[Update 29th March 2021: Explanation offered in a later post]
GameStop's down (though one wonders what the story would have been, had they gone just twenty-fold from about three bucks to the current sixty, instead of breaching US$350 in between) and vaccinations are up, if not exactly fast enough for the public health mavens. That latest news is that the pandemic will stretch for another seven years at current rates of vaccination (though that may not be particularly appropriate framing, as noted by some commentators), with the co-chair of the local COVID-19 taskforce having already put out an estimate of "four to five years" a week or so back. From this, it seems that we may well have to learn to just live with it in some capacity eventually, all the more given reported resurgences in areas with high reported seroprevalence, which we will get back to soon.
To be very frank, the professional response to the coronavirus continues to be not entirely reassuring; on face masks, for instance, we've come from mask effectiveness bordering on a "conspiracy theory", to Dr. Fauci suddenly advising two masks to be worn, to which some Virginia Tech researchers figured that if two was good, wouldn't three be better? This probably pales to the horrible mismanagement on the E.U.'s end, which however might not be getting much airtime in the mainstream news, because no Orange Man was involved. For all the platitudes in the press, nationalism in whatever form appears to only be getting stronger, with countries seemingly smearing and choosing suppliers based on geopolitics - which however is probably nothing out of the ordinary. On this, The State's Times has published a commentary last week on how our initial vaccine purchases were selected, from a shortlist of over 40 companies. While no mention of geopolitical considerations was made, I sure hope that such a possibility is at least thinkable.
Closer to home, while insurance and financial assistance programmes have been introduced to allay fears of vaccine complications, these were possibly not exactly helped by the realization that it seems almost impossible to prove conclusively that any symptoms are due to the vaccines, from a recent case, i.e. future consequences are only more likely to be disclaimed. Well, that and unfortunate human error in application would appear to be relatively minor brouhahas compared to the fog of war at the broader policy level, as possibly admitted by the PM in his (rational) push for global coronavirus test and vaccination standards, and perhaps implicitly by several reciprocal green lane agreements being abruptly suspended. As suggested last October, actual realities may or may not bear much resemblance to what's being painted by the mostly-sensational corporate media blithely (and oft incorrectly) wielding "science!" to club down dissent - and towards this, let us attempt some brief data analysis:
I hope that it remains uncontroversial to state that the performance of various countries towards the coronavirus should be objectively measured and discussed as far as possible, if only to guide future pandemic responses. On this, one of the strangest habits of the corporate media has been their fixation on absolute case numbers, at least in the American context. One particularly dodgy cable news network immediately removing their scary COVID-19 death counter upon the accession of a new POTUS might be an example of how such information has been nakedly politicized - did such data become irrelevant overnight? - which seriously hardly helps to promote trust in whatever latest skewed angle's being sold on the talkies.
While American data on the coronavirus remains mired in inconsistencies, as reported by the Wall Street Journal, it's probably fair to say that it's hardly any better most elsewhere, what with the many powerful stakeholders having vested interests in either downplaying or exaggerating the effects of the pandemic. To this, recall that excess mortality had been consistently proposed as one of the more-foolproof metrics available to gauge the coronavirus' impact. As realized by the PM, standards are all over the place, and with criteria for coronavirus deaths ranging anywhere from "all deaths with any hint of the virus" to "fluid-filled lungs on a full moon" (only slightly embellished), it's honestly difficult to draw any meaningful conclusions about policy efficacy without a lot of adjustment (not that that's stopped the FAKE NEWS)
Excess mortality, remember, gives the extra deaths for a period of time, over what would have been expected normally from past records. While this admittedly does not assure that the extra deaths arose directly from some cause (i.e. the coronavirus), there is an argument to be made that it's maybe more useful to consider all related deaths, even if indirect. For example, if a state pursues a total lockdown policy for months, and thereby reduces coronavirus deaths to near-zero while also causing deaths from other diseases, suicides, starvation etc. to skyrocket, such costs should probably also be taken into account. As such, I had been eagerly awaiting a more complete study based on excess mortality for the past months, with the previous one having been centered on America and Europe.
Excess and Covid-related deaths, example subset
[N.B. Full set comprises fifty-nine countries]
On this, the WSJ has released just such a more-global report a couple of weeks or so back, though the raw numbers (obtained mostly from various national statistical agencies, as explained at the end of the article) are subsumed behind some very natty infographics. Well, not a problem - a critical if less-recognized part of the data science toolset is the extraction of figures from processed formats, after all, which is where we will begin.
The excess (solid colour) and Covid-related (dashed lines) deaths for each country are given as semicircles, with the scales provided at the top right for 10,000, 50,000 and 100,000 deaths. It could be ascertained from the scales, that the number of deaths represented corresponds with the area of the semicircles. Then, since the area of each semicircle derives from its radius and thus diameter, a multiplier between the (square of the) radius and represented deaths, could be determined. The multiplier of 18.262 calculated from the largest semicircle representing 100k deaths was used, as it fell more or less in between the estimates from the smaller semicircles.
The next and slightly more-involved step then involved measuring the diameters (in pixels) for each of the two semicircles, representing the excess (ExPx) and Covid-related (CovPx) deaths reported, for each country. Of course, this is likely to be more accurate for larger semicircles, due to uncertainties at the edges as to whether some pixel should be included. From these figures, the corresponding absolute total excess death (ExDeath) and Covid-related death (CovDeath) numbers could be obtained. This process returns about 470k total excess deaths and 279k excess deaths attributed to Covid for the U.S., for example, which does seem close to what's been reported in other media.
