Thread by @dgurdasani1

Ok, so worth really looking closely at the @NatureMedicine paper that's been cited widely on the impact of COVID-19 on children's deaths.
This has several key limitations, and should not be cited in the way it is being🧵

The analysis is between March '20-Feb '21. Why is this important:
-delta was not dominant
-there were 3 lockdowns during this period
None of this is relevant now with delta, no mitigations, the highest infection rates we've *ever* had in children (4-5% even post- half-term)

The authors carried a review of 61 deaths in children who were positive for SARS-CoV-2/had PIMS-TS and come to the conclusion that only 25 were due to SARS-CoV-2. They conclude that only 25/61= 41% of all children who died 'with' COVID-19 died 'from' it.

"identifying whether SARS-CoV-2 was listed as 1a (the direct cause of death) on the Certificate of Cause of Death and whether the clinical course described was typical of SARS-CoV-2 infection." For the rest, a review was carried out to determine whether SARS-CoV-2 contributed.

It depended on a subjective review of case reports by 3 experts. All reviewers only agreed on only half of the cases they reviewed, indicating the really subjective nature of review. And consensus still could not be reached on 9 which needed further review.

For deaths from PIMS-TS, a multi-inflammatory syndrome that can occur after COVID-19, they relied on diagnoses being in the electronic record. Given this syndrome was only identified and described in April 2020, this would almost certainly have been underestimated.

In fact the authors note the difficulty in identifying PIMS-TS related deaths, because a diagnostic code for this was only generated in Nov 2020. Different codes were considered to represent this before, but it's very likely cases would've been missed at least early on.

The ONS data from the same period on deaths in 0-19 yr olds (0-18 yr olds not publicly available) "involving COVID-19" are 33 (if their end point is 28th Feb) for England and Wales.

These figures are roughly consistent to me with what they may have found, although I can't be completely sure because the exact date range isn't clear, the ONS doesn't provide data separately on PIMS-TS (which would've been underestimated by the ONS too), & the age range is diff.

But what this goes to show is despite what may be a reasonably thorough, if subjective review of the evidence (as the nature of this will always be, even with death certification), it is not generalisable to the current data at all. And it suggests the ONS estimates are accurate.

ONS deaths 'involving COVID-19' include similar classification to what the review has produced- deaths directly due to COVID-19, and with COVID-19 as a contributor cause (also on the causal pathway). Both these would be attributable to COVID-19.

Unfortunately current ONS estimates, suggest a very very different picture - post-delta, and since mitigations have been mostly lifted in schools, which has meant mass infection in kids. ONS currently shows that 68 deaths till end of Oct from COVID-19 in 0-19 yr olds

These are of course underestimates, because death registration gets backfilled & recent weeks are underestimated. This doesn't include the last 2 weeks where there have been deaths in children as per PHE data, at least some of which will be back filled in. They don't include PIMS

These data are lagged, and concord very well with PHE data of COVID-19 deaths within 28 days of a positive test for recent waves, which means once data are backfilled, we have ~101 deaths in children up to now in England & Wales:

https://twitter.com/TigressEllie/status/1458835593992015878?s=20

By contrast with what the paper says which may have been true at the time, the vast majority of deaths with COVID-19 in children are very sadly due to COVID-19 now. There were 94+3=97 deaths in PHE 28 day data compared with 68 in the ONS which is 70% of with COVID being 'from'

If you consider only this year, this proportion is even higher >80%, which suggests that currently most deaths occurring within 28 days of a positive test have SARS-CoV-2 as a cause (direct or contributing cause)

Ok, let's talk about pre-existing conditions now. The paper suggests that 3 of four deaths were in children with 'pre-existing conditions'. The most common pre-existing condition cited is 'neurological condition' which includes learning disabilities and mental health conditions.

Obesity and trauma are also included.
For PIMS-TS- which is likely to have been underestimated, 2 out of 3 children were deemed not to have an 'underlying condition'

What I find most egregious and misleading in the paper is putting down SARS-CoV-2 deaths with a population denominator - suggesting the rate is 2/million. This is wholly incorrect, because the no. of deaths depend on exposure, especially in a pandemic & this number isn't constant

Indeed, if you look at deaths from COVID-19 based on the latest ONS data this would be ~3 times higher, even with outdated data. They then compare with *all other causes of deaths*, as if this is in any way a meaningful comparison- all with population denominators!

Anyone re-assured that so far the death rate in kids has been 6/million *population* (current ONS) with lots of potential to rise? If all children were infected this would be 700 deaths as per the papers IFR estimate (even Whitty said most children will inevitably get infected)

And of course children could get infected again and again, so this isn't even necessarily an upper bound estimate of what could happen. So the infection fatality rate of 5 per 100,000 may seem small, but it's not that small when you allow almost all kids to get infected.

Just for comparison, 30 children died because of flu in 2019 (no lockdowns or mitigations)- comparatively we've already seen many more deaths with COVID-19, despite 3 lockdowns - more recently as mitigations were removed.
ons.gov.uk/aboutus/transp…

Of course the study doesn't look at the most common impacts on children which are long COVID related persistent symptoms (increasing rapidly over time), impact from being orphaned, living with someone with long COVID, and educational disruption from lack of mitigations.

Overall, this study has little value at the current time except confirming the ONS data are robust which of course suggests death rates have increased in children- are much higher than other childhood illnesses, & will increase further unless action is taken.