COVID19 update, April 3, 2020: what does Iceland‘s unique dataset tell us?

Iceland is a small island nation with a thriving biotech industry. As such, it adopted a (to my knowledge) somewhat unique strategy: aside from contact tracing measure, local biotech company deCODE (an AmGen subsidiary) allows anyone who shows up to be tested. (This type of “big data” gathering seems to be part of their business model: they also did genetic testing for about half the adult population in the past )

As of the time of writing of this article about 20,930 people, i.e., 6% of the total population got tested — by far the highest proportion of tests in the population anywhere. (For perspective: to reach the same amount of coverage in the US you’d need to test about 20 million people. The US has tested about 1/20th of that number. )

As of this morning, Iceland has 1,319 cases, of which 4 (four) people have died, 44 people are hospitalized (12 of them in ICU), and 1,031 are in home isolation. 284 have recovered from the infection. 7,166 people are currently in quarantine, 8,945 have completed quarantines.

Now beware of comparing apples with oranges: the official government COVID19 page for Iceland speaks of 1,319 infections, not sick people. (It appears Israel similarly counts everybody who tests positive as a patient.) So the apparent enviably low 4/1319*100%=0.3% is really an IFR (infection fatality rate), not a CFR (case fatality rate).

[Iceland’s chief epidemiologist Thorolfur] Guðnason is quoted here as saying “Early results from deCode Genetics indicate that a low proportion of the general population has contracted the virus and that about half of those who tested positive are non-symptomatic.”.

Note, Iceland did not engage in any lockdowns — gatherings over 25 are forbidden, but otherwise economic life goes on. Instead, aggressive test & trace and quarantines are the order of the day. While the country’s low population density might seem to help, nearly two-thirds of the population is concentrated in and around the capital Reykjavik — which especially by US standards is fairly dense.

Elsewhere I read that not only was Iceland able to trace the source of the infection in most cases, but even to distinguish between different mutants of the virus.

“We can determine the geographic origin of the virus in every single [virus] in Iceland,” he said, adding there are specific, minor mutations for the virus that came from Italy, Austria and the UK. “There was one that is specific to the west coast of the United States,” [Stefansson] added.

[He] wonders whether mutations in the virus are “responsible, in some way, for how differently people respond to it — some just develop a mild cold, while some people need a respirator,” or whether a person’s genetics dictates their condition. “Or is it a combination of these two?” he asks.

“Keep in mind that the screening is now randomized, but voluntary so there is some bias in the data,” the Directorate of Health said in a statement, adding that a “randomized screening program has started and a blood serum screening for antibodies is planned.

COVID19: March 21, 2020 update

(1) Israel saw its first death from the epidemic, an 88-year old Shoah survivor from Jerusalem named Aryeh Even. May his memory be for a blessing.

The total number of patients reached 883. Despite “soft lockdown”, people still went out to the beach and to national parks.

(2) Neuroscientist and sci-fi author Robert E. Hampson: reports on an Icelandic biotech company has been organizing a free testing program running in parallel with the official one, and where everybody can ask to be tested. Presumably the purpose is to gather better data about what the small island nation is really dealing with. (At about 335,000, they can in principle test their entire population.)

The biggest takeaway of the results: about one-half of those who tested positive never got ill at all — a number congruent with what is seen on the Diamond Princess.

(2b) From the same post, a preprint on MedRXiv (DOI 10.1101/2020.03.05.20031773 ) of a detailed epidemiological analysis of the Diamond Princess data and of the China data. The central result: they estimate an IFR (infection fatality rate, i.e., dead per number of people infected) of 0.5% (one-half percent), with a 95% confidence interval running from 0.2 to 1.2%, and a CFR (case fatality rate, i.e., dead per number of people who actually get sick) of 1.1%, with a 95% confidence interval running from 0.3 to 2.4%. I certainly hope it’s closer to the bottom than to the top of these CIs.

Israeli epidemiologist Dan Yamin, interviewed here in Haaretz (archive link) argues for low figures as well, and (very much to the distaste of Haaretz’s small, leftist readership) gives Trump’s response positive marks and the WHO’s negative marks. He also stresses what I’ve been suspecting: that anomalously high apparent CFRs in some countries (notably Italy) reflect a selection bias in testing, with scarce testing resources being applied primarily to cases with an already probable COVID19 diagnosis. (Link via Jordan Schachtel.)

We need more and faster testing, and we need it yesterday.

Happy first day of spring, stay healthy, stay safe, stay calm.

Addendum: somebody forwarded a graph highlighting the flatter case incidence graphs for countries like Taiwan, South Korea, Japan, as well as for Hong Kong. Correlation is not causation, of course, and particularly the Japanese by default practice a considerable degree of social distancing on the micro level, but in all of these places, wearing face masks when one has even the slightest hint of a respiratory infection is common social custom. Wearing a mask prophylactically when one is not feeling ill: somebody who lived in Japan for a long time told me people won’t routinely do it except in flu season, but it is definitely socially acceptable.

Addendum 2: The Johns Hopkins COVID19STATS dataset is online here.

Addendum 3: HonestReporting produced this brief memorial video for Aryeh Even z”l.