Wait a minute – what does ‘peak’ mean?

If there was ever a time when you need to listen to every single word, this might well be it.

The government and mainstream media have been claiming and reporting that we passed the peak on the 8th April – take this compelling article by the BBC: https://www.bbc.co.uk/news/health-52361519.

And yet over the weekend, the few of us who were raising questions about that have been joined by everyone, but in a slightly subtle but critical change in the language: the 8th of April is no longer simply the peak, it is the “peak in hospitals”.

What might be less clear is why this shift has happened.

As we said last week, the data – if correct – strongly contradicted the claim that lock-down was working. We made this statement based on analysis of hospital deaths and new cases.

Even Hospital deaths did not show clearly a downward trend, but more importantly the new case data was certainly at best remaining constant and not falling. If the number of cases was continuing at the same daily rate, then it is reasonable to assume that the rate of death in the subsequent weeks would also continue at the same rate. That is not consistent with us having passed the peak.

What we saw in last week’s data that led us to those conclusions was partially explained over the weekend by the Office for National Statistics (ONS). The ONS released data up to and including the week ending 17th April. This data is important because it includes non-hospital deaths and identifies the deaths that mention Covid-19 on the death certificate.

The chart below shows five data progressions in weekly totals:

• The grey and yellow bars show the total weekly deaths for weeks in 2020 and the average weekly deaths for the corresponding week in the last five years respectively. Both of these data sets are for England & Wales only.
• The blue line is the difference between the above bars, and therefore represents the ‘excess deaths’ – the number of additional deaths this year compared to the five year average.
• The orange line show the deaths that mentioned Covid-19 on the death certificate. The fact that this is below the blue lines illustrates the problem that Covid-19 is not explaining all of the excess deaths – likely both because there are excess deaths for other reasons, and because not all Covid-19 deaths have been documented as such.
• The black double-line shows hospital deaths. Note that this is UK data, which is approximately 10% higher than the England & Wales data shown by the other four data series, and so the corresponding England & Wales curve would sit a little below this line.

Let’s first deal with the contrast with the NHS England data charted in the BBC article above – and widely used to evidence the 8th April peak. That chart seems to show a clear peak on the 8th:

There are two things to notice about this chart:

1. The chart clearly calls out the fact that the data for the last five days on the chart is “likely to change”. If that isn’t a red-flag to tell us we should be showing it I am not sure what is. Cover those five days with your hand, and it is much less clear that the “peak” is a peak at all, as opposed to a statistically reasonable variation.
2. More importantly, these daily figures obviously fluctuate and make it difficult (and for most people subjective) to identify a trend. Compare this with the first chart above that shows weekly data. Wait a minute? Is that right? Yes – there were more hospital deaths in the week ending the 17th April (5,751 deaths) than in the week end 10th April (5,057 deaths, the week that supposedly contained our peak).

Of course, it is possible that the peak hospital deaths were on the 8th April even though the following week had more deaths. If the rate of increase of daily deaths before the 8th April exceeded the rate of decrease of daily deaths after the 8th April by a big enough margin, we would see the kind of data above.

However, that should surely have raised doubts – as it did for us.

But now there is no doubt: we have the ONS data, and this really clearly shows us that in the week ending the 17th April there were 9,496 deaths mentioning Covid-19 on the death certificate in England & Wales, compared to 6,887 the week before.

If we look back the week before that (week ending 3rd April), there were 3,801 deaths. That tells us that the weekly rate of increase in deaths fell only slightly from 3,086 additional deaths between the 3rd and the 10th to 2,609 additional deaths between the 10th and the 17th. It is almost impossible to construct a set of data that has this characteristic and at the same time supports a peak daily death date of the 8th April.

Hence the subtle shift from a narrative about “the peak” to the far less loaded “a peak in hospital” deaths.

The data trend for weekly deaths is almost linear growth between 27th March and 17th April. However, this isn’t very useful as it is too small a sample to draw conclusions. Over the next couple of weeks, this data will become more useful.

So what should we learn from this?

Firstly, listen to the words. They matter. We all like certainty but no one has certainty, so instead of naively believing the narrative that is being promoted, take a moment to consider the old adage that if it sounds to good to be true, it probably is.

Secondly, the narratives we are being fed are not lies. Nor, in my opinion, do they deliberately (in the majority of cases) seek to mislead us. They are over-simplifying complex data and using a visualization that inadvertently misleads. Which then gets further simplified as we re-tell the stories around our family dinner table or via our Zoom nights-in with friends. But with an unfortunate consequence – a shared belief that we are turning a corner, things are getting better, and potentially that we can relax our adherence to lockdown. This is not supported by the data.

Thirdly, it is said that a picture tells a thousand words. In these times, when you see a picture, make sure you have read and understood forensically the thousand words that accompany it. Otherwise, you might find that the thousand words you have inferred weren’t the one intended by the publisher.

So, as a former leader so eloquently put it:

Now this is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning.

Winston Churchill, November 10th 1942

It may be the end of the beginning in the sense that we are starting to get ever-so-slightly more reliable data. We are starting to see the scale and extent of what is happening. We are just beginning to understand the “why” of some of this.

This is positive – we are making progress, and there is no doubt that we will come through this.