Data Analyst.
PhD Engineering.
Honorary Fellow in the Institute for Medical Humanities - Durham University.
The views here represent my own.
2 subscribed
Dec 20, 2023 • 10 tweets • 3 min read
Crying Wolf One Time Too Many
It is exactly 2 years ago that the UK government ignored scientists’ advice to tighten restrictions to combat Omicron.
Cabinet met on 20 December 2021 and decided to persist with Plan B despite SAGE’s apocalyptic scenario for Plan B.
(1/9)
Experts accused the government of ignoring the latest modelling by SAGE which pointed to between 3,000 and 10,000 hospitalisations a day, and between 600 and 6,000 deaths a day without additional restrictions.
How is this seemingly conflicting phenomena even possible?
Both the apparent increase and the apparent reduction in mortality can be explained by the healthy vaccinee effect.
(2/7)
Aug 30, 2023 • 7 tweets • 3 min read
ONS has released new deaths by vaccination data for England.
The data allows us to gauge the vaccine benefit in the absense of the healthy vaccinee effect.
The graphs shown are all age standardised.
(1/7)
ons.gov.uk/peoplepopulati…
The healthy vaccinee effect can be gauged by the difference between the unvaccinated non-Covid-19 mortality minus the ever vaccinated non-Covid-19 mortality.
It starts off high and reduces over time.
(2/7)
Aug 28, 2023 • 8 tweets • 2 min read
Covid-19 #diseasemodelling
How a lack of basic sanity checks resulted in major modelling errors and expert errors during the pandemic.
For balance I list examples on both sides, some underestimating severity, some overestimating severity.
(1/8)
On 17 April 2020, Johan Giesecke (Sweden) estimated a Covid-19 IFR of 0.1% and a first wave attack rate of 50%.
The IFR turned out to be 0.58% in Sweden after the first wave with an attack rate of only 6%.
(2/8)
Aug 25, 2023 • 5 tweets • 2 min read
Royal Society report on Covid-19 interventions:
If you exclude ethical, legal, social, economic and other health effects, the study shows that the most stringent measures (stay at home orders) were the most effective.
(1/5)
The study neglects the scenario of a failed hard and early lockdown, and focuses on countries that succeeded in elimination:
If elimination is possible, going hard and early results in a shorter lockdown.
If elimination is impossible, going hard and early prolongs the lockdown. The end is bound to be tragic.
(1/5)
Lockdown proponents point to New Zealand and China that succeeded in elimination (at least for some time).
Lockdown critics point to Peru, that locked down hard and early, but ended with the highest Covid-19 deathtoll and some of the longest cumulative lockdowns.
(2/5)
Jun 27, 2023 • 5 tweets • 2 min read
Matt Hancock again advocating for harder, earlier lockdowns.
But does an early and hard lockdown necessarily result in a better ourcome?
Lockdowns are costly exercises, so one would think that 16 months into the pandemic, modellers would know what is likely to happen after the final phase of unlocking in England.
Here SAGE-SPI-M shows exactly what they expect to happen after "freedom day" on 19 July 2021:
(1/6)
And here is the actual hospitalisations (thin grey line), right in the middle of the confidence interval during August.
But wait, this confidence interval seems rather wide (100-10,000 hospitalisations per day!).
And the Y-axis is plotted on a log scale.
(2/6)
Apr 27, 2023 • 14 tweets • 2 min read
Why was modelling accuracy so poor during the Covid-19 pandemic?
This thread argues that the move away from ‘descriptive modelling’ to ‘mechanistic modelling’ in the disease modelling community, was partly responsible for this blind spot.
(1/14)
Disease modelling moved away from ‘descriptive’ modelling to ‘mechanistic’ modelling in the early 20th century according to Adam Kucharski in his book The Rules of Contagion (2020).
(2/14)
Apr 3, 2023 • 10 tweets • 3 min read
Were lockdown policies a form of contagion?
Here I use 'The Rules of Contagion' by @adamjkucharski to critically evaluate the spread of lockdown policies.
1/
Lockdowns were foreseeably harmful to the global poor, and this is now widely accepted.
The public health principle, that 'one size does not fit all' was ignored.
I have just completed "The Herd: How Sweden chose its own path through the worst pandemic in 100 years" by @johananderberg.
It is an excellent read and covers all the differences between individual experts from early in the pandemic when Sweden decided to not lock down.
(2/8)
To understand the Swedish response requires knowledge of Sweden's overreaction to the 2009, 2010 H1N1 swine flu virus.
According to Anderberg "the swine flu incident left its mark on Swedish society".
Mar 21, 2023 • 5 tweets • 2 min read
(1/5)
Your belief about how well the lockdown worked depends on your belief about what would have happened with mitigation only.
Based on the ICL Rep 9 scenario for mitigation in the UK, the lockdown worked extremely well!
CC bed demand dropped from 90 to 5 beds per 100,000!
(2/5)
The Swedish Public Health Agency modelled a much lower critical care bed demand for Stockholm (12.5 beds / 100,000), presumably for the measures already in place.
Sweden did not lock down.
Actual ICU beds occupied came in lower (7.8 beds / 100,000).
Mar 19, 2023 • 5 tweets • 2 min read
(1/5)
While prominent epidemiologists in Sweden underestimated IFRs and overestimated attack rates at times, the official reports from the Swedish public health agency were more accurate:
(2/5)
This report estimates an attack rate of 26% for Stockholm metropolitan area (2.4 million people) after the first wave. This implies an IFR of 0.3%.
I was hoping that by now, most people would have moved away from extreme vaccination positions on either side:
The extreme positions are:
1.Ongoing Covid-19 vaccinations and boosters should be mandated as they are completely safe and effective.
2.Covid-19 vaccinations did not work at all, came at significant risk, and no-one should ever have been vaccinated.