Covid vs 1918 Flu

[watu05]

Covid has now caused more excess deaths than the 1918 Spanish flu. How this is measured.

How Covid Upended a Century of Patterns in U.S. Deaths (Published 2021)

The pandemic brought a record uptick in deaths, interrupted a downward trend in death rates and spurred the highest death rate above normal ever recorded in the country.

www.nytimes.com

 

[URedskin54]

This is pretty classic "lying with data." If 1% of the population normally dies in a given year of all causes and an illness brings it to 1.5% that's 50% above normal. If 50% of the population normally dies in a given year of all causes and an illness brings it up to 60% that's only 20% above normal, but it's still 20x worse than the other scenario. Not great stuff from the times.

 

[watu05]

This is pretty classic "lying with data." If 1% of the population normally dies in a given year of all causes and an illness brings it to 1.5% that's 50% above normal. If 50% of the population normally dies in a given year of all causes and an illness brings it up to 60% that's only 20% above normal, but it's still 20x worse than the other scenario. Not great stuff from the times.

Which 'lie" are you pointing to? I understand your point; where is it in the article?

 

[lawpoke87]

Any data point that isn’t tied to % of the population is fairly meaningless not to mention misleading.

 

[URedskin54]

Which 'lie" are you pointing to? I understand your point; where is it in the article?

Like lawpoke said they've chosen a set of data that probably can't be meaningfully compared over time and it is misleading. The denominators are wildly different. % of deaths above normal compared to 1918 tells us a lot more people died of other things in 1918, but doesn't really tell us anything about how bad that pandemic was. The one stat that is probably a fair comparison (% of population who died of the illness), they chose not to include. If they wanted to argue that % of excess deaths is meaningful in some way because medical care has improved so much over time, and therefor we should presume that it would reduce the severity of pandemic deaths in a similar way as deaths of other causes I'm open to that. But they aren't making that argument.

 

[Gmoney4WW]

Which 'lie" are you pointing to? I understand your point; where is it in the article?

Tisk, tisk Watu, that is a Shondiced statistic.

 

[watu05]

Like lawpoke said they've chosen a set of data that probably can't be meaningfully compared over time and it is misleading. The denominators are wildly different. % of deaths above normal compared to 1918 tells us a lot more people died of other things in 1918, but doesn't really tell us anything about how bad that pandemic was. The one stat that is probably a fair comparison (% of population who died of the illness), they chose not to include. If they wanted to argue that % of excess deaths is meaningful in some way because medical care has improved so much over time, and therefor we should presume that it would reduce the severity of pandemic deaths in a similar way as deaths of other causes I'm open to that. But they aren't making that argument.

Thanks, I"ll reread.

 

[watu05]

Like lawpoke said they've chosen a set of data that probably can't be meaningfully compared over time and it is misleading. The denominators are wildly different. % of deaths above normal compared to 1918 tells us a lot more people died of other things in 1918, but doesn't really tell us anything about how bad that pandemic was. The one stat that is probably a fair comparison (% of population who died of the illness), they chose not to include. If they wanted to argue that % of excess deaths is meaningful in some way because medical care has improved so much over time, and therefor we should presume that it would reduce the severity of pandemic deaths in a similar way as deaths of other causes I'm open to that. But they aren't making that argument.

I sent your comments to an epidemiogist.

My take on the NYT article overall: This was a back-of-the-envelope estimate and I felt the authors appropriately qualified the data limitations and findings (at least for my more informed read, perhaps not qualified enough for a more general NYT audience). While there is kind of an implication that they are directly comparing the two time trends (2020 pandemic vs 1918 pandemic), they are not directly making this comparison with hypothesis testing/statistical significance. They are performing a simple descriptive analysis for both time points and qualitatively/narratively describing differences. I am ok with this, although there were a few statements like "...even surpassing the calamity of the 1918 flu" that are overstating their findings. This analysis is answering a simple question of "how much higher was the mortality rate in 2020 than would have been expected based on the prior 5 year linear trend (~2014-2019)?" and then asking the same question for the 1918 period. The two separate analyses are in-and-of-themselves interesting to me, although I wish they didn't put them together in the same conversation. By placing the two analyses "next to" each other and describing them together, the readers are left to jump to conclusions about their relation to each other (such as assuming something like "the excess deaths in 2020 were 4% higher than in 1918" -- which is overstating the results).

I would rather see this analysis performed by the CDC or other epidemiologists since they likely wouldn't have performed such a simple analysis (this one used simple (unadjusted) linear regression to estimate "expected" mortality rate based on the prior 5 years, which is overly simplistic). My problem with some articles like this is that NYT analytic team (and other generalist data science groups) feels emboldened in the COVID-era to produce something close to an epidemiologic analysis using publically available data but they don't adequately inform the public about the limitations to interpretation. For my personal level of understanding, the qualifications/caveats about data were enough for me to appropriately interpret the results, but I don't feel they are caveated enough for the general public.

My take on this particular comment: I believe this commentator misunderstood the interpretation of % excess deaths. I agree that the data available from 1918 is not comparable to the data available in 2020 (for numerous reasons, many of which were discussed by the authors). This is why the authors performed separate analyses for the two time points (and 5 years prior), qualitatively described them, but did not directly compare them. Again to say 2020 had % excess deaths that were 4% higher than in 1918 would be dramatically overstating the results -- in fact from this analysis we can't even really say that excess deaths are higher than in 1918 since the authors didn't account for confounding factors. The fact that the denominators are different and the datasets are very different (e.g., reporting practices were different between the two eras, some states weren't included in the 1918 analysis, causes of death were different between the two time points) doesn't really matter in my opinion for this analysis. We're talking about the % above-normal mortality rate for 1918 compared to the prior 5 years and(separately) the % above-normal mortality rate for 2020 compared to the prior 5 years. Perhaps this commentator thinks the "% excess deaths" is a proportional mortality for % of overall deaths caused by flu in 1918 and % of overall deaths caused by covid in 2020. This analysis would be flawed for the reasons mentioned by the commentor, but I disagree that the trend analysis is inherently flawed.

 

[URedskin54]

I’m not going to try to have a conversation with someone via third party message board posting, but it seems like if this were a conversation between he/her and I we’re taking past each other. What they’re saying doesn’t quite address the critique, or rather they make a distinction between data sets(% above normal mortality and “proportional mortality for % of overall deaths”) that ultimately would have the same flaw in comparing between 1918 and 2020 for similar reasons.

 

[rivalcane]

whats the difference in covid in 2021 vs in 2020. one year.