Friday 10 April 2020

Covid Stats


Yesterday I saw an article showing the projected deaths from Covid, by country in Europe, as forecast by the Institute of Health Metrics and Evaluation, an American research foundation initially seed funded by Bill and Melinda Gates.



Doubt has been cast on the forecast for the UK by Imperial College's Professor Neil Ferguson, who maintains the UK's forecast is actually between 7k and 20k. However, even so, this gives a skewed impression of how countries are coping, as population sizes differ considerably.

I therefore took the above numbers and amortised them over the different populations to derive the chart below - allowing two extra UK columns for the 7k lower (FL) and 20k higher (FH) forecasts of Ferguson.


As can be seen from the orange line, representing deaths as a percentage of overall population, Spain and Italy are forecast to be virtually on level pegging, with France lower and Germany very much lower. Using the Institute for Health Metrics and Evaluation figures from the upper chart, we're not forecast to do at all well compared to the rest of Europe, but the 20k Ferguson higher figure forecasts us to be in the same area as Italy. The 7k Ferguson lower limit shows us forecast to do much better and slightly better than Germany.

Then there's the metric used by the IHME and Ferguson. One assumes the same  counting methodology is used for each country for the IHME forecast, but there are reports that the UK only uses hospital deaths, excluding care homes, and Ferguson's forecasts are therefore skewed. But then where does the IHME get its UK data from, if not the UK? - I just don't know and the IHME site does not state this.

Germany includes care home deaths, but that may be because their testing regime is far in advance of the UK's. Determining the cause of death of care home residents in the UK is fraught with difficulty, as pneumonia is the old person's friend and bodies are not taken to hospitals for autopsies unless there's suspicion of foul play.

It does, however, go to show that comparing raw death data without the nuance of population size can lead one to false conclusions, as can any comparison without a common framework. Doubtless other factors should be taken into consideration, such as percentage of elderly, etc., but that data is not available.

It's strange how when we're presented with a graph or chart, we're conditioned to believe it without looking at the assumptions behind it, but when the same data is provided as a narrative, we're more likely to question it.


No comments: