their 70s; 31 per cent for those in their 80s; and 40 per cent for those in their 90s.

These numbers have been in steady decline since April. For August, remembering that we are counting only resolved cases, only 1.2 per cent of patients in their 60s died. The number for the 70s age bracket, remarkably, is also 1.2 per cent. In the version of the data set I downloaded late Monday, only one person in their (his) 70s developed a fatal case of COVID-19 in all of Ontario in August. This makes this number somewhat unreliable, as astonishing as it is, and suggests that the JuneJuly CFR of five per cent may be a better estimate going forward.

The news is good for the even older, too: the August CFR for patients in their 80s was just 4.8 per cent. The 90- pluses had a bad month, recording a 37.5 per cent CFR, but their death rates are in obvious decline, too: the figures for June and July had been 19 per cent. At the beginning of the pandemic it

everybody really has learned a lot, and in the hardest possible way.

seemed as if the CFR for the extremely elderly would be more like 40 per cent almost as a matter of natural law.

This simple analysis gives the same answer that fancier modelling techniques do, and the numbers are analogous in Alberta. This kind of learning effect is a natural thing to expect, in medicine as much as in engineering or baseball, and I know this because I was a little annoyed in June and July when I played with the numbers and couldn’t yet find evidence of improvement. ( Breaking things down by age groups is important, especially when the age profile of the infected population changes, as it did during the summer.)

The social measures we are all doing our best to stick with, and the defence operations being mounted in care homes for the aged, may also be helping individuals present with less serious cases involving a smaller initial viral load. This means the CFR improvement is not exclusively attributable to health- care heroes, or at least not by me.

But everybody really has learned a lot, and in the hardest possible way. Some of the active cases left out of the analysis may sneak back in as deaths later, but it seems unlikely that this would affect the good numbers from June and July much, and this cannot account for case- fatality numbers in a vulnerable age group dwindling by an order of magnitude. Of course this shouldn’t make anybody less vigilant about COVID- 19. Think how silly you would feel if you died now, rather than back in April when you had an excuse.