Top news, reports and insights for today:
- Daily headline summaries for Wednesday:
- Florida scientist who created a widely praised dash-board for monitoring COVID-19 says she’s been fired for refusing to “manually change data to drum up support for the plan to reopen” (NPR)
- One third of California’s COVID-19 deaths have occurred in nursing homes. Like most states, California has been slow to require comprehensive testing in long-term care facilities even as cases and deaths continue to mount (San Francisco Chronicle).
- First it was toilet paper, sanitizing wipes, and masks that disappeared from stores. Months later, it’s bikes, exercise equipment, freezers, hair clippers, thermometers, Nintendo’s Switch, trampolines and webcams that are in short supply (New York Magazine)
- Deaths rise AND deaths fall. How can it be both?
I received a question from an alert reader who asked me this:
Question: I was reading the daily briefing and came across this: “Last week, 9,484 Americans died of COVID-19, a rise of 13%, compared to 12,125 the week before (growth factor=0.78).” I’m unclear what “a rise of 13%” means? A rise compared to what, given that it’s a *fall* (of 22%) compared to the week before? Is that a typo, or what am I not understanding?
Here’s my answer: Thanks for a good question. I lose track of what can be confusing. There are 2 different things at play here. First, there is the cumulative growth in deaths. Last week, there were 9,484 deaths, which increased the cumulative death toll by 13%. The cumulative death total cannot decrease because death is what demographers call an absorbing state. Second, there is the rate of new deaths in consecutive weeks. That can rise and fall. We use the term growth factor to express the ratio of incident events in two time periods. So, both things are true and they represent 2 ways of tracking the epidemic. Total deaths rose 13% and the rate of deaths slowed by 22% compared to the week before. I wasn’t clear.
- Original Analysis: Total cases are closely related to population density at the county level
I recently read an article by Richard Florida, cofounder of CITYLAB, talking about the relationship between population density and the prevalence of COVID-19. This topic has popped on and off the radar. Infection rates across countries reveal a complicated pattern. For example, Japan and India have very high population densities, but have not had the magnitude of infections that many less densely populated nations have. I have stressed however that epidemics don’t occur on national scales, even if that’s how we tally the numbers. So, I decided to look at the relationship between population density and cumulative COVID-19 cases across all counties in the U.S. The graph I created is below.
Across all counties, higher population density is clearly related to more cases. Population density accounts for about a third of the variability in number of cases. There’s nothing surprising about this. More people = more cases. Why would it be any other way? It’s tautological, since the only way to get more cases is to have more people. I could show you rates (cases per 100,000 population) and the story would be similar. But there are a few other reasons this is interesting. From the standpoint of infection dynamics, we know that the virus is leaping from person to person in chains of transmission that are pretty localized. We believe there may be super spreaders, small numbers of highly influential infections that can trigger large clusters. We know that meat packing plants and nursing homes and other hot spot sources are distributed independently of population density. These are all factors that should weaken the association between density and cases. Yet, it is surprising how tight this distribution actually is. Further more, it is not surprising that the most densely populated areas have much larger cases. We expect a clear and strong association between density and cases in the upper right quadrant of this graph. What is striking is now strong and consistent the association is across the full spectrum of population densities (although this is partially a function of displaying the data on the log scale). In isolated areas, sparse rural communities, smaller towns, and beyond, cases go up as density does. This is a further indication that basic and universal transmission dynamics are at play across community sizes. That is what we expect to see in an outbreak of a novel pathogen when the population has no immunity. That’s also why social distancing remains our best tool to slow the spread of infections.