Critical Info: Deaths in Spain in 2020

Everyone has heard that Spain was one of the countries hardest hit by Covid-19. Spain was also one of the first to impose the strictest of lockdowns, which transformed its busiest and most bustling city of Madrid into a deserted ghost town for two months from mid March to mid May. And in April it was reported that there were so many dead bodies that there was nowhere to put them such that temporary morgues had to be setup to accommodate them. What a nightmare.

At some point this year, official rules for assigning the cause of death were changed and autopsies became discouraged. This was done in several European countries. This probably simplified the work of many health care workers, but these changes have made the statistics relating to specific causes of death unreliable.

The one thing in life that is undeniable is death itself. We know for sure that when a person is dead, then they are no longer alive, and that this final verdict does not depend on the cause of death. It does not depend either on the real cause or on the recorded cause of death.

Total Number of Deaths

So, as obvious as this may sound—even though it has been conspicuously missing from public discourse—the only thing we can truly reliably count are the total number of deaths, because when a person is dead they have passed away, and we can entirely reliably count them as deceased.

In addition, if we are to make any comparison of the present situation with the past, if we are to look at trends over time, there is no better metric to compare a given year to other years in terms of where it stands with respect to the average and thus the expected number of deaths.

In Spain, where I have been living for 13 years, death statistics are compiled in real-time with monthly updates by the Ministry of Health (Ministerio de Sanidad). Here is the graph of the total number of deaths per year from all causes in Spain from 1987 to 2019.

(The dash-dotted line shows the linear least squares best fit. The dotted lines above and below it show the 2-sigma band computed from the variance of the de-trended data.)

We can see, for example, that in 1987 the number of deaths was around 310000, and we can clearly see that there is an overall trend of increasing numbers in time, even though there are visible fluctuations around the average trend line.

But it is surely natural to expect that in the absence of major societal disruptions, the number of deaths will be more or less proportional to the population. And also similarly natural to expect that if living conditions are poor people will leave, whereas if they are good people will be come. So, trends in the population have to be taken into account.

Here is the total number of deaths per year from all causes per million people in Spain from 1987 to 2019.

Hence, from the first plot we can see that last year, for example, in 2019, there were a total of around 420000 deaths, and from the second we can see that this corresponds to 8911 deaths per million people. And looking at overall long-term trend, we can see that these numbers are quite close to the expectation.

What about 2020

For 2020, we don’t yet have the final count. The last update was on Dec 17, and the next one will be in about 2 weeks from now. This means that we have 352 days of data and that therefore in order to compare to the numbers from previous years for which we have 365 days of data, it’s necessary to divide by the number of days. Here is the average number of deaths per day from all causes per million people in Spain from 1987 to 2020.

What does this mean

As far as the impact in Spain of the strict lockdowns, the teleworking for everyone, the social distancing and isolation, the mask-wearing, and all the other measures that have been imposed on everybody for the sake of minimising the Covid-19 death toll, I’m not sure what to say, and I won’t speculate.

As far as the data are concerned, we can say that for the current daily average in 2020 to reach the expectation based on the long-term trend, we need to reduce the number of days over which the data is stated to have been accumulated from 352 to 326. We need to assume 26 days of missing data. This could be the case.

The yearly totals in 2016, 2017, 2018, and 2019 were 409358, 422130, 427956, and 416480. For 2020 (based on a population of 46.75 million), the long-term trend yields an expectation of 410440 deaths. The value we would calculate based on the current estimates is 379686. This represents 30754 missing deaths from the total. This could well be the case.

However, if we take the data at face value, according to these official numbers from the Spanish Ministry of Health of which there are currently 10 updates, the conclusion would be that 2020 in Spain has not been a year plagued by an overwhelmingly large number of deaths. On the contrary, Spain has been blessed with the lowest death rate in almost 20 years!

Nevertheless, it could also very simply mean, that the numbers published are still missing about 31 thousand deaths to reach the expectation, and surely a lot more than that to reach what we expect from the impression we all would have gathered based on the reports from the majority of media outlets to be a massive death toll. The fact is, that we’ll just need to wait a couple of months in order to have the final numbers.

