COVID-19 Age Adjusted Infection Mortality Rate: Understanding the Risks

I am normally not much of a LinkedIn poster but I posted this on FB and a couple friends encouraged me to post it here as the media does a poor job of explaining this correctly. I feel it is important that we all understand what the true rates are so we can incorporate this into our thinking accordingly keeping in mind that the low risk still need to protect the high risk from being infected and we all need to follow the current social distancing guidelines. These numbers won’t be perfect given the data limitations that still exist (I try to point these out) but I am confident they are directionally correct.

The new serology data from New York State is actually the impetus for this post as it combined with the COVID-19 mortality detail provided by NYC actually gives us a way of calculating age adjusted COVID-19 infection mortality rate with some robustness. The serology data (see first pic) shows that as of 4/27, 25% of NYC residents have already been exposed to Covid-19 (or at least 25% of those 18+ that were well enough to go to a grocery store where most of the “random” sampling occurred). To be sure there are issues with antibody testing when it is used to detect something that is rarely present in a population because too high a percentage of the positives will be false positives (i.e., if a test has a 2% false positive rate and the disease it detects is only present in 2% of the population then half of the positives will be false positives). However, because NYC has such a high rate of infection, this issue is minimized. I will say that New York State has not been entirely forthcoming about their calculation method (yet) so it is possible there are issues that make the numbers unreliable but given how high the positive rate in NYC is, I think the math is going to be solid unless the method flaws are egregious.

We also know from this antibody study the percentage of New York residents testing positive for COVID-19 antibodies did not vary a lot by age group in the 4/22 survey although there was more variability in the 4/27 survey (see second pic and note, this was a state wide, not NYC only number). To keep the math simple for now, I assumed the exposure across age groups was constant but in reality it may be that the age-adjusted 65+ groups risk should be adjusted upward by as much as a relative 30%-40%. (with other groups risk being adjusted down a bit).

NYC also provides on its website the number of COVID-19 deaths by age https://www1.nyc.gov/assets/doh/downloads/pdf/imm/covid-19-daily-data-summary-deaths-04302020-1.pdf. This data (pulled for 4/29/20) combined with the population distribution by age, https://www.baruch.cuny.edu/nycdata/population-geography/age_distribution.htm makes for a fairly straightforward calculation of infection mortality rate by age (not the same as the case mortality rate which is what the media typically reports; that only counts people testing positive for viral RNA in the denominator which is a huge underestimation).

Age 0-17: There are just 6 deaths in this group. Although people under 18 were not included in the antibody survey, I feel it is reasonable to assume they are being exposed at the same rate as adults. Given that, the antibody testing suggests 450,000 of the nearly 1.8MM people in this age group have been infected for a mortality rate of 1 in 74,000 (0.0013%). And yes, we could argue that this means we could open schools but there are the teachers and parents to consider in that equation. We all know how much young kids spread germs among themselves and then infect adults…I would swear it is a core competency

Age 18-44: There are 401 deaths in the 18-44 age group and the antibody testing suggests 873,000 of the 3.5MM people in this group have been infected which makes a nearly 1 in 2200 death rate (0.046%).

Age 45-64: There are 2,363 deaths in the 45-64 age group and the antibody testing suggests 528,000 of the 2.1MM people in this group have been infected for a mortality rate of 0.45%.

Age 65-74: There are 2,255 deaths in the 65-74 age group and the antibody testing suggests 172,000 of the 690,000 people in this group have been infected for a mortality rate of 1.3% (perhaps closer to 1.8% if the infection prevalence is truly lower in this group than younger age groups)

Age 75+: There are 4,140 deaths in the 75+ age group and the antibody testing suggests 138,000 of the 552,000 have been infected for a mortality rate of 3% (perhaps closer to 4% if the infection prevalence in this group is lower than younger age groups).

To put these rates into perspective the lifetime odds of dying in a car accident according to the National Safety Council are nearly 1% (0.0094%) which translates to about 0.012% in any given year. https://injuryfacts.nsc.org/all-injuries/preventable-death-overview/odds-of-dying/

This does not consider the impact of pre-existing conditions or nursing home status. According to NYC, >99% of the deaths where the patient medical history was known actually had a pre-existing condition. Unfortunately, it is hard to make heads or tails of this information because the list of conditions is so darn long and not precisely defined. The list of conditions counted include, diabetes, lung disease (does this include otherwise healthy smokers?), cancer (current or past?), immunodeficiency (e.g., HIV), heart disease, hypertension, asthma, kidney disease, GI disease (does that include common problems like IBS?), liver disease, and obesity. With a list that broad, 90% of New Yorkers might fall into it. Hopefully, something a little more useful is made available so we can further calibrate relative risk (a subject for a future post).

This analysis does not consider mortality reductions created by new drugs like remdesivir or other advances in our disease understanding that just lead to better patient management. It also of course does not consider the long-term damage to the lungs that may happen to those who get seriously ill but do recover. That is still not widely understood although clearly it is happening at some level.

And if you think NYC is an outlier, NJ also publishes death by age group on https://covid19.nj.gov/#live-updates and it’s distribution follows the NYC distribution. Unfortunately NJ has not done systematic antibody testing yet so I can’t be sure about the denominator or I would do these same calculations for it.

And lastly, I realize some of you will see this data as an argument to open things up and some of you will see it as an argument to be more cautious. I will keep my opinion to myself (for now) but hopefully, this information will help create a more informed discussion at your dinner table….

Jeff Boschwitz has a Ph.D. in Microbiology and Immunology from Cornell University and has spent 25 years working in and consulting for the diagnostics and pharmaceutical industries with a focus on growth and operational strategy and execution. He is currently working as an independent consultant.