(ME) outbreaks can be modelled as an infectious disease: a mathematical reconsideration of the Royal Free Epidemic of 1955, 2020, Waters et al

Paywall, https://www.tandfonline.com/doi/abs/10.1080/21641846.2020.1793058?journalCode=rftg20
Sci hub, https://sci-hub.tw/10.1080/21641846.2020.1793058

 

Do we know any more on the authors, apart from the details given at the end of the article?

 

Very interesting.

 

Yep, citizen scientists.

 

oooo, fascinating.

 

Excellent. Dig it up by the roots.

 

This looks to me like a very elegant approach.

On the one hand, you have McEvedy and Beard who simply read the medical notes from the 1955 epidemic - choosing not to contact either the patients or their treating doctors - and drew a conclusion that fitted their beliefs. That seems to be a method wide open to bias.

On the other hand, this new study looked for external evidence that would support either an infectious or hysteria basis for the 1955 outbreak. They extracted datasets for the Royal Free 1955 and other outbreaks, and a hysteria outbreak and looked to see if the data fitted the pattern of an infectious or hysteria outbreak.

I'm in no position to judge the validity of the models, though I do know that the SIR (Susceptible/Infected/Recovered) approach is widely used to model infectious outbreaks, including Covid-19/novel coronavirus (e.g. the Imperial College coronavirus model).

However, the approach of looking for new data that would throw light on the question is refreshing.

 

I hate to sound negative, but I would be careful to cite this paper. The analysis is interesting and they raise some valid points, but the conclusions are not supported by the data.

The relation between gamma and beta in the least-square fit will depend on the time window where the fitting is made and also on the number of people at risk (N in the paper). It is not clearly specified how these values are chosen. The fact that you don’t have a minimum in Figure 6 may be caused by too short a time window in the analysis.

Moreover, the peaks in the Royal Free epidemic that are indicated with arrows in Figure 2a do not necessarily reflect separate outbreaks. The SIR model is continuous and assumes an exact incidence and recovery rate. In a more realistic approach, these rates should be modelled as a probability per unit time to become sick or well. When you include such stochastic aspects into the model, the different epidemics vary a lot in length and number of affected people also when the gamma and beta parameters are kept constant. In the Figures below, I show the results of two simulations with identical parameters but with rate modelled as probability per unit time and the number of people actually affected decided by a random number generator. The second example has the appearance of two or three separate outbreaks, but these variations are only due to chance.

Most importantly, the authors should explain which qualities in Figure 6a that cause the absence of a least-square minimum. I believe it is a bad choice of the simulation window, but the rapid variations of the incidence—for example during days 7, 8, 9, 10 and 11—contribute to a poor fit. That may be caused by a combination of poor reporting and chance. There are no cases reported during day 10 and 11, nor during day 17 and 18. These blank spots are exactly one week apart.

I think that the authors miss an important point. Epidemics attributed to mass hysteria are typically short in duration, at most two or three weeks. This also applies to the Kombolcha outbreak that is studied in the paper. The ME outbreaks, on the other hand, lasted for several months. Hence the duration suggests that the ME outbreaks were not mass hysteria.

In conclusion, modelling of ME epidemics is interesting, but I think this paper leaves many unanswered questions.

 

I don't think a believer in mass hysteria would be convinced. Comparing an (alleged) outbreak of mass hysteria among young girls in Ethiopia in 2013 with another outbreak in a hospital in 1955 England doesn't seem like a good comparison. A believer in mass hysteria would say that these outbreaks differed because the social and cultural environment was very different.

I doubt that in the Ethiopia outbreak, there were the resources and knowledge to properly test for environmental factors.

Mass hysteria is not really a scientific concept because it's not possible to prove or disprove. We just assume it exists because we can't find an explanation for some events, and we assume it's not present when we find an explanation.

 

do we know the Eithiopian outbreak was mass hysteria - there was a recently labelled outbreak ( with hallucinations etc) which was found to be due to rice that had been eaten ?

 

Can you tell us more about this (the rice in particular)? Was it the alleged mass hysteria in Malaysia?

 

https://www.s4me.info/threads/bbc-article-re-mass-hysteria-in-malaysia-sw-comments.10831/

 

From Byron Hydes Little red book

https://www.imet.ie/imet_documents/BYRON_HYDE_little_red_book.pdf

 

If one thinks through that court exchange one might come up with interesting hypothoses. Why would counsel have been cross-examining Beard in such terms? Presumably because he had provided an "expert opinion" for the defence. Who would "The Defence" be? Presumably an Insurance Company. One wonders what insurance company. One of the largest insurance companies was taken over by UNUM round about that time. I used to know which.

I think it was in his obituary that it was observed that he liked to spend time with his private practice.

 

I'm not sure, but I can remember it being linked to rice that they had eaten at school.
I will see of I can find anything on it.

 

Here's another... Aldous et al. (1994) An outbreak of illness among schoolchildren in London: toxic poisoning not mass hysteria
https://pubmed.ncbi.nlm.nih.gov/8138768/

 

Interesting study.

The mathematical modeling is above me but what I would like to see is a probability: how unlikely is it that random data fit the SIR model as the data from the 1955 ME outbreaks did? I suspect this is possible to do, using some simulations.

It is also unclear if the SIR model can be used to differentiate outbreaks from an infectious disease with outbreaks that are due to ‘hysteria’ or something else. One could, for example, take data on several infectious outbreaks and data from what are believed to be‘hysterical’ or other outbreaks and give them to blinded statisticians and see how well they can classify them based on fit to the SIR model.

 

Toxic poisoning can happen much more easily than one would expect. All that needs to happen is a farmer being distracted for a moment or choosing the wrong product. We all have made errors like these.

Another way for it is to grains being affected by fungal toxins while in storage. And probably there are many other ways.

Toxic poisoning cases might be overlooked when they aren't severe enough and don't occur in clusters. In the Aldous paper, the poisoning was discovered because by chance, it happened to affect many people at once and because some effort was put into investigating this possibility.

In countries like Ethiopia and Malaysia they might not have resources to investigate the toxicological angle.

 

I wonder if modelling the Camelford incident in Cornwall would indicate hysteria as Wessely stated or poisoning of water supplies. (Aluminium salt I think iirc)