The changing ages of COVID-19 patients.

We heard about COVID-19 back in January and February. It was an old person’s disease, a killer. Recently, people as widespread as Scott Gottlieb and Ron DeSantis and Natalie Dean have speculated about the potential implications of younger people getting the virus. We are also hearing about how hospitals now have younger people. The observations that the age demographics of confirmed cases have changed, and hospital admissions have changed, are true. The inference that young people are now catching COVID-19 when they were not in March and April is wrong. Here are some data and details.

The following is a screenshot from late June from Florida state data on the age demographics of confirmed cases.

This is their new, younger, age demographics. Were we to look at the same data in mid-May, it would show a much older demographic, with many fewer younger cases. I know that criteria to test have been changing. In the beginning, you had to be severely symptomatic to be tested. Today, testing is free and no-criteria in Florida, and has been since May 16th. People who are severely symptomatic are biased older. This forms a classic statistical sampling problem, when the criteria to sample (symptoms and age) are correlated with the output variable (age demographics). We can make a first check on this using death demographics. These are not subject to the sampling bias problem. If younger people are being infected, it should change the proportion of all deaths in each age group. Here is a snapshot of death demographics from the USA.

So, you can see there is no age shift in death demographics, and that the proportion of deaths is greatest for the 85+ age group, with comparably fewer deaths under age 50. This data is national, from the CDC, and replotted as proportions.

Florida did something else interesting. They did an early surveillance antibody study, with first results May 29th. They indicate the percent of people who tested positive for antibodies in each age group. Using this data, we can show the age demographics of all infections in Florida prior to May 29th (probably about 10 days earlier, actually). I took their numbers, and adjusted them for the Florida population in each age group, and made the following plot of all ages of infection in FL from mid May and earlier. You can see two things that are really surprising. First, these are even younger than the current confirmed cases. Thus, there is no shift in age demographics. Young people are not starting to get infected with COVID-19, they have always been getting infected. Second, the 5–14 age group has the largest number of infections despite having the fewest clinical complications. These are important bits of information for public health officials to have to help them form safe back-to-school plans.

To bring this back to the classic statistical problem on why younger people are not being infected now, here is a simple illustration that my PhD advisor used to inform us. Suppose I were to ask you to measure the height of everyone who walks through the door who is more than 6 feet tall, and you find the median is 6 ft 3 inches. Then, I ask you to change and measure everyone’s height, and you find the median is now 5'7". My question to you is: Did the height of your population suddenly change with everyone shrinking 8 inches in height?