1. Exquisite Tweets from @ChrisGiles_, @NickStripe_ONS

    PreoccupationsCollected by Preoccupations

    NEW: The excess UK deaths linked to coronavirus are likely to be around

    41,000

    as of yesterday, not the 17,337 of the hospital deaths announcements

    Here is the explanation for that estimate, which I will update daily

    1/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    The methods are simple and involve no forecasting of how the Covid-19 pandemic will progress.

    It is just an updating model - taking the best data from @ONS, the best concept (all cause mortality), and removing the delays in producing that data in a conservative way

    3/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    It is easily checked and every week, when the ONS data is published, will be a test to see whether it is on track or not.

    So far, it has proved remarkably accurate and we haven't published until now so we could check it was not giving a false steer

    4/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    We know that linked to Coronavirus there have been many excess deaths since mid-March - In England and Wales, there have been 15,161 cumulatively in the ONS data already published.

    Excess deaths is death registrations compared with the recent five year average.

    5/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    The England and Wales chart, compared with the previous 50 years shows this is not a normal seasonal flu

    Carl Henegan, professor of evidence based medicine at Oxford university, says: "I don’t think we’ve ever seen such a sharp upturn in deaths at that rate"

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    But the excess deaths are just as pronounced in Scotland - figures from the National Records of Scotland

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    But these data come with big delays. In England & Wales, the data published on April 21, refers to death registrations in the week ending April 10 and there is an average delay between death and registration of 4 days.

    So it relates to deaths in week to 6 April

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    For Scotland, the delay (yesterday) was 17 days (new data published today), for NI it was 14 days yesterday.

    So the excess deaths figure of around 17,000 across the UK in official figures is over 2 weeks out of date.

    The challenge is how to update to the present

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    This is what the lags look like for England and Wales

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    The updating mechanism that worked best was to use the daily reports of hospital deaths of those who tested positive for Covid-19 and scale them to the community with knowlege of the proportion of people who typically die in hospital (this has ranged between 46% and 48%)

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    Let's recap the assumptions made explicitly
    1) @ONS can count death registrations
    2) Excess all cause mortality is a good measure of deaths directly or indirectly linked to Covid-19
    3) There is a stable pattern between hospital Covid-19 deaths and all deaths

    That is all

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    Getting all the data together and thinking hard about lag strutures makes it all more complex, but the results have been tracking the cases, especially in England & Wales

    - note here yesterday's all cause mortality registrations were about 2,000 down due to the bank holiday

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    So, this is where the 41,000 figure comes from.

    - 17,000 excess deaths so far but that is out of date and the following two weeks had stable and high levels of hospital deaths

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    And if you assume a stable pattern in location of deaths, the numbers for those outside hospitals now suggests over 10,000 people have died in care homes more than would have been expected for this time of the year so far

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    I will update the figure every day using the latest daily hospital announcements. And there will be weekly shifts up or down according to the new official excess deaths numbers.

    Today, there will be new numbers for Scotland

    This official data will provide a check

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    I've kept assumptions conservative, so I would expect revisions to be more likely to be upward than downward. This was a deliberate choice.

    This is not supposed to be an alarmist exercise.

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    I will be transparent about whether the model is going wrong - that will help refine it.

    Remember, there is no forecasting of government policy or the disease here. Just an attempt to update lagged quality data with information we already have from partial data

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    I will not try to forecast the course of the pandemic.

    I am not an epidemiologist

    This is a statistical exercise

    ENDS/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

  2. Just to be clear Chris' model is, in his own words:

    "Just an attempt to update lagged quality data" [i.e. ONS] "with information we already have from partial data" [i.e. daily DHSC data]

    Both are official sources being used to drive his model

    Chris will assess how it performs

    Chris Giles @
    NEW: The excess UK deaths linked to coronavirus are likely to be around

    41,000

    as of yesterday, not the 17,337 of the hospital deaths announcements

    Here is the explanation for that estimate, which I will update daily

    1/

    Reply Retweet Like

    NickStripe_ONS

    Nick Stripe

  3. This, from the head of life events at @ONS is entirely accurate - all the data is official. No one is claiming stuff is being hidden. It's just that we know the up to date data is incomplete, so it's an attempt to fill in the gaps

    Nick Stripe @
    Just to be clear Chris' model is, in his own words:

    "Just an attempt to update lagged quality data" [i.e. ONS] "with information we already have from partial data" [i.e. daily DHSC data]

    Both are official sources being used to drive his model

    Chris will assess how it performs

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    NEW: Today's update of the @FinancialTimes UK death model suggests that by 22 April there have been roughly

    43,100

    excess deaths linked to coronavirus.
    The 2,000 increase from Tuesday reflects:
    a) 759 hospital deaths scaled up for community deaths
    b) new Scottish data... 1/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    The new Scottish data shows more excess deaths than the model had predicted for deaths up to 10 April registered by 13 April. These adjustments are to be expected because we are getting data in all the time and many of the model's assumptions err on the cautious side.

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    With some back revisions, the latest UK chart estimates looks like this

    Reply Retweet Like

    ChrisGiles_

    Chris Giles

    The free to read article explaining my methods is here ft.com/content/67e6a4…

    And the initial thread from this morning is here

    twitter.com/ChrisGiles_/st…

    ENDS

    Chris Giles @
    NEW: The excess UK deaths linked to coronavirus are likely to be around

    41,000

    as of yesterday, not the 17,337 of the hospital deaths announcements

    Here is the explanation for that estimate, which I will update daily

    1/

    Reply Retweet Like

    ChrisGiles_

    Chris Giles