3
$\begingroup$

Please note

  • This question is based on data from the UK government's web site (data from the ONS). I am not taking this data as accurate, but am using this as it is the only source of data I can find, and is what (presumably) the government themselves use when deciding what to do. My question is about trends, not the accuracy of the government's data.

  • I'm also not asking about the wisdom of government policy, that is a political issue, and I'm only interested in biology and statistics here.

  • Finally, I realise that my analysis is all quite rough and ready, as there are many indeterminate factors involved. Again, I'm interested in an overall picture, so a rough analysis is probably reasonably informative.

Here in the UK, the data on new virus cases seems to imply we have a serious problem...

enter image description here

This was plotted using data from the government's web site, and shows that the number of new cases has risen dramatically since the beginning of September. I used a 5-day average to smooth the line (same for the other graphs below)

However, if you look at the number of deaths, it seems that the increase is nothing like as rapid...

enter image description here

Whilst there has been an increase since the beginning of September, it is very much more modest, and not necessarily indicative of a serious problem.

Now, I'm not a medical expert in any way, but it seems to me that the fact that the number of deaths is growing more so much more slowly than the number of new cases indicates that things may be not actually be getting worse, but may be getting better. Sure, more people are getting the virus, but the percentage of those who are dying (which I would have thought is the most important issue we need to consider) is not increasing.

To check this, I divided the number of deaths by the number of new cases, and it showed a very different picture...

enter image description here

As you can see, there isn't a noticeable increase since the beginning of September. To examine this more closely, you can zoom in on the past 8 weeks or so...

enter image description here

Whilst there was a jump in the middle of Sept, it hasn't really increased since then, and may even be decreasing.

So, is my analysis wrong, or is the virus actually becoming less of a problem in the UK?

$\endgroup$
10
  • 2
    $\begingroup$ One obvious problem with your graph is that people don't die immediately after diagnosis. Many people diagnosed in September will die in October, but you aren't counting them. You can see possibly the beginning of an October death spike in what data you do have. $\endgroup$
    – swbarnes2
    Oct 12 '20 at 23:52
  • $\begingroup$ @swbarnes2 That would also make the beginning of a good answer. $\endgroup$
    – Bryan Krause
    Oct 13 '20 at 0:13
  • 3
    $\begingroup$ This would be improved by cutting out some of the rhetoric, e.g. "winding people up into a panic", which suggests that you already have an answer in mind. You might also consider that (if the UK was like the US) many of the early deaths were among extremely vulnerable populations such as older people in nursing homes. Those people having been killed off already, so they aren't a factor in ongoing statistics. $\endgroup$
    – jamesqf
    Oct 13 '20 at 16:10
  • 3
    $\begingroup$ @jamesqf I agree with the first part of your comment, but definitely not with the second. There are still plenty of very vulnerable people around (at least in the US) that avoided being infected during the first round, but are still quite susceptible to disease should they be exposed to it. Round I was certainly not a clear-cutting operation among the sick and elderly by any stretch of the imagination. $\endgroup$
    – MattDMo
    Oct 13 '20 at 16:17
  • 1
    $\begingroup$ @jamesqf You're right, I did have an answer in mind, but was trying to ask objectively, as I'm quite prepared to be proven wrong. I've edited the answer to remove the comment you mentioned. $\endgroup$ Oct 13 '20 at 20:09
4
$\begingroup$

The issue is using the number of positive tests as a measure for the total number of active cases at a given time. At least in the U.K., the early testing capacity was much lower than it was now: enter image description here

Because of this, many people who were infected were not being tested and so weren't included in the total number of infections. Therefore, the total number of infections in April/May time was likely a large underestimate. It's been estimated that there was something in the range of 150,000-200,000 new cases per day in that time period.

On the other hand, testing capacity in the U.K. at the moment is much higher and will catch a much higher proportion of the total cases.

Therefore, your crude calculation of 'death rate' = deaths / cases is skewed by the error in the estimated number of cases, making it appear the 'death rate' is going down because we underestimated the total number of cases early in the pandemic.

The other thing to note is that the 'death rate' is dependent upon demographic factors, most obviously age. Early in the pandemic in the U.K., a large number of cases occurred in care homes, where the residents were particularly vulnerable. On the other hand, cases are high in students at the moment, who are a low-risk group. If the age distribution of cases is relatively skewed towards older people, then the 'death rate' will appear higher than vice versa, even though nothing may have biologically changed with the virus.

Note: Sorry, I don't have citations for a lot of this. I will try and find them and add them at a later point.

$\endgroup$
3
  • $\begingroup$ Thanks for that, very informative answer. My only comment would be that while everything you say makes sense when comparing data from the spring to now, but if you only consider data from (say) the beginning of Sept, as I did in my question, your points wouldn't make such a big difference. Or would they? I would be interested to hear what you think. Thanks again $\endgroup$ Oct 13 '20 at 20:04
  • $\begingroup$ Just by looking at the data, I would say that those deviations could well be within the margin of error (i.e. confidence intervals) of the numbers of cases and deaths recorded. Of course, there is a large amount of error in the estimation of the number of cases, so I expect those confidence intervals to be quite large. IN that case, the deviations we see may well just be 'random noise' that don't correspond to anything meaningful. $\endgroup$
    – user438383
    Oct 19 '20 at 22:31
  • $\begingroup$ Fair comment, thanks $\endgroup$ Oct 20 '20 at 13:23
1
$\begingroup$

Schools opened

What happens in the beginning of September. Something that would expose a large number of people to the virus, particularly people young and healthy enough to have a low risk of death due to virus but who still have decent chance to catch the virus.

$\endgroup$
3
  • $\begingroup$ Excellent observation, never thought if that. I wonder if there is any data showing deaths by age, as that would help confirm or contradict your suggestion. $\endgroup$ Oct 13 '20 at 20:05
  • 1
    $\begingroup$ @AvrohomYisroel you also have to keep in mind that the death rate amongst young people is a lot lower than for older folks, but young people can pick up the virus and spread it to older people they come in contact with, while not necessarily getting very sick or dying themselves. I think one lesson out of this is that there are all sorts of variables at play here, and you can't necessarily draw universal conclusions just out of the data you've analyzed so far. $\endgroup$
    – MattDMo
    Oct 14 '20 at 18:03
  • $\begingroup$ @MattDMo Agreed, very valid observation. I said right from the start that my analysis was rough and ready, I'm now seeing quite how rough it was! $\endgroup$ Oct 14 '20 at 19:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.