However, as mentioned, absolute numbers are probably not the most meaningful metric to reflect a country's performance, with excess deaths as a percentage of expected annual deaths seemingly a rather more useful measure. To estimate this, country annual death statistics for 2019 (AnnDeath) were extracted from the UN's World Mortality 2019 data booklet (the exception being Liechtenstein [population: 38,378 from Eurostat], who were unlisted; their annual death figure of 280 was thus derived from the annual death rate of 7.2 per 1,000, from 2018). From this, the total excess deaths as a percentage of annual deaths (ExDPer) and Covid-related excess deaths as a percentage of annual deaths (CovDPer) could be computed.
One more detail might be taken into consideration: the original death figures were reported at different end-dates for each country, as far as could be interpreted (e.g. "as of Dec 5" for the U.S., Dec 12 for Mexico, Nov 21 for Brazil, etc.); as such, it appears fair to adjust the ExDPer and CovDPer to represent the same time period of one full year, assuming a start date of Jan 1. This gives the adjusted total excess deaths as a percentage of annual deaths (S_ExDPer) and Covid-related excess deaths as a percentage of annual deaths (S_CovDPer) respectively. Admittedly, this does not make too much of a difference for many countries who reported theirs up to December or late November, though there are exceptions such as Georgia, who had an impressive negative excess death count on June 27.
What happened in Georgia after that, from around October onwards, was perhaps slightly less outstanding...
The spreadsheets with all the data described above are freely available, and also for convenience sorted by the S_ExDPer column, which represents the adjusted total excess deaths as a percentage of annual deaths - which one supposes should be the most-relevant metric of interest here. This figure is about 17% for the U.S., which is perhaps not too surprising, and it seems to indicate that Latin America has been hit the hardest by far, with Peru and Ecuador both over 50%, Mexico at 36%, Colombia at 21%, and Chile and Brazil not far behind - which has maybe not been emphasized too much in the news.
One might naturally be curious as to where Singapore stands, according to this methodology. As it turns out, we're smack dab in the middle: 32nd of 59 countries, with an S_ExDPer of 5%.
But wait, one might say, surely this is misinformation? Hasn't the government been lauded internationally for their exemplary coronavirus response (excepting foreign workers), with only 29 deaths officially attributed to the coronavirus thus far, the last back last November? Well, I present the numbers as they turn out, if that's what you're asking. Notably, with a diameter of 2 pixels estimated for Covid-related deaths in Singapore from the infographic, this translates to an absolute number of 18 deaths, which is both unavoidable from the image resolution, and not that far off. However, one would expect the estimate for total excess deaths - which is what's being discussed here - to be rather more accurate, with 15 pixels measured.
However, it wouldn't be a very good rankings list if we weren't top of it, and as it so happens, this is the case for a perhaps quite relevant metric - the ratio of (adjusted) total excess deaths, to total Covid-related deaths (S_ExCovRatio). From the data, Singapore is top by far with a value of 56.44, which indicates that for every about 56 deaths over what had been expected (which itself seems to be kinda average globally), only one was directly attributed to the coronavirus. South Korea's a distant second at 24.16, and Malaysia third at 8.23. Most Western countries appear to have a value between 1 and 2 (e.g. the U.S. at 1.69, the U.K. at 1.17, Spain at 1.55 and Italy at 1.42), with a number (e.g. Romania, Switzerland, Iceland, etc.) having a value of essentially 1 (i.e. all excess deaths were directly attributed to the coronavirus). Interestingly, some countries had values less than 1, which may suggest that they had been overeager to attribute deaths to the coronavirus; Sweden (0.56) is one such exemplar, given their policy of counting all deaths with the virus detected, as Covid deaths.
Selected country ratios
One caveat - three countries (Mongolia, Taiwan, Georgia) essentially had S_ExCovRatio undefined due to zero Covid-attributed deaths implied by the infographic (a Google search suggests two for Mongolia, and nine for Taiwan, and we've covered the incomplete data for Georgia), so we might not own the title of "having (by far) the most unexplained excess deaths, scaled to population" if their external figures are considered - which however might also not detract from the main question raised: it seems that a good five percent more of local residents had died last year than had been expected, but essentially none of them had been attributed to the coronavirus. If the above is true, what then exactly had these extra 5% perished of?
To further affirm these implications, one might consider the context - for example, the WSJ infographic has Japan reporting a large number of negative excess deaths, right at the bottom; isn't that even stranger? However, this happens to imply just eight thousand-plus negative excess deaths, which is basically a rounding error compared to Japan's annual mortality of some 1.36 million, i.e. Japan basically recorded more or less precisely their expected death rate in 2020, despite the pandemic. Definitely, the above analysis should be considered a brief sketch, and there are many valid limitations that might be appropriate to air, in a formal publication, for instance:
I'd be more than willing to entertain specific critiques along such lines; one supposes that there would be many respected research groups independently performing retrospective analyses on excess deaths in the near future, however, and assuming they begin from the same datasets, one would expect the conclusions drawn to be broadly similar. When that happens, it would, I don't know, probably be good for various health authorities to have their explanations and rebuttals ready for public consumption...
Next: Longish Rest
Copyright © 2006-2022 GLYS. All Rights Reserved.