(THREAD) Suppose we want to figure out how many new #COVID19 cases there are in #Indiana each day on average, BUT… we don’t want to use *any* of the testing/positives data reported by ISDH. Can we do that and what would it look like? [1/X]
One way would be to use “infection hospitalization rate” or percent of *all* infections (not just detected ones) that result in hospitalization. How do we find this? We need to know total cases out there and number of hospitalizations from these cases at some point. [2/X]
Phase 1 of the Fairbanks study (random sample of all Hoosiers) gives us an estimate of *all* infections out there. Their conclusion was that by April 29, there had been 187,802 Indiana residents infected with #COVID19. Fairbanks results are here: https://www.cdc.gov/mmwr/volumes/69/wr/mm6929e1.htm [3/X]
Next, suppose hospitalizations usually occur one week (on average) after infection. Based on Regenstrief Institute hospitalization data, one week after April 29th (or on 5/6) a total of 7,423 Hoosiers had been hospitalized for #COVID19. [4/X]
Put these two together and we get an "infection hospitalization rate" of 7423/187802 = 3.95%. Or that 3.95% of all #COVID19 infections (ALL infections, not just those we detect) in #Indiana on average result in a hospitalization. [5/X]
Next, suppose the infection hospitalization rate stays mostly the same over time (a reasonable assumption unless the virus is mutating substantially). Then we can use this rate to see how many infections are implied by our hospitalizations at any given time. [6/X]
Calculating implied average daily #COVID19 cases from daily hospitalization data using this method gives us this graph of approximate average daily infections in #Indiana. [7/X]
While this is a *very* rough estimate, it suggests that during March we peaked at about 4,500 infections/day. Recently, daily infections have been rising, but we're only at about HALF the level of daily new cases compared to the first peak. [8/X]
This is a *very* different story from what the ISDH data on daily #COVID19 positive cases tells. Looking at daily positives data suggests that we are currently at about 35% *higher* daily positive cases now than compared to the first peak. [9/X]
So, which story fits better? Are #COVID19 cases spreading at a 35% HIGHER rate or HALF the rate they were during the first peak in March? There’s a *huge* difference between those two. [10/X]
Calculating implied infections from hospitalizations makes a lot of big assumptions *but* it has the advantage of not being affected by changes to how much we’re testing and who we’re testing (both of which are changing a lot right now). [11/X]
You could also use infection fatality rate and deaths to calculate implied infections in the same manner. If you do that, you end up with this graph, very similarly shaped to implied infections from hospitalizations. Again, way fewer infections now that March peak. [12/X]
I'm more partial to the implied infections from hospitalizations method though due to a larger sample size and I think "infection fatality rate" may be more likely to have changed than "infection hospitalization rate." [13/X]
Just to be clear, these are *very* rough approximations at best of total infections, but when our testing patterns are shifting and the relationship between positives, hospitalizations and deaths isn't making sense anymore they may be worth looking at. [14/X]
To see what I mean by "relationships... not making sense", look at this graph. Back in March/April all three rose together. Now (since June) positives are way up with hospitalizations only rising a little and deaths effectively flat or down. [15/X]
You can follow @MicahPollak.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled:

By continuing to use the site, you are consenting to the use of cookies as explained in our Cookie Policy to improve your experience.