web analytics

Comments on “triple threat” article

I have been asked a question related to the COVID-19 data I included in my The Conversation article on COVID, flu and RSV:

It is related to the first figure in my article:

AK/UKHSA

In this graph, the cases (“Pillar 1 episodes”) go down while the positivity (the proportion of tests which are positive) goes up. These two data seem to contradict each other.

So, are the COVID numbers going up or down?

The answer might be more complicated, but I suspect the main reason is the testing intensity. If the “true” number of cases is roughly the same, but the number of tests goes down, the “reported” number of cases will also go down.

If we tested completely randomly (or is compulsory), this should not matter to the proportion of positive tests.

Of course, the more we test, the more accurate the predictions are and the smaller the variability, but the results will be consistent:

But, in reality, testing is not random. People will only be subject to testing if they have symptoms or are more at risk. As a result, the fewer the tests, the more likely it is they catch positive results:

Here, we focused on the “cluster” of cases in the middle. So, what can we say about the “true” number of cases of COVID-19 in England in 2021 and 2022?

There are other sources of data, for example the Zoe app, catching those with symptomatic COVID. The graph below only captures the last three waves, March-June 2021, June-September 2022, and September-November 2022. The first two are about the same, whereas the most recent one is smaller.

The number of deaths is another indicator, with the advantage of not being dependent on testing.

If you line up all waves since the summer of 2021, the number of deaths looks very similar, except for the two large omicron waves:

The most recent wave (September-November 2022) is similar to the June-September 2022 one but appears to have peaked earlier and at a lower value. In contrast, the March-June 2022 wave was much higher, almost as high as the omicron peak in October-December 2021.


There was a question about Australia’s influenza data. In my article, I used internet sources to state that in 2022 Australia experienced the worst flu season in 5 years.

I did not do a detailed analysis of hospitalisation and other data; apparently they show that the season was not as bad. But, I did use WHO FluNet notifications:

Print Friendly, PDF & Email