Extending excess deaths

[J-Dunn]

The new excess death data is a great addition.

Would it be possible to extend it backwards for context?

The tail of last year's flu deaths overlaps slightly with the start of COVID19 and it is important to see shape and magnitude of recent years' flu excess deaths to understand relative significance of COVID.

Excess deaths has been the classic way to assess the impact of influenza epidemics and there is a large peak most years. Seeing at least the 2019 excess deaths would help to put the magnitude of COVID in context.

About 1/80 of the population dies each year of one cause or another ( assuming a median age of death of 80 y.o.), many of diseases such as influenza or covid which those close to the end are unable to resist. Excess deaths towards the end of this year may well go below average since the baseline is the typical "average" flu death toll.

Having a broader window is essential to correctly viewing this data in context. Hopefully that will not be too difficult to add since you have already done the hard work of finding the relevant data sources.

This is an excellent project and fills some of the gaps left by ECDC only giving "cases" and deaths. Great work, many thanks.

[edomt]

@J-Dunn We have now extended our excess mortality file with more columns with data from previous years, which should let you run the analysis you're describing: https://github.com/owid/covid-19-data/blob/master/public/data/excess_mortality/excess_mortality.csv

[J-Dunn]

Thanks for being so quick in picking this up. That's very helpful.

It would make sense to extend the table to week 52 to include the full year for the previous years data. Because the current year stops in Sept, we only 3/4 of the historic data for each year. Even if we don't have collated data for recent weeks in the file, it would still be good to have complete years for the background seasonal average. We could get then put COVID fatalities announced via national sources against the excess deaths for the time of year.

I can guess by the magnitude that this is excess deaths for 7d period but this does not seem to be documented anywhere. It may be good to document what the date column represents. Is it the last day or mid point of the 7d period; a Friday ? Are the source data arranged as week 1 , week 2 , etc. ? It's detail but it would be nice to have this documented.

You may wish to add a line to the change log as well.

Thanks again.

[edomt]

Dates refer to the last day in each reporting week for most but not all countries. For a more detailed description of the HMD data, including week date definitions, see the HMD metadata file at https://www.mortality.org/Public/STMF_DOC/STMFmetadata.pdf.

[J-Dunn]

Many thanks for the reply and extra info.

Is it possible to include the rest of the historic data? It is not very logical to crop each of the previous years' data at the current limit this year's data. An important part of the context of the current epidemic in each country is to see the peak in excess deaths in previous years which has been traditionally attributed to influenza.

[edomt]

We're working on adding this.

[J-Dunn]

Many thanks for being responsive to these requests. They make the dataset much more useful.

This is what the French epidemic looks like in the context of previous years' flu spikes:

It's a bit of shame it's not more up to date at the end, we can't really assess the current "second wave" but I suppose you have to wait for late comers.