The importance of excess mortality in a model, assumptions and scenarios

Accountability

In the Beta version, three buttons could be turned:

1. Vaccine effectiveness: number of deaths prevented, estimated for Omicron virulence
2. Excess mortality factor: The share of vaccinations in excess mortality (all mortality above the expected baseline, minus covid mortality)
3. Expected virulence of the virus in relation to e.g. Wuhan/Omicron/BA.5 (we chose Omicron as a reference)

In het eerste artikel op Maurice.nl werden 1. en 2. beide hypothetisch op 100% gesteld. Daar sloeg menigeen op aan: Vaccins voor 100% de schuld geven van de oversterfte werd zelfs "misdadig" genoemd. Gelukkig ageerde niemand tegen de eveneens discutabele claim van 100% vaccin-effectiviteit! Voor punt 3. is Omikron als referentie genomen.

Adjustments
In response to targeted criticism, we have implemented two additional assumptions that related to the past period, the basis on which further calculations are made:

4. The past Vaccine effectiveness on the Omicron variants serves as a reference. It was estimated at almost zero in the Bèta but that, like any assumption, can be substantiated but is also debatable. We can now easily make calculations if we assume higher observed Omicron-AE.
5. We had assumed RIVM death figures. Several experts (thanks Hans) pointed out to us that although the figures from CBS are not plausible, it is more realistic to temporarily increase the RIVM figures by about a third. Initially, we wanted to stick to the official figures institute as much as possible, but here we allow ourselves a small correction.

Example of points 4. and 6. What difference does it make?

De verschillen als gevolg van deze twee toegevoegde factoren zijn relatief klein. Gezien de aannames en schattingen waarmee moet worden gewerkt beoordelen we de cijfers achter de komma in dit geval als schijnnauwkeurigheid. Het gaat om het grote plaatje. Om de schijn te voorkomen dat we de vaccins zouden willen benadelen, hebben we beide punten meegenomen in alle onderstaande scenario's.

ad 4) There are approximately 1,500 Omicron deaths. With a VE of 50%, for example, we have to calculate with 3,000 potential Omicron deaths, half of whom have not died thanks to the vaccination. In the first model, 1,500 was calculated. Despite the fact that the VE against Omicron is very low and is decreasing very quickly, especially by vaccine standards, we still consider an average historical VE of 20%.

ad 5) Er zijn circa 1.500 Omikron-overlijdens te betreuren. Naar aanleiding van kritiek op de rivm-registratie hebben we die voor dit artikel opgehoogd met 33% ter compensatie. Het zijn er dan 1.995. Natuurlijk is die 33% een gecalculeerde schatting die ook weer tot discussie leidt - een discussie die ook intern voortduurt.

Presumably that percentage has to go down or RIVM turns out to be correct after all. This is less essential for the overall picture than it seems. Below in graphs 1 and 2 an example of the differences we talk about (click on the graph to zoom). The difference is most visible in the 70-79 age group. We do not attach enormous value to this degree of precision given the margins of uncertainty that we already have to calculate.

Other comments

Cohort distribution
There are certainly other improvements that can be made. The mortality rates were substantially different in the previous model due to a different synchronization method of age groups. We are now working with an improved version and on further refinement. Unfortunately, institutes report in random age classifications. That is why choices have been made in reconciling reports. Responsibly splitting from 0-65 to 0-49, 50-59, 60-69 was a challenge. Thanks to foreign studies, we have found a well-founded distribution. (link to VAERS study)

With the help of the VAERS figures (number of deaths due to vaccine per million vaccinated people), it is possible to calculate how many deaths can be expected, taking into account the NL situation in terms of number of inhabitants and vaccination coverage. But VAERS only counts what is reported, for 11 months, of which probably the first 4 without damage (December 2020 to November 2021). So 7 months of damage against NL 12 months. That has also been settled. That number is then the real minimum of vaccination damage you can expect in NL. There are 959. But in total we think we will see 10,585. Assuming that this is due to underreporting at VAERS, we make a pro rata distribution.

Optimization therefore has our attention. A further division of 0-50 years is also desirable. Of course, many more improvements and refinements are conceivable. In our opinion, the model already offers sufficient insight for an initial orientation.

Brief explanation of the estimated corona deaths
To determine the excess mortality, in our previous article we had taken the excess mortality figures published by CBS and reduced them by the Corona mortality as reported by the RIVM. We received comments on this, because the RIVM figures would be an underestimate of the actual corona mortality and the CBS figures more representative.

If we look at it over a year and we take 2020, a year in which there has been no vaccination yet, we see that the extra mortality over that year does not differ very much from the reported corona mortality of the RIVM: 11,520 and 12,682 (CBS).

Excess mortality is a period of increased mortality compared to the forecast. If extra mortality is detected in a year's time, this means that people have died who would normally live to see the turn of the year. We cannot say how many months shorter they lived. These are people who belong to the 12,682 of the CBS.

Er is echter nog een andere groep mensen die is gestorven door corona, terwijl ze normaal gesproken gemiddeld ook nog maar een half jaar hadden kunnen leven. Dat aantal zien we dus niet in de extra sterfte per jaar van het CBS terug, dat was al verrekend. Hoeveel mensen dit precies zijn is moeilijk te zeggen, ook krijgen we de bekende "aan of met corona" discussie. Wij hebben de doodsoorzaak statistieken van het CBS bekeken en gekeken naar onwaarschijnlijke afnames van bepaalde doodsoorzaken en de rapporten van stichting NICE over de opnames. We schatten hieruit in dat de sterftecijfers van het RIVM 33 procent te laag zijn. Dit hebben we verrekend.

According to Statistics Netherlands, many more people have died of corona than the excess mortality shows. To calculate net mortality, i.e. mortality from causes of death other than corona, we take total mortality and subtract it by corona deaths. Net mortality will then be almost 8000 people lower than the mortality forecast for 2020 (compare with earlier Virusvaria article from Nov, 2021). That is strange and it strongly suggests that the reduction in deaths was not due to "other causes" but "with covid" and corona was not the cause of the deaths. For this reason, the RIVM figures seem to be a more reliable source of information, especially after correction by this 33 percent.

Period
We calculate from the beginning of 2021, which means that we include several waves in the calculation. A vaccine can prove itself better the more cases of that disease occur. And vice versa: the less (serious) illness, the more difficult. If the disease does not occur, only the possible harmful effects remain.

Quantitative approach
Needless to say: this is a purely arithmetical approach based on assumptions. To what extent you can sacrifice a healthy person (perhaps involuntarily) to save several vulnerable people and how that ratio should be with a vaccine, whether a vaccine should cost the lives of healthy people at all and other ethical questions I have already asked previously raised. We compare 1-on-1 here, without quantitative value judgment or QALY calculations. That's another chapter.

Logarithmic vs linear
Graphs 1 and 2 are presented on a logarithmic scale. As a result, the differences in the groups with low mortality are also visible. Differences between small numbers appear larger than they actually are, differences between large numbers are masked. Behind graphs 2 and 3 are the same figures.

This calculation model was developed in Excel by Herman Steigstra and is co-managed by Team Maurice and Virusvaria. If you have developed something in Excel that could be useful for a better estimate of unknown quantities, we would like to hear from you.