COVID-19/All-cause deaths

This learning resource is about all-cause deaths under the learning resource COVID-19.

  • Aggregated data of all causes of death add additional noise to the COVID-19 data,
  • if you want to analyze and assess the developement of a specific cause of death it is required to analyze the development of data for the specifc cause of death. Because scientifically you try to remove the noise from the data instead of adding noise to the data for COVID-19 directly.
  • Death is a very late indicator for epidemiological response activities. In epidemiology for infectious diseases it is the key to act as early as possible (containment of the disease, avoid lockdowns) and not wait until the even the aggregated mortality of all-cause deaths shows a significant increase.
  • the disease control tries to flatten the curve that the health system is able to deal with the number of cases. The heath service delivery is reponsible for more than one disease so an extreme increase of COVID-19 treatments in hospital assigns medical staff and resources to COVID-19 treatment that might be necessary in other areas. So early detection and early interventions for COVID-19 are required.
  • Nevertheless is makes sense to keep an eye at other causes of deaths that COVID-19 might have an impact on.
    • reduced Influenza caused death due to risk mitigations strategies for COVID-19,
    • impact on traffic accidents,
    • impact on suicides,
    • impact on drug abuse,
    • long term impact on death cases due to postponed treatments,
    • ...
In that case all-death cases can be rough indicator. E.g. if COVID-19 caused death can be controlled at the current level, but we see a significant increase in all-cause deaths dispite the disease control. Again the significant increase dispite the disease control activities must be clarified by looking into the specific causes of death.

Source of the DataEdit

What follows are all-cause death charts for multiple countries, most of which were created from data from Human Mortality Database (HMD), mortality.org[1], a collaboration between organizations from Germany, the U.S., and France: Max Planck Institute for Demographic Research; University of California, Berkeley and INED, Paris. Some charts were created from other sources.


All-cause death charts help answer the following questions, region by region:

  • What elevation above the normal mortality could be observed during the covid pandemic? (Not all elevation in necessarily caused directly by the covid infection.)
  • COVID-19 non-pharmaceutical risk mitigation strategies are not virus-specific, so even mortality caused e.g. by influenza may be reduced in comparison to previous years. What are the changes in general on the specific causes of death that are aggregated in all-cause death?
  • How rare are the peaks in mortality in 2020 compared to other years? (Peaks in mortality seem at least somewhat indicative of peak hospitalizations and the associated healthcare overload.)
  • Could it be that the great majority of covid-coded deaths are merely covid-positive deaths rather than covid-caused death?< (This is not entirely easy to get right: elevated deaths can in principle be contributed to by lockdowns and by improper treatment that causes more harm than good.)

There are repetitive disclaimers under the charts about the data being preliminary.

Registration delay: The data from 2020 suffer from registration delay. The last two weeks suffer especially badly, but other weeks are also not free from the registration delay effect.

Scripting to create these charts are at /Scripts.

The number of years covered varies, depending on how much data HMD was able to obtain from the countries. Some data starts in 2000. The more years included, the stronger conclusion one can make about how rare the covid death elevation in 2020 is compared to other years.

Learning TasksEdit

The learning tasks is a stastical learning task of aggregated data and disease specific data. If you want to follow current developement of COVID-19 follow the actual number of cases and the recommended risk mitigation strategies of public health agencies, scientific results of centres for epidemiology and disease control with the epidemiological expertise.

Now we analyze all causes of death in learning tasks with the following charts that show aggregated mortality of all causes (e.g. death without any disease or incident, accidents, drug abuse, suicid, communicable and non-communicable diseases, ...), region by region:

