Thursday, October 22, 2020

Reports of Sweden's deaths highly exaggerated

by George J. Dance 

As the governments of Europe began reimposing lockdowns this week, the pro-lockdown media have been reminding us what happens to bad governments that fail to shut down their societies; the favorite cautionary tale, as always, being that of Sweden. "The Swedish COVID-19 Response Is a Disaster," Time magazine headlined October 14. On October 19, CBC Radio's "As It Happens" news show featured a "Swedish virologist [who] says her country's COVID-19 strategy has failed, but nobody will admit it." "As White House eyes 'herd immunity,' Sweden's no-mask approach is failing to contain COVID-19," Yahoo News chimed in on October 20, the same day a Washington Post headline reminded us that Sweden may have saved its economy, "But too many people have died."  

The story those sources tell is grim. "Sweden and the U.S.," Time declares, "are the only countries with high overall mortality rates that failed to rapidly reduce those numbers as the pandemic progressed." The Post warns that "deaths fell to low levels in August and early September but are now rising again." CBC tells us the Swedish health authority has been forced by the death toll to abandon its voluntary approach, and now "allows regions to institute local lockdowns 8 months into the pandemic" (although the story actually clarifies that these new local measures are not lockdowns: "there will be no legal or financial consequences for non-compliance"). Yahoo News quotes a "group of Swedish health and science researchers and professionals, calling itself Vetenskapsforum (Science Forum) COVID-19," as claiming: "Sweden is actually today among the highest countries in the world when it comes to deaths per capita from COVID-19." Scary claims indeed; but do the numbers back them up? 

Time reports (and CBC repeats) that "As of Oct. 13, Sweden’s per capita death rate is 58.4 per 100,000 people, according to Johns Hopkins University data, 12th highest in the world (not including tiny Andorra and San Marino)." But that is not exactly news in October; we already knew six months ago that Stockholm (like Milan, London, New York, and Montreal) would experience high deaths. All five cities, being international air hubs, were hit hard at the outset, and their seniors were not adequately protected at the time. In addition (unlike the other four), Sweden did not lock down, which scientists using Neil Ferguson's Imperial death model confidently predicted would lead to "a median mortality of 96,000 (95% CI 52,000 to 183,000)" dead Swedes by June.  

Sweden's actual COVID death toll by June - more than 5,000 - was tragically high. But things have changed there since. The virus is still active: as per Worldometer's dashboard, there were 58,000 total cases of COVID in Sweden by the end of spring (June 20), and 106,000 four months later (October 19). In the same four months, though, there have been only 618 deaths with COVID, an average of 5 per day. So what is the evidence that the health authority "failed to rapidly reduce those [mortality] numbers?"

Time's evidence consists of "a study published Oct. 12 in the Journal of the American Medical Association, which pointed out that, of the countries the researchers investigated, Sweden and the U.S. essentially make up a category of two: they are the only countries with high overall mortality rates that failed to rapidly reduce those numbers as the pandemic progressed." Time illustrates that claim with a chart (which repeats the claim, but without the 'rapidly' qualification): 

While the JAMA article and Time chart may be new, the data presented is old (perhaps due to the time needed for peer review) and outdated; the numbers go up just partly into September. I decided to look at the complete data, past and present, for myself. I used Worldometers' coronavirus dashboard for Sweden for the death total on the last day of each month, and then used simple subtraction to calculate the mortality for each month. Finally I divided each monthly total by 10.099 to calculate the month's mortality rate per million. 


As I hope my table shows, Sweden has in fact consistently and significantly lowered the mortality rate with COVID in every month of the pandemic since April (ultimately by 98%,). The authors of the JAMA article may have concluded otherwise, but as there is no discussion of Sweden in their study, that appears to be merely Time's spin on the results. In any case, that conclusion – that Sweden experienced a "failure to reduce coronavirus mortality rates as the pandemic progressed" – is falsified by the actual numbers.     

Admittedly, my table goes up only to the end of September; so it does not address the Post's October 20 claim that "deaths are now rising" or Science Forum COVID-19's finding (reported the same day) that "Sweden is actually today among the highest countries in the world when it comes to deaths per capita from COVID-19." So, what of those claims? 

Data for October is not complete, so the mortality rate for the month cannot be calculated; but it is still possible to look at Worldometers' daily averages. On September 30, the seven-day average of daily deaths with COVID in Sweden was 2 (two). On October 20 (the "today" when both claims appeared), the seven-day average was 2 (two). Average daily deaths had not risen by October 20, and they were certainly not among the highest in the world. I rate both those claims as simply false. 

Sources (accessed Oct. 21, 2020)

Jasmine M Gardner, Lander Willem, Wouter van der Wijngaart, Shina Caroline Lynn Kamerlin, Nele Brusselaers, & Peter Kasson, "Intervention strategies against COVID-19 and their estimated impact on Swedish healthcare capacity," medrxiv, April 11, 2020.  doi: https://doi.org/10.1101/2020.04.11.20062133

Kelly Bjorklund & Andrew Ewing, "The Swedish COVID-19 Response Is a Disaster. It Shouldn’t Be a Model for the Rest of the World," Time, October 14, 2020. https://time.com/5899432/sweden-coronovirus-disaster/

"Swedish virologist says her country's COVID-19 strategy has failed, but nobody will admit it," As It Happens, CBC Radio, October 19, 2020. 

