Comparing the Coronavirus Situation Across States

It’s important to preface posts like this with the disclaimer that I am far from an expert on the coronavirus. I’m not an epidemiologist. I’m not a public health professional. I’m not a statistician. If you’re an expert and you want to correct anything in this post, please do so. If you’re not an expert and want to discuss anything in this post, please start that discussion. This is an attempt to learn more than it is an attempt to educate.

Recently, I’ve been trying to understand how the severity of the coronavirus problem relates between different places in the country. It’s impossible to get a perfect read on this. States have different testing programs. States report data differently. There’s diversity within states: I live in Austin, where the population is denser and travel in and out of the area is more common than it is for much of Texas; my parents live on the outskirts of the Chicago suburbs, where density isn’t as high as within Chicago itself, but is higher than it is downstate.

In an effort to understand the differences between areas better, I looked at two variables this morning. I looked at both variables by state because, well, it’s easy to look at things by state. As has been said, situations differ within a number of states (again, it’s fair to assume the situation is worse in Austin than in other areas of Texas). But this at least provides a reflection of the average situation across each state, which is helpful for my understanding, since I can make some inferences about Austin from the broader Texan state of things.

The first variable is how quickly people are dying on a per capita basis. I took this from The COVID Tracking Project and measured it as the difference between total coronavirus deaths to date, reported on July 4^th, and total coronavirus deaths to date, reported on July 11^th. I have some concerns about using these specific dates, because the holiday weekend may have delayed some reporting and slightly increased death counts for this past week, but I wanted current data, and I wanted seven days of it to iron out the likely-larger effects of variance by weekday. My guess is that the death rates are comparably affected across states by the Fourth of July holiday, so, because this is a comparison between states rather than an overall assessment of the severity of the situation, the Fourth of July backlog concern seems negligibly important.

The second variable is the transmission rate of the virus: R_t. Rt.live is my go-to for this, as it seems to be the best estimate easily available of the current rate of transmission in every state, specifically defined as how many people are expected to be infected by each infected person (in a state with a transmission rate of 1.10, each person infected would be expected to infect 1.10 people; in that same state, each 100 people infected would be expected to infect a collective 110 people). Of course, this will also likely be delayed slightly, but as with death rate, it’s the best estimate available.

On the chart below, I’ve described these variables roughly as ‘how bad the situation is’ and ‘how better/worse the situation is getting.’ States with transmission rates below 1.0 are seeing their overall situations improve. States with transmission rates above 1.0 are seeing case levels rise. This—the transmission rate—is the front end of the timeline. At the back end of the timeline we have the death rate, which can offer a reflection of how bad the situation currently is. Again, in states with transmission rates above 1.0, this will likely rise in the weeks to come, whereas in the few states with transmission rates below 1.0, the number should soon drop.

The immediate takeaway is this: Things are comparably good in Connecticut and Maine. They’re decent in D.C., Rhode Island, New Hampshire, Arkansas, and North Carolina, and likely improving. Elsewhere, they’re getting worse. Arizona and Mississippi are in a bad place, with Mississippi likely worsening more quickly than Arizona. Wyoming, Montana, Hawaii, and Idaho, all states with small populations and therefore presumably a more inconsistent death rate, could see problems worsen soon. Florida, California, and Texas—three states prominently in the news for rising caseloads—are indeed on the leading edge of bad situations that are getting worse. New Jersey has a bad situation that’s not exactly stable, but is fairly close to stable. Louisianans should be very concerned. I won’t go through all the states (seventeen, plus D.C., is enough), but that’s how to interpret this graph. You want to be where Maine is, or at least where North Carolina is. Keeping that transmission rate below one is what “flattening the curve” means, while opinions differ over where on the x-axis (deaths per day over the last week, adjusted for population) the curve should be flattened. Lastly, it’s worth emphasizing again that these variables are weeks apart, so it may take a while for deaths to really increase or decrease in accordance with transmission rate.

Here’s the data in table form, sorted from where the situation is the worst to where it seems to be the best. If you’d like to compare transmission rates directly between states, and see how they’ve changed over time, rt.live is the place to do that. We may revisit these numbers soon if this seems to be an effective way of gauging situations.