  • (Pattern in Curves) Look at the pattern in these charts, what are the causes of these pattern? Try to find scientific evidence for these patterns?
  • (Environmental Conditions) Now let us analyze the communicable disease (because this learning resource was created under COVID-19 - there is more to examine - see first question).
    • How do environmental seasonal conditions have an overall impact on all causes of death (accidents, suicides, diseases, ...)?
    • What are environmental conditions that could increase spreading of communicable diseases in general and what are environmental conditions that help us to reduce the number of cases?
    • Explore the current scientific knowledge about COVID-19 and the impact of environmental conditions on aerosols, droplets directly and on the behaviour of people (staying more inside, closed windows in colder season, ...).
  • (Aggregated mortality and specific disease data) Now we compare aggregated mortality and disease specific data.
    • Aggregated data (as the title of this learning resource suggests) add up all causes of death and not only the data of a single disease. Take a non-communicable disease (e.g. breast cancer and remove the mortality of breast cancer from all years in mortality chart of all cases and add the breast cancer data of mortality just in 2020. Do you see an elevation above the normal mortality? If you do not see any elevation, would that justify to stop medical treatment because you analyzed aggregated data? What is the scientific reason to look at specific disease data, that is not aggregated to follow the developement of the disease and to access the impact of specific risk mitigation strategies.
    • The charts below aggregate all cause of death. Assume we aggregate not all causes of death but only causes of death of the class of influenca viruses to be more closer to virus disease. What are the seasonal pattern of influenca caused death and how do you explain that (look for epidemiological scientfic evidence). Compare Influenza and COVID-19 in terms risk mitigation strategies for the vulnerable patients. What are similarities and what are the difference between Influenza and COVID-19? If you apply the precautionary principle what are consequences for the responce in public health and risk mitigation strategies? Keep in mind that the seasonal data for class of aggregated Influenza have a longer scientific history.
    • Look at the peaks in mortality in 2020 compared to other years? Explain why is it necessary to collect the data from hospitals with the available medical equipments to treat patients for specific medical support for COVID-19 and not look at aggregated data of mortality to adjust prepardness for communicable and non-communicable disease, treatment if injuries, ... (e.g. analyze mortality that is caused of peak hospitalizations and the associated healthcare overload by a higher demand for treatment of respiratory disease that cannot be covered by health system.)
    • An infection causes "work" for the immune system even the patient is asymptomatic i.e. the COVID-positive patient is showing no COVID-19 symptoms. In aggregated mortality you get 0 or 1 for an covid-coded deaths. To what extend the disease contributed is difficult to see in aggregated mortality. A comparision of aggregated covid-positive death
      • in less vulnerable cohort without any other risk factors with
      • a more vulnerable population with additional risk factors
    • helps to understand how a COVID-19 infection has an impact on the number of deaths in the group? This is not entirely easy to get right in aggregated mortality data: deaths can in principle be contributed to by
      • improper treatment, partially disconntinued treatment without any medical advice, ...
      • unavailable medical resources for treatment (e.g. Intensive Care Units, ...)
      • unavailable scientific knowledge about the best treatment, which can be identified by medical studies, ...)
      • the missing communication about risks, Risk Literacy of patients (e.g. avoidance of necessary adviced treatments, that are recommended for the increasing the life expectancy of the patient)
      • etc.
  • (Technical Learning Task) Learn about the technical approach in providing the charts. What lessons can be learned to update other COVID-19 specific charts in this learning resource and keep them up to date with automated scripts.

Difference COVID-19 positive death and COVID-19 caused deathEdit

A COVID-19 positive test is counted if a someone, who died was tested positive. A COVID-19 caused death is much more difficult to count. Patients that had other diseases before and get an additional burdon to their immune system may die because of COVID-19. The can be extended to people with a more vulnerable immune system. In all that cases COVID-19 is an additional burdon to their health (if it minor, major of irrelevant) is difficult to count. COVID-19 is a new virus and it is difficult to add 0.1 or 0.7 to the death count instead of 0 or 1. How is it possible to identify the impact of COVID-19 on the death of patient in comparision to the other causes of death.

Basic ExampleEdit

Let us assume we have 3 causes of death: A, B and C (C stands for COVID-19 in this example). We want to quantify the contribution of C to A and B. We assume that the baseline before was unbiased of COVID-19. We can define to study and test patients with disease A and disease B for COVID-19 and identify if the number of death are significantly higher in the COVID-19 positive group than in the COVID-19 negative group.

Limitation of Test CapacityEdit

To perform those type of studies you need the test capacity. If the test capacity is a limited resource then decision making must be applied on the allocation of the limited test capacity. Basic introduction to assignment of limited test capacity in epidemiology can be other learing resource. We just give a basic example: We can use the limited test capacity

  • for quantification of the impact COVID-19 to other death causes or
  • for medical staff to protect other patients or for testing patients that show the symptom to make evidence based decision making for these patients.