Update from Feb 6 2021

The Ministry of Health published an updated total for 2020 on January 22. Given that we now cover all of 2020, we don’t need to average over the number of days used to estimate the total, but can simply look at the total per million people. Here is what this plot looks like:

The total number of deaths for 2020 is now stated 401359. This corresponds to 8584 deaths per million people, and it is below what could be expected from the long-term least squares fit shown as the dash-dotted line. Let’s wait another month to see if the total number settles.

17 thoughts on “Critical Info: Deaths in Spain in 2020

  1. Very good, no? However, the main reason for the measures taken (in the Netherlands at least) is to avoid saturation of hospitals, and not to limit the number of deaths. Also, thanks to the measures in place, chances to die of other causes than Covid seem to have diminished as well (for example, there are much less influenza deaths thanks to these measures)

    Like

    • I think that in order to make sense of this, we would need to look in great detail at all the data from all causes and at higher resolution, which is a massive amount of work. The main point was in fact to contrast what we might have expected with what is actually the case. From everything that has happened and everything we have been hearing since March, everyone would naturally conclude that we’ve had a massive increase in the total number of deaths given that we would have, I think correctly imagined, that we should expect all regular deaths as in previous years PLUS all the deaths due to Covid-19. But this is NOT what we see in these data. One likely scenario to explain the Covid-19 deaths in light of the lower overall mortality is that the vast majority of people who died were so ill and weakened that they were going to die of other causes anyway, and they just happened to die of the Covid-19 instead of something else. But for sure, in order to draw more detailed conclusions, one would have to invest a lot more time in data mining.

      Like

  2. Are you sure the Plot 3 tells you what you think it does? Because you just scaled the Plot 2 by 365 (I guess, because the plots look identical, except for the scaling), so you are still looking at annual average, not daily. To see the daily death rate, you should count and plot the deaths for each day. Then it’s quite possible the data would show that many of those deaths happened in a very short time, i.e. during the initial weeks of the first and the second wave. I guess this is why the morgues were full. But if you average them over one year, nothing pops up. What I’m trying to say is that while the annual average death rate might be the lowest in the last 20 years, many of those deaths likely happened (twice) in a very short time, which created problems in morgues. But to know it, we should see a plot of the daily (or at least weekly) death rate.

    Like

    • I have no doubt that what you describe is probably right. But the main point is exactly that: looking at the whole year, it is exceptionally low, which I would have never expected. Indeed, I like probably everyone else, would expect to see much higher mortality in 2020 given the news and information we’ve been getting every day now since March.

      Like

  3. Hola Guillaume, I have to disagree with your analysis, as I am afraid it is based on provisional data. I recommend you to read this article: https://factual.afp.com/los-datos-de-defunciones-en-espana-en-2020-no-son-aun-definitivos , where you can see how the actual number of deaths in 2020 will not be registered until several months into 2021. How otherwise you can explain the excess number of deaths shown in this web site which tracks more faithfully the impact of COVID19 in Spain in real time: https://momo.isciii.es/public/momo/dashboard/momo_dashboard.html ?

    Like

    • Hi Pedro, yes, we’ll have to wait a month or two more to have all the final numbers. However, you will agree that the closer we get to the final, then the closer we get to the final count ;) The excess death rates are based on an expectation value computed as an average over many decades. It is useful to look at data in real-time and have a reference to compare against. But in the end the thing that matters the most is the total number of deaths. And in this case, I think what these data indicate, even if this would have to be verified with further analyses, is that based on the total number of deaths, it would seem like the people who died of covid-19 would have died of something else within a very short period of time. Hence, even if there is peak of excess deaths in March-April, overall the year’s total is less than expected by quite a bit, for whatever reason.

