I bought some death data

"Why are you so hopeless with Covid, Victoria?"

I bought some death data from the Australian Bureau of Statistics (ABS).

Data Requested

I commissioned a compilation of all the deaths, by week of occurrence, that have any association with Covid-19.

This means,

• all the deaths attributed to Covid-19 as an underlying cause, plus

• all the deaths attributed to Covid-19 as an associated cause, plus

• all the deaths with underlying or associated causes or file notes referring to Covid-19 in the categories

  • Personal history of COVID-19 (U08) (254 in 2022 so far),

  • Post COVID-19 condition (U09) (183 in 2022 so far), or

  • Multi-system inflammatory syndrome associated with COVID-19 (U10) (3 in 2022 so far)

I asked for the deaths by State, but not Tasmania, the Northern Territory or the ACT because the ABS said that, as they are small, they’d have a lot of randomization-for-privacy error.

Data Validation

I’m at the data validation stage. Share it with me. Look at the graphs below while I sing that Sesame Street song “One of these things is not like the others”.

The graphs compare Accumulated Covid Deaths (per capita (per 1000)) on covidlive.com.au (grey) vs the ABS “Anything Covid” Death data I commissioned (blue), for each State (NSW, Vic, Qld, SA & WA).

Hopefully you spot it. There’s a second place consolation prize if you pick the wrong one. Sorry it’s so small but I want them lined up..

Did you pick the second graph, Victoria, as being “not like the others”, with the grey CovidLive line being a LOT higher than the blue ABS line? Here enlarged with the actual death numbers on the y-axis, with thousands more covid deaths attributed to our names than we perhaps deserve:

Have you ever compared the deaths by state on CovidLive and thought “Why are you so hopeless with Covid, Victoria?”

All I want to say here is that maybe we’re not that hopeless. Maybe the Victorian Health department is hopeless. Maybe they’ve been amplifying the deaths through to 2022 (+29%) and +38% through to 2023 (albeit with the various reporting, ABS & coroner delays).

I was so shocked. I didn’t at first believe the ABS numbers. CovidLive numbers were my gospel truth. I knew every nuance of them, adjusting and smoothing, allowing for delays and reapportioning bulk notifications.

I asked all the questions… The ABS was quite surprised themselves to see the difference. Was there a huge delay at the Victorian Coroner? The ABS didn’t think so - they would’ve heard if there was a signal that big. Is there a system that isn’t merging properly? The ABS will have an ask-around, after the Christmas-New Year break.

I’ve seen that Victoria can be different. Recently there was a project to retrospectively bring in suicide deaths that had been hiding away somewhere. Maybe something like this will happen and we’ll discover the missing covid deaths six years from now as quiet retrospective additions to the tables.

Consolation Prize

Did you pick Western Australia? The fifth graph.

Unlike the other three states they look like they HAVEN’T been publicly reporting all of their Covid-related deaths, down by 16%. Unless all the people in Category U09 live in Western Australia, and their long vax from their extreme boosting is being called long covid, or something? Or maybe they don’t report if covid is clearly only an associated cause?

Reasons - Complexity, Bias & Beauty

I’d seen enough to know that the Total Covid numbers could tell some very clear stories, and that they would be informative and useful.

However, the CovidLive numbers that I was using as a surrogate required a lot of manipulation, reapportionment, and allowance for reporting delays of different lengths at different times to have them in a reasonable form.

There were two problems with using this data. The first is that I would have had to explain and justify what I’d done, which would’ve been tedious and complex. The second, is that such manipulation could leave open the space for questions about whether I had shaped the data to fit a hypothesis.

So to avoid the complexity of explanation, to avoid question of bias, and also to see the inherent mathematical beauty of real numbers, the solution was thus to order the data from a higher and independent (and hopefully more accurate) source.

I hope these numbers will serve us well.

The End.