Updating Priors To Accommodate New Evidence

Written by Martin Neil and Norman Fenton on February 5, 2024. Posted in Current News

Revisiting our published work on covid infection prevalence and fatality rates from May 2020.

In May 2020 we published one of our first papers on covid-19, in the Journal of Risk Research.

Given it isn’t a medical, biological or epidemiology journal this might look like an unusual venue for such a topic, but we published there because they were specifically looking for articles on covid ‘risk’.

Also, and perhaps more importantly, we were starting to see the wave of censorship coming our way, and (rightly) thought publishing it anywhere that would take it would be better than nowhere.

Our main motivation for writing the paper was the shock at the over-the-top hysterical media and professional reactions to the preprint published by John Ioannidis in early 2020, which was ultimately published here in 2021.

Ioannidis was claiming that covid prevalence was high and the fatality rate was significantly less than the CFR (case fatality rate) being quoted from China, at 3.4 percent. We wanted to double check his numbers by running a similar, but more sophisticated analysis, using as much available data as we could get.

It is now almost four years since publication, and it has been quite a ride since then. We’ve learned a lot about the ‘pandemic’ and many of our scientific positions have changed since 2020, some in quite subtle ways and other changes are of seismic proportions.

It is very technical in places and so quite specialist. The abstract says:

Widely reported statistics on Covid-19 across the globe fail to take account of both the uncertainty of the data and possible explanations for this uncertainty.

In this article, we use a Bayesian Network (BN) model to estimate the Covid-19 infection prevalence rate (IPR) and infection fatality rate (IFR) for different countries and regions, where relevant data are available.

This combines multiple sources of data in a single model. The results show that Chelsea Mass. USA and Gangelt Germany have relatively higher IPRs than Santa Clara USA, Kobe, Japan, and England and Wales.

In all cases the infection prevalence is significantly higher than what has been widely reported, with much higher community infection rates in all locations. For Santa Clara and Chelsea, both in the USA, the most likely IFR values are 0.3–0.4 percent.

Kobe, Japan is very unusual in comparison with the others with values an order of magnitude less than the others at, 0.001 percent. The IFR for Spain is centred around one percent. England and Wales lie between Spain and the USA/German values with an IFR around 0.8 percent.

There remains some uncertainty around these estimates but an IFR greater than 1 percent looks remote for all regions/countries. We use a Bayesian technique called ‘virtual evidence’ to test the sensitivity of the IFR to two significant sources of uncertainty: survey quality and uncertainty about Covid-19 death counts.

In response the adjusted estimates for IFR are most likely to be in the range 0.3–0.5 percent.

We analysed a hodgepodge of all the sero-prevelance and mortality data we could get at that time from the UK, Spain, Japan (Kobe), Germany (Gangelt), and the USA (Santa Clara in California, and Chelsea Massachusetts).

We were aware of some, but not all, of the issues around the trustworthiness of the mortality statistics but made adjustments for this in ways that Ioannidis did not.

When, we look back at the paper we now realise we would interpret the results quite differently – with the benefit of hindsight. So, let’s look at some of our main results.

Infection prevalence and community infections

Our first estimate was of the proportion of people infected as detected from seroprevalence testing, infection prevalence rate (IPR) (note that the rate here is measured on a scale of 0 to 1 rather than as a percentage 0 to 100)

For each location the associated graph represents our belief about what the true IPR (shown on the horizontal axis) was at that time given the available data.

So, for example, the graph for Gangelt Germany shows that we believe the rate almost certainly lies between 0.1 and 0.22 (i.e. 10 and 22 percent) with a mean rate of 0.16 (i.e. 16 percent). Notice how very different the (mean) estimates are from each other.

Also, some are very uncertain, such as Chelsea and Gangelt, whilst others are very tight and sharp, such as Spain. The results suggest that they were doing a huge amount of testing in Spain and finding that around five percent of the population was antibody positive.

