Lies, Damned Lies and Statistics



This article is a continuation of 'Statistical analysis methods to count COVID cases' article. 

Public health England has stated. They will start to publish three different daily death figures. 

  • One with a twenty-eight-day cut off after a positive test. 
  • One with a sixty-day cut off and the current figure. 
  • Anyone with a positive test who then goes on to die

The most accurate figure will still remain. The slightly delayed data from the Office for National Statistics. They count anyone with COVID mentioned on the death certificate. But the most important thing to remember. The trends of all these different data sets' broadly match. They all imply a downward trend which has to be a good sign.

One thing that is not decreasing at the moment are the number of cases. In early July, the rolling average of daily confirmed cases in the UK before July had fallen to around 500 or 600. In the past few weeks, it has been rising past 600 past 700 past 800 and now past 850. That potentially spells trouble. Because more cases today mean more hospital admissions. More deaths later in the summer.

What can be learnt from increased testing over this period? 

Of course, if you test more you will find more cases. Carl Heneghan Professor of evidence-based medicine. At the University of Oxford wrote a blog post on the subject. He compared the number of tests process weekly in the UK to the number of cases we are finding. And concluded that the ratio was not increasing with time. 

But not everyone agrees. Professor Christina Pagel. Director of the clinical operational research unit at University College London. 

Professor Christina Pagel has said. What is wrong with tests is that test processed is not actually the number of people you tested. The issue with testing is there is quite a large distinction between the number of people who get tested. And the number of tests the government claims it is doing. So, we need to look at the number of people tested and compare that with the number of people who test positive. 

So, what do we learn from that?

Professor Christina Pagel states, there are two main ways to get tested in the UK. 

Pillar one tests generally hospital tests. If you look at that hospital infection rates. They have come down. They have almost halved since the beginning of July. So basically, it is saying in hospital COVID is less prevalent.

Professor Christina Pagel went on to say. The other way of getting tested is called pillar two community testing. That is the type of testing you have if you have any symptoms either you go to a testing center or a home test is sent to you. The number of people tested has gone up over July. It gone up by almost forty percent.

What you see from those who have been tested positive. Is in the first two weeks of July positive rates are still coming down. But then from the 16th July onwards the last two weeks it has been going up again. For instance, since the mid of July although the number of people being tested has gone up by twelve percent. The number of positive people has gone up by forty percent.

So, you can’t explain the number of cases we have seen just by increased testing. This kind of makes sense if you think about it. From a timeline you know we opened our pubs and restaurants and in other areas on the 4th of July. We've seen in other countries. It takes three to four to five weeks for that to show up in cases and now it is showing up in our cases.

Should we be worried about a second wave is in the offering?

This is not a second wave. The increases are small. What we should be worried about is what we have learnt from other countries. If COVID is under control for reallyreally long time it can still really quickly spiral out of control.

What is of concern is if you wait until you are actually sure that case are increasing, you have waited too long to act.

Because the one thing we have learnt from other countries. If a country acts quickly while cases increase they do the best in bringing down the number of cases. So, it is right to be worried and to act if we were worried. But we should not think we are on the verge of another March. We are not. And we should not think we will ever get there because we are testing so much more. Professor Christina Pagel went on to say we would act a lot quicker. 

So, what is the current positivity rate and how has it been increasing? 

The current positivity rate in the community is 1.66%. That has gone up from 1.35% it is a small increase and that compares with the United States where COVID is not surging. Areas in the US where COVID is rising they have a positivity rate of ten percent to fifteen percent. 

Professor Christina Pagel states. The worrying thing about things going up now is that it is August. There is no school there is no University. People are on holiday everyone is outside as much as they can be. So, this should be the easiest time of year to keep a lid on COVID. Because the one thing we have learnt is that it really does not spread outside.

What we should worry about is when everyone comes back in the autumn. 

If you just count students that is 2 ½ million people moving around the country in the UK. You are creating a situation where you have a lot of movement. You have a lot of people in buildings and a lot of potential for that virus to spread.

But Professor Carl Heneghan Oxford University says not so fast. Cases may be going up slightly and so to are cases as a proportion of people tested in the community. We cannot draw any conclusions from that.

Remember these cases aren't from a random sample of the population. The testing program is as it should be. Tests focus on areas such as Manchester and Leicester where the virus has been spreading.

Professor Carl Heneghan states. When you go case finding you must consider the population, you are testing. How many more people are you going out there and testing.

A good example might be somewhere like Leicester. Leicester has seen a significant up surge in testing. As soon as you start testing you are likely to discover a problem. Leicester implemented 10,000 tests within two weeks. Leicester went on to announce they have a problem. So, it is possible to conclude whenever you go somewhere, and you test more are going to pick up more positive tests.

For now what that means is debatable.

So, what evidence do we have of what is really going on. Is it because we're looking for more cases we find more COVID cases.

Professor Carl Heneghan goes on to say. If you look at the 1st of June, we processed about 100,000 tests. By the 1st of July it is up to 150,000 now it is up to 200,000 tests. We have doubled the number of tests in the last two months. Professor Carl Heneghan thinks at very low levels you must be very circumspect. About what the data means.

In all these areas where the cases have been going up what you are really interested in is the exponential rise. 

Remember the test cases should double every three to four days. In doing that you will see in about a two-week period what is happening. Because you are getting a doubling effect. None of these areas have seen that. 

What they have seen when they have tested is more cases are detected and then a flat lining effect. There is a second issue a regression to the mean effect happening. What we are going to see as we continue with testing in areas where increase testing occurs are spikes. Then flatlining regression to the mean. What we really should be interested in is exponential growth. That is when highly accurate data really matters.

Both Christina Pagel and Carl Heneghan agree.

We are unlike to see exponential growth in cases.

Professor Heneghan points out that there are many things we still do not know about our testing. How many false positive cases might be thrown up. How much virus someone needs to have before a positive test is revealed and so on.

What we trying to do is detect the signal from the noise. 

If we got a noisy background while testing. Plus, we cannot be clear about the cycle threshold. Plus, we are not clear about the problems with the testing program. We can conclude specificity becomes incredibly difficult. In other words, we cannot state which direction the infection is now taking.

So, what would you need to see in the data that would worry you? 

Carl Heneghan stated the bits of data that he looks for are coronavirus triage data. The reason it is a sensitive measure of what is happening with 999 calls and 111 calls. The data tells him how many people are actively engaging in healthcare. Since they are semantically unwell. If you look at the 999 calls, they have flatline for about two months now.

The second set of data he watches is the admission data. The number of people in hospital. If you look at the admission data, it is continuing to come down and it is going down by over ten percent.

The third measure is death. Currently a downward trend for about three weeks. 

About the author 

Christopher Bird

Building your own Power App, BI solution, or automated workflow can be a mind-blowing experience. It can also be a nightmare. Particularly when you begin with a blank screen. My advice, get professional help as and when you need it. That's what successful people do.

{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}
Subscribe to get the latest updates