      Like

      • If you read carefully the information provided at: https://factual.afp.com/los-datos-de-defunciones-en-espana-en-2020-no-son-aun-definitivos you will easily see that the decalage in the statistics is extremely important and a fundamental factor to take into account to perform any valid analysis. In the above link you can see, as an example of what happened in previous years, that the number of deaths published by INE by 20-Nov-2019 were only 266,413 people when the final number of deaths in 2019 reached 416,602 ! This means more than 150,000 deaths to add to the provisional numbers published in late November. (more than 30% of deaths in 2019 were still not recorded by end November). With a similar huge correction factor to apply to the numbers published so far for 2020 I am sure you will easily reach a number very close to the 70,000 excess deaths that have been recorded by MoMo in 2020, which gives a much more realistic and close to real time accounting. The plots shown by MoMo at https://momo.isciii.es/public/momo/dashboard/momo_dashboard.html speak by themselves. As expected, the two excess peaks are concentrated in the periods 10/03/2020 to 09/05/2020 (~44,600 excess deaths) – first wave of the pandemia – and 20/07/2020 al 20/12/2020 (~26,000 excess deaths) in the second wave. And for the rest of the year the number of deaths recorded were very closely matching those of previous years. In any case, we just need to wait a few months to confirm the final numbers….

        Liked by 1 person

      • One extra piece of information, which may help understanding the problem. As an example, if you compare data from 2019 and 2020 published by INDEF in similar periods, you can easily conclude that the excess deaths in 2020 is, as expected, considerable. In particular, the number of deaths reported by INDEF on 29 July 2020 reached 181,777, while on 26 July 2019, the reported number was just 141,299, this is 40,478 less than in 2020, not surprisingly, very close to the 44,000 excess deaths attributed to the first wave of the COVID-19 pandemia.

        Like

  4. Hi Guillaume, thank you for this analysis.
    I agree with you that interpreting the Covid-19 data has become increasingly difficult and the only really reliable number is number of deaths (and maybe people in ICUs as well).
    It does surprise me though that Spain would not have any excess deaths due to Covid19. I was checking data from EUROMOMO that keep track of death rates in many European countries and plot excess mortality, and in their graphs Spain does have 2 very big peaks of excess mortality during 2020: see https://www.euromomo.eu/graphs-and-maps towards the bottom of the page. There is a small dip below the expected in summer and also at the end of the year, but that does not seem to compensate at all.
    Do you have an idea what could cause the discrepancy?

    Like

    • Hi Anik, yes, I’ve also been looking at Euromomo since March. Excess mortality is the number of deaths above the expectation based on several decades of data, somehow scaled to the current population. So, we cannot get from this what the actual total number of deaths is. It shows that in March-April there were a lot more deaths than expected based on previous years, but in the end what we are really interested in is the total. Now that the year has come to an end, it shouldn’t be long before we have the very final count. So, when we have it, in a month or so, I strongly suspect it that the total averaged on the number of days, as shown in my last plot, will be basically the same as we have now.

      Like

  5. I simply repeat here the comment I made on January 8, 2021. It is well known, simply by looking at when the final data was eventually available in previous years that in order to have the statistics right from INE you will have to wait for the updated numbers that will be provided SEVERAL MONTHS INTO 2021. Still the numbers provided in January are provisional. You just need to look at the numbers provided e.g. for 2020 on 24 March this year if you take the same source of information and you will find that on this date only 77 deaths were reported! which is clearly telling you that numbers are usually final only around 3 months into the year. But I don’t understand why you insist on using the numbers provided by INE and not the ones provided by https://momo.isciii.es/public/momo/dashboard/momo_dashboard.html where the plots provided speak by themselves. Are you a kind of negationist?

    Like

    • Hi Pedro, I actually started working on more realistic and reliable estimate of what the total is most likely to end up being based on the evolution of the total number for 2019 over the course of 2020, but I haven’t yet finished. I’ll post the update when it’s ready. And I don’t consider myself a “negationist”, (which I’m not sure is an accepted word in english as of yet). I’m simply a “seeker-of-evidence-based-truthist” ;)

      Like

  6. Last update: 23 March 2021
    … and finally, the total number of deaths in Spain in 2020 reaches 483,237 people according to the government statistics, confirming the predicted excess of around 60,000-70,000 deaths with respect to previous years, perfectly alligned with the MoMo statistics, as I anticipated to happen. I think you can now safely update the blog entry (or publish a new one?) with the final numbers and new plots and the necessary revised analysis as they totally change the initial conclusions derived from your preliminary analysis and will avoid misinterpretation by the negationist world.

    Source: https://www.mscbs.gob.es/estadEstudios/estadisticas/docs/indNacDefunciones/2021_Defunciones_1.pdf

    Like

Comments?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s