The disease burden plots. Cases per fatality (recent) vs cases per capita per week. The curved lines are contour lines for expected fatalities per capita per week and are spaced at 2X changes. 1/
These plots are an attempt to use simple statistics to evaluate disease burden at different places, at different times. The cases per fatality is fatalities in the past two weeks and cases 20 days earlier. If you divide cases per capita by cases per fatality, you get 2/
fatalities per capita, or disease burden. The curved lines show functions where the expected fatalities per capita stays the same. They are spaced at 2X changes. 3/
For example, Maine (ME) has about one fourth (2 lines) the disease burden of Nebraska (NE) right now, because they are separated by two lines. VT to ME is an 8-fold change. 4/
The darkest line lower right is the disease burden in New York City in April at its peak week. The line two to its left (4X lower) is Arizona in summer. Hopefully this shows you how bad things are in your geographic region. 5/
If not, let me know, and I will add it (except for those countries that are super-low in transmission - they compress the rest of the plot). 6/6
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Rt vs percent infected in US states and Canadian provinces. How I calculate these (thread). 1/
Rt is the ratio change in new cases over one serial interval (reproductive cycle). I use 5.2 days for SARS-COV-2. To calculate it, I take the last 10 days of new cases, Hamming window it, and take its ratio to the same measure 14 days earlier. 2/
Then, raise that ratio to the (5.2/14) power to convert a 14 day ratio to a 5.2 day ratio. 3/
US daily 7d average cases and deaths. Rt in recent weeks:
1.23
1.21
1.16
1.14
1.11
Most recent Rt is decreasing, which means we are approaching a peak sometime in 1-3 weeks. The 7d average fatalities should be a new record in 21 days. 1/
I've been making these plots since June, but recently noticed (and commented) on the stability in the CFR since mid/late July. The program spit the numbers at me daily - I made a plot of it this week. 2/
If CFR is stable, then this plot, which used to show percent infected vs fatalities, will look the same as cases per day vs fatalities. So I replotted it as cases per day. And, note, I adjust the cases per day prior to July 20 based on changes in CFR before that. 3/
For the ratio of cases to total infections, the best approach would be surveillance data, which we do not have. We do have reasonable estimates of infection fatality ratio, and it can be compared to the measured case fatality ratio to get that ascertainment ratio. 1/
In the US, recent CFR is 1.67%. So, if you believe the CDC and IFR is 0.72%, there are 2.31 cases per infection. If you think we are doing slightly better than that (I use 0.65%), 2.53 cases per infection. Maybe a lot better (0.5%), and you get 3.34 cases per infection. 2/
The way that CFR has stayed stable since late July, and cases and deaths have proportionally tracked, tells me we have not gotten much better at saving COVID-19 lives in that time period. And, a 2.53 ascertainment matches the Arizona summer wave serology data quite well. 3/3
Others will follow in coming weeks not months. There will still be lots of cases and deaths, but that exponential growth will not continue too long.2/
Once a spot gets worse than AZ was this past summer, they respond, and do not let things grow to become as bad as NYC in April. Before we peak we may hit a new deaths per day record in December. 3/
Is Trump the Infecter in Chief? Here is a plausible scenario based on public data (link at end). Sept 24th, Ronna McDaniel, who does not know she is positive, travels with Trump to Charlotte and infects him. On the 26th she finds out she is a contact and isolates. 1/
Two days later he hosts two events at the White House, the Gold Star military family event, where he infects an Admiral, and the Barrett nomination where he infects about 10 including three senators, a sitting governor, and a university president. 2/
Four more at debate prep the next day which was reported in a small, poorly ventilated, room. 3/
The CO2 report from school one. Most of the rooms in school were reading ~800 with students in them, maybe some up to 900. 1/
An outside trailer used as a teacher office was at 3000 ppm CO2. It dropped to 600 within 5 minutes of windows/doors being left open. Classrooms were also really responsive to opening windows on opposite sides. 2/
Even events in the gym did not really get CO2 higher, but windows and doors open could drop CO2 in any room relatively quickly. Windows and doors in common use buildings, people. Windows and doors. 3/