Why the jump in deaths on Tuesdays?

May 13, 2020 7:38 AM

The number of Covid-19-related deaths in Illinois jumps up by ~ 100-200% every Tuesday. Any idea why? It's extremely pronounced for the past three weeks. Before that, there are days with fewer deaths spaced several days apart, but it's not as clear a pattern.

I've been following this graph on the Sun Times.

I tried briefly looking at another state (NC) and it looked like there was a clear weekly pattern, but I couldn't easily see it in detail.

posted by amtho to Science & Nature (9 answers total)

Response by poster: My bad - the jumps are actually on Tuesdays, from very low numbers on Mondays.
posted by amtho at 7:39 AM on May 13, 2020

It could be from less testing on Sundays, with Tuesdays showing the lag.

My state has some reporting problem where Sunday numbers aren't the day's total for the past four weeks.
posted by NotLost at 7:45 AM on May 13, 2020 [6 favorites]

It's the date of reporting. Weekend deaths are mostly reported on Monday, compiled and released on Tuesdays.
posted by DarlingBri at 7:45 AM on May 13, 2020 [18 favorites]

Response by poster: Don't they report the actual date of death? I know people are super busy and stressed, but is is possible to see dates of death rather than dates of reporting?

Otherwise, looking at a graph like this isn't as useful at this scale. You'd be better off just seeing deaths per week rather than per day - any perception of trend or curve is essentially an illusion.
posted by amtho at 7:53 AM on May 13, 2020

It depends on the state. Massachusetts reports deaths by day of death as well as day reported. I agree that day of death is much more useful as there's a lot of noise in the day reported deaths due to things like weekend effects.

To get a sense of how deaths are trending if you can't get day of death data, I recommend using a rolling 7 day average - that way you should always have one of each day of the week in the average and normalize for things like weekend effects.
posted by peacheater at 8:02 AM on May 13, 2020 [7 favorites]

I do believe that this is a function of how the counting system works in your state, as well as how hospitals report. Not all hospitals use the same reporting methodology. Here in Washington, for example, there is a different number of hospitals reporting their ICU/hospitalizations of Covid every day. The state does a pretty good job of showing us all of the data to help us understand the numbers, but it's never a full picture.

Most numbers reported are as of midnight the day before, so if the number is high on Tuesday, that means that Monday had the highest numbers, which likely includes Monday + some number of deaths that did not get reported over the weekend.
posted by pazazygeek at 8:18 AM on May 13, 2020 [2 favorites]

You are correct by the way , that the day by day data is not useful for making accurate decisions! The smoothed 7 day average helps, and unfortunately just because a respected body chooses to release a particular graph doesn't mean it's the best way to reflect the underlying truth of the world.
posted by heyforfour at 8:31 AM on May 13, 2020 [2 favorites]

Don't they report the actual date of death?
It depends. Some places are reporting date of death (and, for cases, diagnosis date) - and others are doing report date instead. Some places don't have the capacity during a pandemic to do the additional data cleaning and processing to go from report date to diagnosis date.
posted by entropone at 10:00 AM on May 13, 2020 [1 favorite]

looking at a graph like this isn't as useful at this scale. You'd be better off just seeing deaths per week rather than per day - any perception of trend or curve is essentially an illusion.

Yes.
posted by PMdixon at 4:43 PM on June 12, 2020 [1 favorite]