Melissa Rossi, "As White House eyes 'herd immunity,' Sweden's no-mask approach is failing to contain COVID-19," Yahoo News, October 20, 2020. 

Lars Comfors, "We saved our economy in Sweden. But too many people died," Washington Post, October 20, 2020. 

3 comments:

  1. Sweden is being criticised for its high death rate by the same liberal groups who praised the country for its open door policy for the wave of immigration in 2015/16 in sharp contrast with its Nordic neighbours who closed their borders. Yet what is not recognised is that these same immigrants, because of their precarious health conditions, have been and still are disprortionally reflected in Covid deaths. Remove these from the death statistics and Sweden doesn't look very different to its neighbours.

    ReplyDelete
  2. Here is my take. It's a bit different with the same conclusion.

    Here is the hypothesis of the Times article:
    ----------------------------
    The Swedish COVID-19 experiment of not implementing early and strong measures to safeguard the population has been hotly debated around the world, but at this point we can predict it is almost certain to result in a net failure in terms of death and suffering. As of Oct. 13, Sweden’s per capita death rate is 58.4 per 100,000 people, according to Johns Hopkins University data, 12th highest in the world (not including tiny Andorra and San Marino). But perhaps more striking are the findings of a study published Oct. 12 in the Journal of the American Medical Association, which pointed out that, of the countries the researchers investigated, Sweden and the U.S. essentially make up a category of two: they are the only countries with high overall mortality rates that have failed to rapidly reduce those numbers as the pandemic has progressed.
    -----------------------------
    I went immediately to the referenced study, read it, and was immediately confused. To start, outside of table data, the study makes no mention of Sweden at all, let alone grouping it with the US in any special category. That kind of invalidates the primary reference the Time authors were using to support their hypothesis: "But perhaps more striking are the findings of a study published Oct. 12 in the Journal of the American Medical Association, which pointed out that, of the countries the researchers investigated, Sweden and the U.S. essentially make up a category of two".
    The study points out no such thing at all. The study isn't very long and I read it three times AND searched for the term SWEDEN to make sure the only references were in the table data. The authors aren't off to a good start. It is the authors the Time article that must be looking at the data in the study, and drawing their own conclusions from it, quite independent from the study authors whose focus was entirely on something else. So, the study has nothing to do with the TIME hypothesis other than the fact they are lifting the data from the study.
    So, now I'm curious if they used the data they cherry picked from the study in the right way. This is where it really goes downhill fast. The data they picked was COVID-deaths / 100K population on three arbitrary dates. Then they go and compare the countries listed and claim Sweden and the US have the least improvement over those dates.
    So many problems.
    First, why in the world would you use straightforward COVID-19 death rates from a study which was looking at all-cause mortality compared to the US? Why not pick that very simple data from the source?
    Second, the data they picked from was not overall mortality rates as they claim, but the COVID-19 mortality rate. The study does talk about overall mortality rates, but that isn't the data they used.
    Third, if you're going to compare how well a country did WRT to deaths over time, you're going to compare the deaths from the time the pandemic started in that country, not 3 arbitrary fixed dates in time!
    I attached a screen shot from the Financial Times comparing country deaths/m population over time NORMALIZED time scale to when the pandemic started in that country. You can see very clearly, that although Sweden had a slower down slope they don't have the continuation of the second wave every single other country is currently suffering from.
    Fourth, if they wanted to do a meaningful comparison of overall all-cause mortality rates across countries, they need to normalize the data to the demographics or more meaningfully compare the countries to their own historical overall mortality. In the latter case, I've published many graphs showing that Sweden's overall mortality rate for this year didn't change their 5 year average so it is well within historical norms. The US is in a class of it's own, as you have to go back almost 50 years to find a death rate so high.

    ReplyDelete
  3. Now, I'm starting to wonder who the authors are, because this looks like a high school level mistake. You don't even need to study Stats 101 to see the problems with this. It turns out the authors are both Swedish, and neither one seems to have a background in math, one being a journalist, and the other a molecular biologist. It looks to me like two Swedes with an axe to grind. Maybe they saw the new study, scanned it and saw Sweden and the US as the only countries with a double digit number in column 5 (see my screen shot of the study table)? Maybe they then decided to create a pseudo-scientific hypothesis and link to the study to fool people into thinking their assertion was backed up by scientific study?
    Even if the molecular biologist and chemist hasn't studied math since first year, that person should know better than to cherry pick numbers without any thought at all and pretend that a study says something it doesn't. The fact this is a professor writing this is scary.
    Time magazine actually published this. What a joke.
    I lifted the multi-country comparison from Matt Strauss' twitter feed here. His caption: Awkward
    https://twitter.com/strauss_matt/status/1319666353335373826

    ReplyDelete