There has been much discussion over the COVID-19 attributed deaths being reported out of Florida vs. what the real number is. I decided to dig into the numbers on my own with two truths.
- I’m not a good blogger/writer so for that I apologize in advance
- I’m pretty good with numbers, especially CDC data. I tracked Morbidity and Mortality Weekly Report (MMWR) data for about 12 years. Long story, prior life.
Methodology
First I needed to obtain the data. So going to data.cdc.gov/… I exported the dataset for the MMWR 2019 and 2020 (cool site, pull your own state data!). There is merit to obtaining more years for a deeper comparison; however, you open the data up to more reasons it will not be comparable. For example, 2018 was a stronger flu year in Florida so you would have to be able to normalize that vs. 2019 and 2020. To keep the exercise as simple as possible I limited the data to 2019 vs 2020. This requires the assumption that factors would be consistent (flu season, other deaths etc.) . To my knowledge, the flu seasons for the two years were reasonably consistent (i.e., no known significant strain misses in the vaccine). Additionally, while there has been some evidence that deaths from workplace and vehicle accidents were down during the lockdown period let’s assume that’s not a factor and the numbers are comparable. Note there is a lag in the data collection here. I pulled through Week #21 (week ended 5/23/2020). Week 22 and 23 was posted but appeared incomplete. I will update it over time but I truly believe that is partial as of today due to the known lag so it is currently not included.
With the data I simply ran a pivot table to show the MMWR week # for each year. This allowed me to do a simple comparison as to what happened with deaths YoY. I called this ‘Flat’ as I also ran numbers on a population adjusted basis. I used 1.1% population growth for Florida. This can be modified but you get the point. This quantified the excess deaths in 2020 as compared to 2019.
The data included a field where they counted death data where COVID-19 was the underlying cause of death (COVID-19 (U071, Underlying Cause of Death). Since the week #s are difficult to align to the reported figures by the State I made the assumption that this approximates the deaths being reported by the State. I did compare this to the data I’m extracting daily from the Florida Dashboard and it appears reasonable. You will also note there are no counts attributed to COVID-19 the for week #11 and earlier (week ended 3/14/2020). This actually ended up being the most curious piece of the analysis.
I then took the excess deaths YoY and compared that to the COVID-19 attributed deaths. Any delta represents more deaths in 2020 as compared to 2019 that cannot/are not explained. Aside from any other drivers this is the pool of potential undercount.
The Numbers
Note that I hid MMWR weeks 22 and on which is why you see a larger total death number for 2019. That is the full year total vs. partial year 2020.
Worksheet here: www.dropbox.com/…
But What Does it All Mean?
My comments are as follows:
- It looks like it was here WAY before we started counting. Weeks 1-11 had an increase in death count of 1,882. That’s 4%. If not COVID-19 than what is causing that spike? Could be flu based on seasonality but certainly merits a deeper dive.
- Through week #21 we have an increase of 4,446 deaths (5%) but have attributed only 2,105 to COVID-19. Even if you adjust for population growth you have 1,372 unexplained excess deaths.
- If I narrow the scope to when the State actually started reporting COVID-19 deaths (weeks 12-21) you are still left with 459 excess deaths. Population adjusted this appears more reasonable but again, looking at this data I think COVID-19 was killing people here way before we reported it.
My Final Thoughts
I think the number is higher, my math is showing 60-110% for the year vs. what we’ve reported. That may be high and it’s predicated on weeks 1-11 in large part. Knowing what we know now about the virus potentially being here earlier the number certainly looks much higher. I just don’t know how you prove it at this stage. We know they weren’t testing dead bodies, especially earlier when we had no tests. But I have to believe the 950+ excess deaths in the 4 weeks BEFORE we started reporting are largely COVID-19 caused. I’m sure years later when the death data is final (no longer provisional) and posted by the CDC the excess deaths across the US will be much higher than the death count we are working with (NOTE: the final CDC death data is only up for 2018, 2019 is still open to give you a feel for the data lag).