The most important question we face around the COVID-19 outbreak is just how deadly is it.  We know the virus is highly contagious, we presume that there is little built in immunity since the virus is “novel,” or new to our population, but there is considerable spread as to the estimates of the actual death rate.  Some will argue this is like the flu and around 0.1%.  Other estimates are around 1% and I’ve seen higher numbers than that.  Even if this really is like the flu and has a 1 in 1000 death rate, a novel virus that spreads quickly and infects most of the population would kill 300,000+ in the USA.

Death rates are calculated as (number of deaths attributed to COVID-19) divided by (total cases of COVID-19).  While there are some arguments to be had with regards to how deaths are attributed to COVID-19, particularly outside the United States, I think we have a fairly good handle on this number.  If you are sick enough to hit a hospital, you’ll get tested and we’ll know if you had the virus.

But the denominator in the fraction is really hard to determine, particularly for a virus that is known to be asymptomatic with many cases.  The vast majority of these people will never get tested.

As of 4/22, the raw death rate in Ohio is 4.3% (610/14,117).  Here’s a sampling of a few countries raw death rates for comparison:

Country Raw Death Rate
United States 5.6% (47,684/849,092)
Spain 10.4% (22,157/213,024)
Italy 13.4% (25,085/187,327)
Germany 3.5% (5,319/150,773)
United Kingdom 13.6% (18,100/133,495)
Sweden 12.1% (2,021/16,755)

All data courtesy of

Some of those numbers are truly scary, but they are artificial due to a global testing bottleneck and the prevalence of asymptomatic cases.

The good news is that we have some better data coming online that helps us understand the actual population of infected patients which will help us understand the real death rate.  I posted previously about the Stanford study in Santa Clara, CA and we now have a similar study in LA County that again showed through antibody testing a potentially much greater infected population than originally suspected.  These are tremendously important, but they are only testing a small sample of a population and extrapolating from there.  That’s where our current prison outbreak in Ohio becomes so important.

The best way to truly understand death rate is to just let the virus spread and see what happens, but ethically, you wouldn’t want to allow that to occur if you could prevent it.  Clearly our leaders in Ohio agreed as they began testing and isolation within all of our prisons once they had evidence of an outbreak.  Unfortunately, for two prisons, the Marion Correctional Institution and the Pickaway Correctional Institution, that effort came too late and upwards of 75% of inmates are infected in both institutions.  As sad as these outbreaks might be, they represent amazing natural experiments.  We now have 3500+ cases in two small counties where we can track the outcomes.  Given that the virus runs its course in about 2 weeks and the testing began on 4/11 and was complete by 4/18, we’ll have a good estimate of a death rate by 5/2.  I’ll be watching this closely.

Otherwise, now that testing of the inmates is basically complete, we’re back to the status quo in Ohio of slow decline in cases.  4/22 brought us 359 new lab confirmed cases.  When we look at onset date and pull out the inmate tests, we have a clear downward trend (yellow line) over two-weeks with a slope of -7.7.