Death Cases and COVID-19Edit

This part of the learning resource will look into death numbers and risk mitigation strategies. The preparedness plans of public health agencies, health care facilities, ... are more complex. Please follow and analyze peer reviewed public health journals on that topic and the recommendations of the public health system that must be updated according to new scientific knowledge in virology, medical treatments, .... This learning resource is a wiki, can be outdated, incomplete, ... .

Death Counts in early PhaseEdit

COVID-19 was new virus, when cases and death are described by medical staff for the first time. Tests were not available at that time. Medical staff or the public health system just observe epidemiological patterns of a contact induced disease (infections). E.g. people that were not sick before show all of the sudden with the same symptoms of the patient that they visited 2 weeks before and pop up in the same hospital. At that time medical decision making introduces protective measures to prevent other patients to be infected (percautionary principle).

Early Observations and MortalityEdit

For a new virus the available tests for other viruses are unappropriate and a test must be designed and scientifically analyzed. This takes time Nevertheless if the health systems observes mortality in conjunction with an unclassified disease that seems spreads within among humans, then preventive measures for other patients are introduced to prevent the virus from spreading, even if the public health system and virology cannot classify the disease at that time. The containment strategy tries to keep number of infections low, to avoid other epidemiological interventions like lockdowns, that have more sideeffects on society, economy, .... All-cause deaths will not be helpful in this early phase because the number of death cannot help contact tracing.

Classification of a new Virus and MortalityEdit

Risk mitigation of the public health system looks (among many other tasks) at "who had contact with identified new infections" (contact tracing). This Virology at the same time has the task to identify the genectic oode of the new virus (COVID-19) and classify the genetic code. The classification is necessary to estimate if the virus is high pathogenic (e.g. if a virus is classified as H1N1 virus, H5N1, ... . The classification is important especially for a new virus, because assessment of the risk is associated with genetic simularity to other known viruses. The challenging part is not the sequencing of the genetic code. The challenging part is the scientific analysis which part of the genetic code for "behaviour" of virus in the host. If early containment of the virus is the key, all-cause death or even the COVID-19 positive deaths will not be helpful for risk assessment. Nevertheless mortality is observed by public health agencies during the decision making process. The classification of the virus has uncertaintities. The similarities to other known viruses do not excluded the possibility that is more harmful or less harmful. The assessement includes if the virus is high pathogenic and/or high infectious. Similarities of virus can help in the early phase and all-cause death cannot help for the assessment. Furthermore ethical consideration must be respected. Waiting until all-cause death show a signicant signal of a pandemic, is again (a too) late indicator for implementation risk mitigation strategies. Nevertheless it is important retrospectively to look on all-cause death and other death causes to see what was the impact of COVID-19 on other causes of death.

Genetic Code and EvolutionEdit

Viruses can change their genetic code over time (e.g. create new subtypes). If the genetic code of the virus changes, also immunisation might fail. For some diseases the vaccination will be preventive for a long period of time. E.g. for influenza viruses the vaccination must be adapted from year to year. So the preventive measure must be adapted from year to year. It is the objective to prevent an epidemiological spread of a disease in the early phase, because new viruses spread with an exponential growth in the beginning, because a limiting factor for the growth like immunity of the population or vaccination is not there (see basic introduction to SIR-Model).

All regionsEdit

Some countries and regions are transcluded in sections below, while some are only in their separate pages, to prevent chart rendering problems and speed up page loading. Regions covered:

See also Category:Pages with graphs.

BelgiumEdit

Weekly all-cause deaths in Belgium, based on mortality.org data, stmf.csv[2]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in Belgium for 0-14 year olds, based on mortality.org data, stmf.csv[3]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded were excluded to prevent the worst effect of registration delay.

Observation: The above 2020 drop in the values could result from very pronounced registration delay, not visible to this degree for other countries; could it be something else?

