Root causing differentials in Sweden

From Why did Sweden have more covid deaths than its neighbors? by Sebastian Rushworth.  

One of the arguments that has been used in support of strict lockdown is that Sweden has had significantly more covid deaths than its nordic neighbours. On the 19th of November, Sweden had registered 637 covid deaths per million people. For comparison, Denmark had registered 140, Norway had registered 57, and Finland had registered 69.

But, as I wrote about recently, the studies that have been done have not been able to find any correlation between severity of lockdown and the number of covid deaths. Which must logically mean that Sweden’s higher death rate was not due to the fact that it didn’t institute a severe lockdown. So, if that is the case, why did Sweden have more covid deaths than its closest neighbors?

One hypothesis is related to travel.  Stockholmers travel more internationally than other cities in Sweden and Stockholm has suffered a disproportionate number of cases.  In addition, there is quintessentially Swedish quirk of school calendars which I know from my youth and which provide a natural experiment.

The first hypothesis is that Sweden, and in particular Stockholm, imported many more cases of covid-19 from abroad before measures were put in place to stop the spread between countries. The main reason for this is that Stockholm has a half-term holiday (“sportlovet”) in late February, when many people go skiing in the alps. The other nordic countries have similar holidays, but they have them earlier. So any Norwegians, Danes, or Finns who went skiing in the alps, would have gone there before the pandemic exploded in that region, while the people from Stockholm were there when infections were spreading at their worst.

The two other large-ish cities in Sweden, Malmö and Gothenburg, provide a useful control for this hypothesis. Both cities have their half-term holiday a week or two before Stockholm, and both were hit far less severely than Stockholm in the first wave. Stockholm experienced 40% of Swedish covid deaths, despite having only 24% of Sweden’s population.

A second hypothesis is demographic.  Sweden has a much larger portion of its population who are dark skinned refugees who, owing to dark skin and little sunlight, are especially prone to Vitamin D deficiency which is associated with high Covid rates. 

Sweden has a much bigger population of immigrants than its nordic neighbors. 19% of Sweden’s population is foreign born, as opposed to 14% for Denmark and Norway, and only 8% for Finland. What this means in practice is that Sweden has a bigger population of people with darker skin, and it has been clear since early in the pandemic that darker skinned people in western countries are much more likely to develop severe covid than lighter skinned people.

The primary hypothesis which he support is that the higher death rate is mostly a function of all-causes-death rate lags.  

Sweden had a much larger vulnerable population at the beginning of 2020 than its nordic neighbours. This can be seen in multiple different ways in the statistics.

The first is that Sweden has a large nursing home population. Relative to population size, Sweden’s nursing home population is 50% larger than Denmark’s. And as I’ve mentioned previously, in Sweden, people don’t go to nursing homes until they are near the end of life.

The second way this can be seen in the statistics is by looking at overall mortality for the immediately preceding year, 2019. If unusually few people die in one year, then unusually many will die in the following year, since there is a carry forward effect (due to the fact that humans are not immortal). 2019 was an unusually un-deadly year in Sweden, and the early part of 2020 (pre-covid), was also unusually un-deadly, which means that there was an unusually large number of very frail old people in the country when covid struck. This same effect was not seen in Sweden’s nordic neighbours – for them 2019 was normal in terms of overall mortality.

To clarify exactly how big this difference is, let’s look at the numbers. In Sweden, overall mortality in 2019 was 2,5% lower than the average for the preceding five years. In Norway, mortality was exactly in line with the average. Denmark and Finland both had mortality rates that were 1% above the average. Denmark, Finland, and Norway were in a much better position in relation to covid from the start. Sweden was always going to have more deaths, regardless of the actions it took.

This is akin to forestry management.  If you suppress natural fires (by analogy, death rates) for any period of time, you end up with more fuel so that when an uncontrollable fire erupts, you end up with worse and or more widespread fires.  A low all-causes mortality rate in 2019 ensured that there was a larger at risk population in 2020.

The mainstream media report the headline numbers which are often either meaningless (owing to definitional issues) or misleading out of context.  The two primary measures (among other useful ones) are the total all-causes-mortality rate.  If it is rising it may because of a lag effect which needs to be taken into account but otherwise can crudely be accorded to Covid-19.  If it is not rising, either there are mis-categorizations of death or there is some causal death substitution occurring (as when lockdowns decrease Covid deaths but engender higher deaths from suicide or overdoses).

The second primary measure has to be in actuarial adjusted years of life lost.  Losing primarily people in their seventies and eighties is an emotional and wrenching tragedy but not societally consequential were the loss primarily 18-25 year olds as was the case in the 1918 Spanish Flu.  Years of life measurements capture the enormous differential in consequence of the latter over the former.  

To make this cruel reality explicit, let's assume that the Covid adjusted years of life loss is 6 months (we are losing the very elderly who are also very sick with co-morbidities).  If we lose 300,000 people, then we have lost the equivalent of 150,000 person years.  

If, however, we lose 300,000 25 year-olds with a life expectancy otherwise of 75 years, then our cumulative years of lost life is (50 years x 300,000) 15,000,000 years of adjusted lost life.  100 times a worse tragedy than we are facing now.