Wednesday, April 22, 2020

Undercounting the COVID-19 Deaths


The excess number of deaths that are not explicitly being attributed to SARS-CoV-2 (coronavirus) has become mainstream news recently. The New York Times recently put out a series of graphs that look at localities that provide their COVID-19 death statistics in comparison to the average number of deaths in the same time period over the previous five years. Those five years include the late winter-early spring of 2018, when a particularly lethal flu virus dominated.

The U.S. deaths for flu for the 2017-18 season was initially calculated at 79,400 and then reduced to 61,000. The former number would have made it the worst flu season since modern tracking techniques in the mid-1970s and even the adjusted numbers made it one of the worst. Due to the 2017-18 bad flu season, in the initial parts of 2020 the number of deaths showed a slight decline when averaged out over previous years.

The United Kingdom provides mortality statistics, including those deaths due to coronavirus weekly, with about a two week delay. These numbers are gathered separately for England and Wales (combined), Scotland, and Northern Ireland. I present these below as a peek into both the number of increased deaths attributed to COVID-19 and the number not attributed.

I updated these tables using the figures most recently available as of 4/29/2020.

England and Wales.

First, England and Wales which have a combined population of 59,115,800 (2018). The raw numbers came from the UK Office of National Statistics, England and Wales .


*The initial period of deaths are presented as average deaths per week. The determination of excess deaths not classified as COV-19 (1 - (column 2/(column 4 - column 3)) as percentage) is performed the two most recent weeks when total deaths rise above noise. An excess death figure of 100% means that the increased deaths over the week were twice what was attributed to COVID-19.

England and Wales are having a considerable degree worse time with COVID-19 than Scotland or Northern Ireland, below.

Scotland

Here are the numbers for Scotland which has a population of 5,438,100 (2018). Here the raw numbers are from the UK official site for Scotland. Note: Scotland does their counting using Week Beginning rather than Week Ending and start on a different weekday.


Northern Ireland.

Northern Ireland has a population of 1,881,600 (2018). The raw numbers came from the UK official site for Northern Ireland.


The Scotland and the Northern Ireland figures are not included in the New York Times article cited above. I used the figures for only the most recent weeks in estimating the undercounting of COV-19 deaths. In each instance, the undercounting occurred to a much lesser degree over subsequent weeks and, in the most recent weeks, there was even overcounting.

Previously:
SARS-CoV-2 Infection and Mortality Rates
The Coronavirus: Potential Treatments and Drugs to Potentially Avoid

Thursday, April 16, 2020

SARS-CoV-2 Infection and Mortality Rates

COVID-19 is the name given to the disease caused by the SARS-CoV-2 virus (commonly called coronavirus). SARS-CoV-2 is listed in a recent article by LiveScience.com on the viruses most deadly to humans, not because it has the highest rate of mortality, but because it rapidly becomes widespread. The mortality rate given in the article is 2.3%, but open to conjecture.

Worldwide numbers for infections by SARS-CoV-2 have topped 2 million but are likely to be a gross underestimation. The politics of 200 different countries, the rates in which test for infections take place, how countries they define their positive cases, and how they define deaths by COVID-19 are not only variable, they are confounding. For example, Chile counts those who have died due to SARS-CoV-2 virus as being recovered. They are no longer infectious.

This leads to two important issues are: how many people have SARS-CoV-2 but are either untested or asymptomatic; and, how many people die of SARS-CoV-2 but are not listed among the total deaths.

How Many Total Infections?


In regards to the first question, an interesting observation was reported in the New England Journal of Medicine. One New York City hospital tested all of the women (215) admitted for childbirth and found an infection rate of 15.4%. Of these, 87.9% were asymptomatic. Several factors suggest that late-term pregnant women may not be representative of infection rate in the population as a whole, but nevertheless, this is an intriguing number comparing infections to those asymptomatic.

This report states the United States is detecting only 9% of the coronavirus infections.

Such reports are political fodder and it is difficult to be sure if the input is unbiased. One model generated by Chicago economists says that 1.5 to 14% of the infections have been detected depending on lag time involved in detecting the infections.

Why political fodder? If the risk from SARS-CoV-2 is much smaller than reported, then the number infected (and perhaps immunity) is much higher. In such a case the risk of opening the economy is lessened. Some are gung-ho to lift safety measures in place.

The ultimate risk of SARS-CoV-2 is both from its morbidity and mortality rates and its degree of infection. A virus that causes a 0.5% death rate (low end) and saturates at 70% of people infected (high end estimates), would kill 25 million worldwide along with a million Americans. A similar number would come from a 1.5% mortality rate and 23% infection rate.

