6 thoughts on “State Level Excess Mortality Updates – Sept 8, 2021”

1. Thank you again

   LikeLike

2. To follow up on my previous post regarding BMI and more generally co-morbidities. This graph (https://www.cdc.gov/obesity/data/maps/brfss_2019_ob_white.svg) doesn’t address all co-morbidities but it does address obesity which is itself a co-morbidity but also highly correlated with other co-morbidities so a decent proxy. Full site: https://www.cdc.gov/obesity/data/prevalence-maps.html

   Note a pretty strong correlation between your graphs and the obesity prevalence graph from the CDC.

   LikeLike

   • I’d be curious to just see a simple scatter plot of states’ obesity prevalence on one axis and Covid excess mortality on the other. I’d bet there’s an R2 above ~0.5. Or a dual factor correlation with obesity and fraction of population over 65 years old I bet would explain ~80% of the excess mortality.

     LikeLike

     • Hmm it seems while logging in my first post was lost. Well to try again… Thanks for the post and I’m glad someone is bothering to post this information. What I find frustrating about most of the analysis these days is whenever different states are brought up it’s usually done in order to attempt to assess the efficacy of different non-pharmaceutical interventions. But the most rudimentary of analysis would have to correct for underlying differences of the states first before trying to imply a cause effect relationship between interventions and mortality. Two of the the huge factors would obviously have to be age and co-morbidities.

       A question on the graph you posted and the age adjusted mortality from the CDC. The graph doesn’t really make sense to me. One would expect the age-adjusted mortality to roughly straddle the raw mortality when looking across the states. For example Florida’s age adjusted mortality (since it’s an above average aged state) would be lower than it’s raw mortality number and then a younger state like California should have a higher age adjusted mortality than it’s raw number. The fact that all the age adjusted mortality numbers are lower than their respective raw numbers is very odd to me. Do you know how they are actually calculating it. The standard way is discussed here: https://www-doh.state.nj.us/doh-shad/view/sharedstatic/AgeAdjustedDeathRate.pdf.

       LikeLike

   • Oh man, a lot of good questions here. Apologies in advance if I miss anything.

     First, the age adjustment graph. So I should clarify, that didn’t come from the CDC, it’s actually a group out of Santiago, Chile. They didn’t age adjust to standard US data (what I did with my calculations), they age adjusted to World Health Organization international age standardization data so they could compare across countries. They through in individual states just to show how it would work in the US, but that really should be read as “if this state were it’s own country”. Every state looks “better” because all US states have older populations than the world average, if that makes sense.

     Second, BMI is an interesting factor, and one that could probably use its own post. At least in the US it’s highly correlated with age so it’s not quite fair to treat them both as independent variables. I did a quick scatter of it solo and it appears the R2 between obese BMI and excess mortality is about .1 (for some reason it won’t let me post it in the comments). The problem is there are a few states with very low obesity rates and very high or higher than average excess mortality (DC, NY, Montana) and some with high obesity that have lower than average excess mortality (WV, Alaska, Iowa). I may update the post with the graph later with that and the age scatterplot.

     This is not to say these things aren’t important, as you note. I do think taking individual state characteristics in to account before we talk about the success/failure of various interventions is really important. For example, the initial OPENsafely data from the UK noted that black and Hispanic populations seem to have worse outcomes even after controlling for other factors:
     https://www.nature.com/articles/s41586-020-2521-4/tables/4

     My personal bias is that there is likely a genetic component to who gets severe COVID, in which case some states may simply have more people of particular ancestries. I may be proven wrong, but I like to keep it in mind as a possible confounder. To note: it’s not as simple as white/non-white, as the R2 there is .07. But I’m not ruling out there’s a possible predisposition towards bad COVID that may be out of anyone’s control.

     LikeLike

3. Yeah, great stuff. Just to throw in a few things I know, Alaska has a large percentage of Natives who are obese and also many with severe alcohol issues, but they have a low covid rate, likely due to isolation and underreporting. Native Hawaiians show something similar. What is happening? Dunno. A great example of why state-level data is important. Significant local factors can mask other trends.
   Any obesity or other metabolic syndrome understanding is going to veer instantly into blaming and ethnic data. I can separate that out in the discussion personally, but Kyle and Bethany might both hesitate to put forth contentious things they can be later kicked for, however undeservedly. I’m retired and can say “Screw it.” I will understand if you have to be exquisitely careful. It’s probably good practice for you anyway. As the hostess has been sharing interesting info with me in another forum on the brain chemical influences on obesity, which kick into another gear around 1980 , it starts to get complicated. Such as, If your brain is telling you “You’re not fat and you’re still hungry” because something has screwed with your leptin levels, (Example: WV’s high obesity rate may owe something to exposures resulting from strip mining, and thus unintentional blaming for increased covid vulnerability in that population.) Or increased obesity the farther down you go down in the watershed…why, exactly? Messy. Eludes simple answers, at a time when even a few simple answers that explained 10% of what is happening would be most valuable.
   I have also noted that American states and Canadian provinces that border each other are similar in many covid rates (though the Canadian rates are lower) – except for an intriguing few. I am not saying this to offer a new theory, but only to caution both of you not to leap to an hypothesis without noting that possible disconfirmation.
   Yes, i made things more messy instead of moving the discussion to an answer. I wish it were otherwise, but that’s about all I’m good at.

   LikeLike

Comments are closed.