Read on Twitter
I've been haranguing political scientists about good and bad uses of COVID death data. This seemingly minor intervention by Belgium's health minister provides a great example of why this matters -- so hear me out. Thanks to @rdanielkelemen for bringing it to my attention. 1/ https://twitter.com/PietEeckhout/status/1281123494936358912
The central point is that government reporting of the same type of deaths differs across place and time, even in normal, non-pandemic circumstances -- and in ways that create measurement bias. This thread is about why, and why it matters. 2/
Not all governments have the capacity to detect all causes of death at any given moment in time. When there's a novel pathogen (e.g. HIV, zika, swine flu), some governments will have the ability or willingness to test sooner than others, creating a lack of comparability. 3/
But even given the same ability to detect the cause of a death, doctors and coroners may follow different rules and/or norms about attribution of cause of death when there are multiple possible causes, leading to under- or overcounts of deaths from certain causes. 4/
And different populations may be included when death counts are reported (e.g. some countries include the deaths of people who live in prisons, nursing homes, or mental institutions, while others don't). 5/
Moreover, how long it takes for a death to be counted in a vital records system varies. In some places deaths are recorded every day; in others, it could be once or month or even longer. 6/
(Tangentially, but this is the topic of my new book project, so I can't resist: Govts also record different information about the dead, resulting in varying capacities to detect inequalities by SES, race, etc. in deaths from certain causes, or even in all-cause mortality.) 7/
So if you want to compare the incidence of a fast-moving, novel cause of death across place (or within place over time), you need to be very sure that you're using data that allow you to compare apples to apples. And reported COVID19 deaths are not that -- at least not yet. 8/
At this point you're far better off using excess mortality data like those that have been standardized across country by http://Euromomo.eu . Yes, these data include deaths that are not caused by infection by sars-cov-2 -- but may still be attributable to the pandemic. 9/
(E.g. people dying of heart attacks at home because they are not going to the emergency room, people mis-categorized as having died of pneumonia, etc.) 10/
This matters not only for getting the story right from a scientific point of view (though that's surely important). Let's come back to Belgium for a minute to see why else it matters. 11/
From the early days of the pandemic, Belgian officials got pilloried for doing a bad job with COVID19. They weren't. They were just counting differently: more deaths of more kinds of people were categorized as COVID19 deaths than in other countries. 12/
(I suspect, though I can't be sure, that part of the reason Belgium did a better job counting COVID19 deaths is that they were relatively late to automate their vital records systems. 13/
As recently as the early 2000s, there were long delays in death reporting in Belgium because record-keeping was still being done by hand in many localities. I learned this tidbit while doing interviews for my book Regimes of Inequality https://tinyurl.com/yy5xl2v9 .) 14/
Regardless of why they did it right, it's clear that we should have been looking to Belgium as an example of *what* to do right: doing one's best to discover the real numbers of deaths from a novel cause, and being transparent with the world about it. 15/
Instead, Belgium got a black eye for its high death counts in the international media, among the all-important consultants, and (albeit temporarily) within the int'l orgs that disseminate best policy practices. THIS IS BAD FOR PUBLIC HEALTH. 16/
It matters who gets blamed for deaths. It matters when communicating the truth gets punished. Which is why it matters what data political scientists use in a fast-moving situation where millions of lives are at stake. FIN. 17/17