Yet in Chelsea the mean rate was nearly 30 percent. On the west coast of the USA in California the rate was only 0.8 percent.

From community testing (via PCR) we could estimate the ratio between what was seen in the seroprevalence tests and community testing:

So, for example, for Gangelt Germany we estimate a ratio of between 150 to 250 with mean 200. This means that there were around 200 positive PCR tests for each positive seroprevalence test. The ratio for Kobe, Japan was off the chart (greater than 300). For the rest it was around 10-50 positive PCR tests for each positive seroprevalence test.

At the time we made no attempt to explain these differences, except to say that there appeared to be a lot more covid around in Japan and Germany than in the UK, USA and Spain. Would we come to that same conclusions now? No, we do not.

If we assume that there was a highly transmittable virus travelling around the globe, following some natural transmission pattern, we should not expect to see such disparities in seroprevalence between neighbouring countries, and neither would we expect to see such dramatically different levels of community spread of the virus across countries.

Why would Germany be so different from the UK and Spain? Why would Japan be completely different from everywhere else?

Infection fatality rates

Our second estimate was of the infection fatality rate. Before making any adjustment for the mortality data this is what we estimated for each country/region.

Notice how different they all are. The Japanese infection fatality rate is completely different from all of the others.

People were just not dying from covid in Japan at the rate they were elsewhere. Spain had, by far, the highest infection fatality rate, whist the others fall in-between.

In our paper we said:

There is some controversy over UK fatality counts (Fenton et al. 2020a) and there appear to be many reasons not to trust the fatality count, including:

Ambiguity and confusion about diagnostic criteria for Covid-19

Care home deaths not certified by a qualified medical practitioner.

Hospital and other deaths signed off as ‘caused by’ Covid-19 when they are ‘with’ Covid-19

The number of excess deaths could be much higher because of cases that remain undiagnosed.

Similar concerns apply to Spain and potentially elsewhere. Given that these uncertainties work in both directions we must be careful to include under and over-estimates and not chose one to suit our prejudices or political outlook.

So, we adjusted for this, and produced revised IFR estimates (it is all in the paper – there were no black box ‘fudges’):

These fatality rate estimates now look similar, with each having a mode of 0.3%-0.5%. So, our most likely estimate was a fatality rate perhaps comparable to the flu, and a lot less than was being claimed by Ferguson and others.

Reflections

Scientists are supposed to change their mind in the light of evidence. As Bayesians we live by this mantra. But looking back on it were these the right conclusion to make? Firstly, we had some pretty big premises at the back of our mind, and forming our priors, when conducting the analysis:

• The healthcare responses were irrelevant.
• The seroprevalence and PCR tests were trustworthy.
• The mortality data was accurate.

None of this was true.

Secondly, we now know that in some places the spring 2020 deaths were largely iatrogenic, i.e. caused by the inappropriate response to the perceived pandemic which especially affected vulnerable and elderly people in healthcare settings.

This included isolation from family members, denial of approporiate medical treatment and the introduction of novel end of life protocols.

We also now know that PCR testing was inaccurate, and the results massively exaggerated.

Looking at the data again – with 20/20 hindsight – we should have concluded that:

• The range of seroprevalence positive rates was far too wide to suggest it was diagnosable of anything much, and not consistent with a transmissibility pattern.
• The vast differences between seroprevalence and PCR positive rates undermined many of the supposedly reliable theories underpinning how people got infected, sick and then died.
• The range of fatality rates was far too wide to be consistent with a single common cause – the virus – and hence we should not have applied a uniform adjustment.

In conclusion we were naïve and too trusting of the authorities, but also at that time we were blind to the iatrogenic healthcare effects and the issues with PCR testing.

See more here substack.com

Header image: FIKA – PIKA / Shutterstock

Please Donate Below To Support Our Ongoing Work To Expose The Lies About Covid 19

PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX. 

Trackback from your site.