All-cause deaths in Belgium in weeks 1-28, year by year, based on mortality.org data, stmf.csv[4]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in Belgium in weeks 40+ the year before and weeks 1-28 of the year, year by year, based on mortality.org data, stmf.csv[5]::

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

England and WalesEdit

Weekly all-cause deaths in England and Wales, based on mortality.org data, stmf.csv[6]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Observation: Above, the last week of each year generally seems to have a dip, a discontinuity in the data. This is not observed for other countries. To be researched.

Weekly all-cause deaths in England and Wales, based on mortality.org data, stmf.csv[7], 3-week average:

Weekly all-cause deaths in England and Wales for 0-14 year olds, based on mortality.org data, stmf.csv[8]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in England and Wales for 0-14 year olds, based on mortality.org data, stmf.csv[9], 3-week average:

All-cause deaths in England and Wales in weeks 1-29, year by year, based on mortality.org data, stmf.csv[10]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in England and Wales in weeks 40+ the year before and weeks 1-29 of the year, year by year, based on mortality.org data, stmf.csv[11]::

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

FranceEdit

Weekly all-cause deaths in France, based on mortality.org data, stmf.csv[12]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in France for 0-14 year olds, based on mortality.org data, stmf.csv[13]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded were excluded to prevent the worst effect of registration delay.

All-cause deaths in France in weeks 1-23, year by year, based on mortality.org data, stmf.csv[14]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in France in weeks 40+ the year before and weeks 1-23 of the year, year by year, based on mortality.org data, stmf.csv[15]::

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

GermanyEdit

Weekly all-cause deaths in Germany based on mortality.org data, stmf.csv [16]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded as suffering significantly from registration delay.

Weekly all-cause deaths in Germany for 0-14 year olds, based on mortality.org data, stmf.csv[17]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

All-cause deaths in Germany in weeks 1-25, year by year, based on mortality.org data, stmf.csv[18]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

ItalyEdit

Weekly all-cause deaths in Italy based on mortality.org data, stmf.csv[19]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in Italy for 0-14 year olds, based on mortality.org data, stmf.csv[20]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

All-cause deaths in Italy in weeks 1-24, year by year, based on mortality.org data, stmf.csv[21]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in Italy in weeks 40+ the year before and weeks 1-24 of the year, year by year, based on mortality.org data, stmf.csv[22]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

NetherlandsEdit

Weekly all-cause deaths in Netherlands, based on mortality.org data, stmf.csv[23]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in Netherlands for 0-14 year olds, based on mortality.org data, stmf.csv[24]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded were excluded to prevent the worst effect of registration delay.

All-cause deaths in Netherlands in weeks 1-28, year by year, based on mortality.org data, stmf.csv[25]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in Netherlands in weeks 40+ the year before and weeks 1-28 of the year, year by year, based on mortality.org data, stmf.csv[26]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

SpainEdit

Weekly all-cause deaths in Spain, based on mortality.org data, stmf.csv[27]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay.

Weekly all-cause deaths in Spain for 0-14 year olds, based on mortality.org data, stmf.csv[28]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded were excluded to prevent the worst effect of registration delay.

All-cause deaths in Spain in weeks 1-29, year by year, based on mortality.org data, stmf.csv[29]:

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

All-cause deaths in Spain in weeks 40+ the year before and weeks 1-29 of the year, year by year, based on mortality.org data, stmf.csv[30]::

mortality.org indicates the data for 2020 to be preliminary; above, the last two weeks available from mortality.org were excluded to prevent the worst effect of registration delay. The above is not adjusted by population size.

Age-standardized mortalityEdit

ONS article linked below has weekly age-standardised mortality rates in 2020, which includes Montenegro, Serbia, Wales and Northern Ireland, and shows comparison between the selected country and England in a chart. The mortality rates are per 100,000 and are age-standardized.

A staggering observation is that the overall normal-mortality differences between countries make much more of a difference than the covid does, as apparent e.g. from comparing low-rate Switzerland and high-rate Serbia; they do so year after year and are going to in near future. Switzerland's mid-term low values (as opposed to peak values) are at or below 15 weekly deaths per 100,000 while Serbia's are about 25 weekly deaths per 100,000.

Links:

See alsoEdit

External linksEdit