What killed people in the 1st LD?

I’ve shared this plot a few times on twitter, and it has seen limited praise, and crucially understanding. What’s depressing is, of all the research I’ve shared, this is likely to be the most “mic drop” moment yet in the debate over whether excess nonC19 deaths in the first wave were caused by lockdown, or just untested C19.

Let me explain.

My original aim in building the plot was to line up separate sources of data to cross validate them, but I ended feeling my own jaw drop as I brought the data together. These sources were:

• blue lines: ONS doctor certified deaths, and ONS excess deaths, shifted back 19dys to imply infection.
• red lines: cases as measured by PHE using PCR tests only, split <70 and >70 and shifted back 10dys to imply infection.
• black line: daily antibody rates as measured by the REACT study, shifted back 21dys to imply infection.

In summary the rate at which those who survived accrued antibodies seems to move in lock step with the rate at which we has excess deaths, suggesting that they were all C19.

The hard part was reasoning through how to line up the data that is all on different scales. Let’s consider each in turn.

The confusion with PCR testing, as I discuss in this, is the govt was not trying to build a diagnostic architecture to find symptomatic cases, it was trying to build a search architecture to find spread, which would include finding pre-symptomatic, and post infectious, infections.

Add to this complexity, testing rates did not achieve their maximum till the 2nd/3rd waves, with the green line in the bottom plot in this being my best estimate at the proportion of true infections needed to explain deaths, never getting above about 1/4.

Crucially, this varies over time, as testing has increased 500x, and t&t performance degrades with prevalence, finding fewer, making lining PCR implied infections tricky.

My conclusion was the best thing to do was scale the 3rd wave cases, <70 and >70 separately, to line up the peak heights with those of the 4x bigger serology numbers for the REACT and ONS serology implied infections data.

In doing this, the main aim is to assess the degree to which each age group was being under tested in the 1st wave, and it is clear the govt did prioritise testing the those >70, but both were hugely under tested.

Now the remaining questions are — why are excess deaths above C19 labelled deaths in the 1st wave? And, below in the 2nd/3rd?

The latter is covered in this piece, but is best summarised in this example. Essentially, C19 will have taken most normal deaths early from this year, then next, etc, so the more C19 deaths we’ve had, the more deaths now will see under trend, confusing matters.

Which leaves us with just trying to understand what happened in the 1st wave, and this is where daily infections as implied by antibodies come into their own. Unlike PCR testing, which is not definitive of infection, these make it very clear that an immune system has seroconverted generating an IgG response.

To give you an idea of scale, for the 138k deaths in England & Wales, the study has estimated 15.8 million seroconverted infections.

This is the denominator in the plot earlier estimating that only 1/4 of real infections are being found by PCR testing, incidentally this provides a sanity check that PCR cases are not overestimating infections, as they have never been more than 60% of antibody counts, an underestimate.

But, how do we scale these to put on the plot?

Well, I went with the simple answer, like PCR implied infections, of scaling these to match the height of the 3rd wave for both ONS C19 death implied infections, and PCR implied ones.

When we do so, the most striking thing that comes out, is that the 1st wave lines up with excess death implied infections, not ONS C19 certified ones.

This is huge, as it suggests all the nonC19 labelled deaths in the 1st wave, ones militant sceptics believe were LD induced deaths, were just C19 deaths, and in flying blind at that time, our system just missed them.

I suggest you spend some serious time kicking the tyres on this as I have, it doesn’t matter how much I mull it over, I can’t think of a flaw in the reasoning.

Sure, testing is involved in clinical judgement, so more testing will mean more C19 certs in the 3rd wave than the 1st, and it is possible some of these are mislabelled, but by scaling the antibody counts match the 3rd peak height, we we are making the 3rd to 1st wave comparison agnostic to this problem. We’re simply asking the question, how many ONS certed deaths would we have had in the 1st wave using the same diagnostic criteria as the 3rd wave.

With the answer being, pretty much all of them.

It also means that, contrary to conspiracy theories, doctors didn’t overmark in the 1st wave, they just struggled to cope.

I really feel for them, in the 1st wave they planned for a far worse spike, shutting far more services than they should have, and we bash them for it.

In response, and in a real testiment to their duty and resolve, by the 3rd wave they handled 2x the admissions, never saw the hospital fatality rate go about 2/3 of the 1st wave, maintained treatment levels for all nonC19 care to the highest possible level they could, despite a reduced bed capacity brought about by isolation procedures, and higher staff/patient ratios needed for ICU, and this despite 6% more staff.

Inspirational.

As to what antibodies say about the 1st wave excess deaths. For sure nonC19 deaths happened in the panic, but in aggregate, I challenge anyone to explain the data differently. I’m all ears.