What We Can Learn from the Diamond Princess Cruise.


The Diamond Princess cruise ship is instructional when trying to determine unknowns regarding SARS-CoV-2 infection rates. I agree with this quote in the scientific journal, Nature. “Cruise ships are like an ideal experiment of a closed population. You know exactly who is there and at risk and you can measure everyone.” John Ioannidis, an epidemiologist at Stanford University in California.

According to the CDC's March 27, 2020 report, the ship set sail on January 20th from Yokohama with 3711 passengers and crew. When stopping in Hong Kong on January 25th, one passenger disembarked due to being symptomatic. This passenger was later confirmed to have SARS-CoV-2.

The crew averaged 36 years of age and the passengers 69 for a combined average age of 59.6. Due to concerns regarding SARS-CoV-2, those on the ship were not allowed to disembark when it docked in Japan on February 3rd. On February 5th passengers were quarantined in their cabins.

According to the Nature report, over 3,000 tests were run on those aboard the Diamond Princess, with some tested more than once. This says that the testing was not comprehensive. A total of 712 tested positive. Back to the CDC report: Of those testing positive, 46.5% had no symptoms at the time of testing. Furthermore, and more revealing, it was estimated that 17.9% never developed symptoms.

At the time of the CDC and Nature reports, 9 of those aboard the Diamond Princess had died. I have found more recent figures place the number of deaths at 12, these from Sydney Morning Herald and Wikipedia. A higher figure makes sense: the previously mentioned articles spoke of some of those from the Diamond Princess as being in critical condition. Using the figure of 712 testing positive this has a mortality rate of 1.3 to 1.7%. On the one hand, the cruise ship did skew toward elderly passengers. On the other hand, the attendant publicity likely ensured top quality care for those who were aboard.

There are other cases of cruise ship outbreaks with some continuing to be in quarantine.

Excess Deaths in Nembro, Italy (citation) The red line is the number of deaths in the time period in 2020. The green line is the number of official COVID-19 deaths (all 2020). The blue line is the average number of deaths in the same time period but from 2015-19.

Undercounting Mortality


To what degree are deaths due to COVID-19 undercounted? One way in which this can be addressed is by looking at the excess number of deaths taking place in a location where SARS-CoV-19 has become prominent in comparison to previous years. A number of studies have been made.

One is a look at the mortality statistics collected in England and Wales. This is not a small population sampling: together they have a population of 60 million. The following table lists the number of deaths attributed to COVID-19 and compares those to the increased number of deaths. England began its lockdown on March 23rd. The table below is my creation taken from the aforementioned dataset.

Weekly deaths in England and Wales
 The small decrease in the early part of 2020 may be due to a particular virulent strain of flu virus dominant in 2017-2018.

 For the week ending April 3rd, those deaths officially deemed to be due to COVID-19 accounted for only 57% of the increase in deaths (that is a total of 33.6% of the 59% increase).

Because the report notes that the final week's cause of death may be incomplete, it is worth noting that for the week ending March 27th, the official COVID-19 deaths account for only 53% of the increase (5.3% of the 10%). It might also be noted that a 59% increase in death does not correspond to a virus that was as lethal as the annual flu. The annual flu had been around in previous years.

Two factors may contribute to the total deaths figure. One is an increase in deaths due to other causes due to, for example, an overwhelmed health care system. The other could be an underestimation due to the decrease in certain deaths such as traffic fatalities due to the lockdown conditions.

 According to the UK Office for National Statistics, these data will be updated on April 21st. Considering the rapid rise in the number of weekly deaths, that update will be crucial to providing an even better picture.

This phenomenon is hardly limited to England and Wales. This report from Netherlands analyzes increases in deaths compared to previous years and find that the official COVID-19 numbers only account for half of the deaths.

Similarly, from Spain, a study described in El Pais, says that while the official number of COVID-19 deaths was 3439 during the time period of March 14th through March 31st, there was an increase in total deaths of 6613 when compared to the same time period in the previous year.

This news report describes several U.S. cities as having significant undercounting.

What's the takeaway from this? First of all, although there are a variety of models that come to different conclusions, there is some hard evidence out there about the undercounting of infections and deaths. The very fact of the spikes in deaths exist indicates that SARS-CoV-2 has a much higher lethality rate than a typical influenza virus. Deaths are probably double the official figures in places where deaths are meticulously counted. In other parts of the world, where the government either wishes to lowball the pandemic or else just doesn't count, the statistics on death are meaningless.

Previously: The Coronavirus: Potential Treatments and Drugs to Potentially Avoid
Next: Undercounting COVID-19 Deaths.