Not exponential
Firstly, does nobody else find it alarming that Dominic Cummings had retained a photograph of a whiteboard from a brainstorming session with the Prime Minister and others and just published it on Twitter?
There will be endless claims of who knew, or should have known, what and when; how they interpreted and/or should have interpreted it; what the actual facts were at various stages; and what subsequent interpretation of the facts reveals. From these we'll have multiple diverging opinions about what would have happened if only: <fill in whatever scenario springs to mind>. I'm certain that prevailing opinions about all of these will change over time. In 20 years Wikipedia (or whatever it evolves into) will still be torn over what actually happened and what was and should have been done about it.
I'm happy this process has started and I'm hoping it will include all the advisers (and former members) from SAGE and reveal any dissenting opinions. People are becoming more concerned about what should have been done than what will be done next. In other words they think it's over.
I like many others was engaged in curve-fitting available data from the first UK lockdown onwards. I was looking at data published by ONS (ie for England and Wales), by PHE and the data collated by the UK Government 'dashboard'. I would hope that advisers to the government and health services had quicker access to the data than we as members of the general public had.
To gain context I retrieved ONS weekly mortality data going back to 2010 and population estimates and projections. For added context I retrieved ONS annual main cause of death data (as reported on death certificates) via the NOMIS system going back to 2013. In addition I looked at ONS' excess winter mortality figures back to 1950 and compared that with annual death rates for the standard population subsets of male/female, age: <1, 1-14, 15-44, 45-64, 65-74, 75-84 and 85+.
The first wave of Covid-related deaths (as reported on death certificates) in England and Wales up to September follows a Gompertz curve (cumulative deaths). It is such an exact match as to be astonishing.
A general formula for a Gompertz curve is: a * EXP(-b * EXP(-c * time))
The ONS data up to 10 April 2020 (published 21 April) fits a Gompertz curve with the formula: 6.22E+04 * EXP(-1.23E+02 * EXP(-6.25E-02 * <day no>)) where <day no>=1 is 30 Jan 2020. For context 10 April 2020 would be <day no>=72 by which time ONS was reporting 15,785 deaths. The formula 'predicts' 62,200 deaths.
Even by plotting just the first thousand deaths (ie as at 23 March 2010, <day no>=54) it was obvious that we were not following an exponential curve.
The ONS data up to 4 September 2020 (published 15 September) fits a Gompertz curve with the formula: 5.17E+04 * EXP(-7.14E+01 * EXP(-5.66E-02 * <day no>)). ie it predicts 51,700 deaths. As I said above the fit is astonishingly close.
Rather interestingly the data from October onwards is chaotic and does not fit a Gompertz curve at all. I'm sure this will attract different interpretations but it's almost as if multiple different things were happening rather than a single-cause epidemic.
Biting the hand that feeds IT
