COVID-19, Are we Plateau Yet?
Like everyone else I’m sheltering in place with too much time on my hands and way too much curiosity about the data behind the pandemic. In my first post on the subject 10 days ago I looked at case fatality rates (cfr), population mortality rates (pmr) and case doubling rates (cdr).
Ten days seems like a lifetime ago: at that point New York and Italy were in the exponential growth phase of the pandemic. Today the data kind of looks like a plateau. Plateau is a nice French word that derives from plat meaning flat. Mathematically the plateau in new cases also represents the inflection point in the total cases curve: after the plateau the number of new cases and deaths “should” decline gradually, in other words, it represents hope.
Having ridden a bicycle up many hills I am suspicious of plateaus. When I’m exhausted and I get to a flat part on a mountain I’ve never ridden I start to wonder if this could be the top. It rarely is. Frequently an even steeper section awaits around the corner and I’m usually already out of lower gears to shift into. It’s the hope that the summit is close that keeps me pedaling.
We’ve never seen a pandemic like COVID-19 before so this is a new hill for all of us. Are we near the peak or is there a tougher climb up ahead?
Changes in the Case Doubling Rate (CDR)
If we look at either the county or state level things are looking better (green), particularly near the coasts. On the other hand many more counties in the middle of the country now have enough cases to show up on the map and many of those are red (cases are doubling faster). If you’ve been reading in the news that rural counties are starting to be impacted more severely, this chart tells a little bit of that story.
At the state level things look even better. There’s lots of dark green except for South Dakota and Rhode Island. (Alaska and Hawaii are also green if you’re curious)
The notebook that generates the above charts can be found here.
Comparing States
Another notebook lets me compare any set of states to look at regional patterns in more detail. Since I live in California I’ll start with the western states. Here’s the top level summary:
California has by far the most cases and deaths of these four states but Washington has been the most impacted, with 65 deaths per million people. Surprisingly though, only 0.13%, or about 1 in 750,000 people are known to have been infected. Even if the known infection rate is much lower than the real infection rate that tells us, thankfully, that much of the population of these states has yet to be exposed.
The daily growth charts tell the plateau story more dramatically:
For those of you who may not be used to reading log plots in scientific notation, 10⁰ is 1, 10¹ is 10, 10² is 100 and so on.
The plateau is most evident in the top left chart which shows new cases per day. Since this data is very noisy it has been smoothed using a 7-day moving average. Deaths lag cases which is why the new deaths in the bottom left chart don’t show any obvious plateau yet except for Washington.
Now, the sobering part here is that most models tell us the peak death rate should come at about half the total deaths which means you can roughly double the deaths column to estimate the ultimate number of deaths in each state. Should this hold, California can expect to see another 700 or so deaths in the weeks ahead.
Let’s look at some other regions. New York has been the hardest hit state, not just in the US but in the entire world. Let’s compare it to its immediate neighbors:
The death toll in New York state is highest in the world at 483/million. New Jersey is relatively hard hit as well at 262. The overall known infection rate in New York is just below 1% which means it will surely exceed that in the next few weeks.
Fortunately new cases have reached a plateau in all these states excepting perhaps Connecticut. The plateau for deaths is again lagging new cases. But if New York is halfway home it means it can expect another 9000+ deaths which is horrible to contemplate.
The Midwest
Michigan has also been hard it, especially recently. Let’s compare it with its neighbors:
Fortunately the number of cases appears to be peaking in Michigan, Indiana and Ohio. Illinois is still growing.
The Gulf Coast
Louisiana has been in the news a lot recently. It’s mortality rate is a bit higher than Michigan’s but still far from NY/NJ.
Fortunately new cases appear to have peaked in Louisiana and are clearly on a downward trend. Alabama’s cases are still growing but they are about 10% of Louisiana.
Four Corners
As much of my family lives in Colorado I was very curious about the Rocky Mountain states. Wyoming remains the only state without any fatalities. The four corner states look like:
It’s fascinating how few deaths Utah has seen compared to its neighbors. I haven’t heard a good explanation as to why.
Three of these states have plateaued in terms of new cases with Utah showing a clear downward trend.
The Southeast
Florida has been in the news thanks to their Governor’s resistance to implementing social distancing measures. They are indeed doing worse than their immediate neighbors but nowhere near as bad as states in the Northeast.
The rate of new cases in Florida appears to have peaked albeit at over 1000 new cases/day. The same appears to be true in Georgia and South Carolina but North Carolina is still climbing.
Around Missouri
I’ve been watching Ozark and Tiger King lately and one of my best friends is sheltering in some remote corner of Missouri so here are four neighboring states in that general vicinity:
The casualty and infection rates here are fairly low.
Missouri appears to be near a peak, new cases in Oklahoma are on their way down, and the other two states are pretty flat. On the other hand it was just Easter and if the churches in this area were really packed then we could see a resurgence of cases in the next few weeks.
Other Areas
The aforementioned notebook includes a function that lets you generate the chart and summary table for any combination of states. I recommend not doing more than four states at a time because the lines get pretty jumbled otherwise.
It will be interesting to come back to this data and see if these plateaus were real or if there’s more climbing ahead. The nation is tired and anxious to put this pandemic behind us and start the descent.
I’d like to thank the New York Times and the COVID Tracking Project for making their data openly available. These resources have been invaluable.
