[vhfpmr (OP)]

I’ve been considering whether the correct metric for comparing infection levels between countries should be absolute number, or per head of population, but I can’t make up my mind.

If the outbreak is seeded by a single infected individual, the disease will spread exponentially at a rate determined by the number each carrier infects and the time they take to do so, for as long as the uninfected population remains large enough to approximate infinity.

However, presumably most outbreaks are started by multiple travellers arriving at substantially the same time, and the number of travellers will be proportional to the population.

It's not clear why these plots are such a markedly different shape, either:
https://ourworldindata.org/grapher/daily-covid-cases-3-day-average?yScale=log&time=..&country=GBR+ESP+ITA+FRA+BEL
https://ourworldindata.org/grapher/daily-covid-cases-per-million-three-day-avg?tab=chart&yScale=log&time=..2020-03-15&country=BEL+FRA+ITA+ESP+GBR

[alancalverd]

The travellers will have come from somewhere to do something. As the pandemic began in China we will find national sources where business travellers abound, so large numbers in Europe but few in the less prosperous parts of Africa.

However once a national outbreak is established, its trajectory depends on social isolation or lack of it, so will be very variable between countries with equally dense populations depending on their compliance with lockdowns. It will also be selective, so we can expect in a primitive society such as the USA  that Republicans who flock to salute their Fuhrer and Christians who flock to seek protection from their god, will be disproportionately hit by the second wave.

And then there's the problem of reporting. If you have no cheap, reliable test, you aren't going to record COVID as cause of death distinct from seasonal flu or pneumonia.

The only reliable statistic is the time series of total daily deaths, minus the average for the last couple of years on the same days. 

[vhfpmr (OP)]

But the choice of which metric is best for comparing countries is independent of these factors, you need to decide which measurement is the best way of viewing the data in order to decide whether religion etc are having effects.

After the first person arrives in a country there will be a period until the first diagnosis is made, during which time two things happen:
1) The disease spreads from the first infector
2) More infected travellers arrive

It seems to me that the best metric would depend (in principle) on which of these two has produced the most cases during that period. In practice of course, you're unlikely to be able to tell which is greater, so it doesn't follow that one method is necessarily any better than the other.

There are some right wing fanatics on Twitter who have chosen to use cases per head in order to 'disprove' that the UK is a bad case, and argue that choosing to compare absolute numbers is an example of the left of deliberately lying. That seems like the pot calling the kettle black.

The Oxford Uni plots above strike me as strange, unless the population has changed significantly during the last few weeks, the two plots ought to be exactly the same shape on a log scale, but they aren't. Presumably they're not choosing different datasets just because one set is divided by the population.

[alancalverd]

After the first person arrives in a country there will be a period until the first diagnosis is made

There's the problem: diagnosis. We simply do not have meaningfully comparable data because even in civilised countries 80% of cases never present to anyone capable of making a diagnosis, let alone having the facility to do so with any degree of certainty, and even if he had the kit, there's no statutory requirement to place the result on public record.

A friend of mine let it be known that she was a qualified nurse when her husband was posted to Papua New Guinea, and some patients walked for four days to get a wound dressed - the idea of mass virological testing is absurd in some countries, and hasn't begun even in the UK!     

We know, for instance, that there are many COVID deaths in UK care homes, but there is no actual data - it's a sensible guess if residents start dying with similar symptoms but there is no point in postmortem testing and most of the infected staff will recover. The only credible statistics we have are for deaths in hospital, which probably account for around 3 - 5% of all infections and are thus not a reliable indicator of the number of cases.

One of my good buddies has just been discharged from 10 days in hospital. Both he and his wife had clear COVID symptoms and her infection was clearly traceable, but she recovered completely and therefore does not feature in any public record - or even her own GP record unless she insists on reporting it. He needed significant treatment short of active ventilation, so the hospital record shows a successful recovery from "probable COVID"  but the incidence in that household was twice whatever has been recorded.

[vhfpmr (OP)]

The only credible statistics we have are for deaths in hospital

So you could still ask the same question, is it most helpful to view total numbers, or normalise them to population.

[alancalverd]

Neither. If a hospital service is overwhelmed (UK) or inaccessible (PNG), more people are going to die outside than inside, but we won't have credible diagnostic statistics. In the case of an elderly care home, there may be little point in transferring a seriously sick patient to acute intensive care, especially if they have a pre-existing terminal illness. Assuming normal progression from emergency -> acute or discharge -> intensive care or discharge, the number who die in a given hospital is simply a fraction of the number of ventilators available.  That fraction may be indicative of the quality of intensive care, or of the previous acute care, but it is too small a sample of the population to make reliable inferences about the country outside the hospital.

The statistic "deaths in hospital of COVID-positive patients" is credible but frankly useless. One is reminded of the lost balloonist who yells out to a farmer "Where am I"  to which he replies "About 50 feet above my field."

[vhfpmr (OP)]

Since I started this thread, the concept of total excess deaths over and above the long term mean has gained some popularity as a 'gold standard' for comparing countries, precisely because it goes a long way toward addressing your objections about data reliability. However, since the excess can be measured either as an absolute number, or a percentage of a norm that itself is proportional to the population, it begs the question I asked at the outset.

Is the UK 43 times worse than Austria here, or 4.7 times worse:
https://www.ft.com/content/a26fbf7e-48f8-11ea-aeb3-955839e06441

If I had asked whether speed is measured in miles, or miles per hour, your argument is like telling me that it's irrelevant because I don't have a reliable speedometer. The issue is the unit of measurement, not the practicality of the measurement.

[alancalverd]

The norm is about 1/75 of the population each year, with the denominator varying between 65 and 85 in countries that you might consider civilised.

[vhfpmr (OP)]

There was a scientist on the Today program one morning this week (from UEA I think, but I can't find it now), who's been comparing the effectiveness of individual lockdown measures. She said she had controlled for the variations between countries, but didn't say how, and then used the fact that each country implemented different measures at different times to estimate which measures were most effective (by regression presumably?).

She found that the most effective measures, in order, were:
1. Closing schools
2. Banning mass gatherings
3. Initial closure of shops

The following made little difference:
Closing the remaining shops
Staying indoors

Now there's talk of reopening schools, it seems to me that there might be some mileage in getting someone to review her work.

[alancalverd]

I think "staying indoors" is just a simplification of "restrict your interactions". You are unlikely to catch a respiratory infection in the middle of a field with nobody around, in your own car, or even in a large garden, but modern society crowds people together at work, in shops, etc., so "stay indoors" actually means "don't mix with strangers".

I'm in contact with people in a small mining town in Australia. The "native" population and the few resident managers would have no problem as they are hundreds of miles from any source of infection, but the shift workers who fly in for "2 weeks on/1 week off" may be infective, so the permanent residents have to isolate themselves from the migratory group.