Long #COVID19 thread:

2020-07-18 Florida #COVID19 update:

I've updated the model since last time with several new graphs and predictions for the key metrics being tracked (the ones being shown on the "dashboard"): total cases, new cases/day, etc.

As I mentioned the other night, the gross two-week predictions (2WP) are tracking closely, but under, the reported number of cases. This particular prediction I am referring to is the red curve as shown that intersects Florida's population in early October, ... 2/

...which as I've said, has not moved much. I'll reiterate: As fantastical or hyperbolic as that sounds, *if things do not change behaviorally by people in the state of Florida*, we will see this infection rate then. This is not "Ask The Magic 8-Ball" - this is simple algebra. 3/

The R0 or r-values have come down slightly, trending towards 1.0. Good job, folks. This will help flatten the curve. Let's try to sustain that. Overall predictions still hold steady: This is a good indication that the basic models are accurate enough to be ballpark figures. 4/

They are not changing wildly, so every day I grow more confident in these forecasts.

Some bad news: By the end of the year, we are looking at anywhere from 5M-12M cases, and 55k-69k new cases *per day*, and a death toll of approximately 12k-15k. 5/

A few themes and questions that need addressing for the general public:

1. "The numbers are faked/the government is lying to us/you can't trust the media": This is one of the most tiresome complaints I see and hear, and frankly, it's lazy. 6/

You as well as I can open up a browser on your phone and go to any of the websites and links that I've posted--and I've posted a bunch--and the data are not manipulated. Granted, we're in the middle of a global viral pandemic and the paperwork part of things is, shall we say, 7/

expected to lag (due to the volume of cases and medical staff and first responders trying to save lives), so there is always lag. Even the daily data that I and everyone else receives is late, but it's that day's data and it's consistent. 8/

There is no excuse at this point for not being informed. There is no "news media conspiracy" to make up or alter numbers, so stop that, please. Yes, there are occasional glitches, such as when the FDOH site -- which I monitor daily -- temporarily said there were over ... 9/

... 652,000 cases (since corrected) and only a 5.7% test positivity rate (later corrected to ~11%). 10/

As far as the government lying, yes, Florida state and Federal officials are lying, but they're lying about broad themes, not numbers, per se. Watch Trump, Pence, or DeSantis speak, if you can tolerate it. 11/

2. "It's hardly fatal at all, so why is everyone making a big deal about this??? I want to go back to NORMAL!!!" (stomps feet): Yes, it's true that COVID-19 has a case fatality rate in Florida of somewhere around 1.5% -- so you may think you have nothing to worry about. 12/

However, there are numerous and mounting reports both anecdotal and medical that suggest that those infected suffer a horrible time of it, for a prolonged period of time for weeks or even months (since we now have observed data with which to make these assessments), and ... 13/

... there are long-term deleterious effects that aren't even known yet because it is a *novel* coronavirus. There is no deterministic factor that says, "Oh, you're safe, you'll only get the sniffles", so it is absolutely not worth anyone risking this. 14/

3. "I don't know anyone with it, so I think this is a hoax.": Setting aside for the moment how colossally dumb a statement/belief this is, it is understandable that people may not know anyone who has it, has had it, or has died from it - only 1.6% of the Florida population... 15/

... is infected, which means 98.4% don't* have it (*we don't know if we're asymptomatic carriers or if we're totally in the clear). But we've also only tested not even 14% of the population. 16/

This is quite equivalent to the analogy of assuming you know what's in a completely pitch-black room by shining a penlight on one spot on one wall, when what you need are floodlights -- you don't know what you don't know yet. 17/

Anyway, I think you get the picture. There are gobs of factual data being put out by a multitude of sources, no one is manipulating them for nefarious purposes, and it's everyone's responsibility to help get this under control. 18/

This is why many other countries are already able to get back to some kind of normal and Florida is mired in 5th place *in the world* as an epicenter of this pandemic. 19/

Upcoming enhancements to the model include tracking of testing and hospitalization trends. Next model update will likely be next weekend since this takes a lot of time to update and pull together. Thanks for reading. I hope you find it useful or at least a mild curiosity. 20/

Your weekly dashboard figures: One, I corrected an error in how I was computing weighted averages, and two, I now have projections for an end of year (EOY) death toll. (Numbers at the bottom of the dashboard are current numbers as of today.) 21/

A composite plot of various R-values. 22/

This is a single series of R-values; as you can see, it peaked around the Fourth of July and has since gotten back to where we were when I started tracking a month ago. However, this also sadly says that we've made no progress and haven't gotten closer to the goal of r=1.0. 23/

My earliest linear models are now clearly wildly off the mark as far as what the year-end predictions for total cases will be. The actuals are the purplish traces on the left, climbing steeply, slope unchanged over the last month. 24/

Here's the linear models again, with the FORECAST model: We're looking at 1.8M (1,800,000) cases by the end of the year just from this model alone (the peak of the green mountain). 25/

In a month, the percentage of Florida's population that has gotten infected has increased by over 1.0%. One percent of 21.48 million is almost 215,000 people. 26/

Here's a chart I keep referring to: The FEM1 and FEM2 curves still predict total infection months later, but the gross two-week prediction (2WP) is holding steady. I've formatted the lines with dashes to indicate they are projections, not actuals. 27/

The purple line at the bottom, almost lost in the noise, are the actuals. It's hard to see, but over time, should separate out to differentiate from the three other prediction curves. 28/

A more dramatic rendering of the case growth over time. 29/

This plot shows the %age of growth in cases over time: daily, cumulative average since the start, and the seven-day rolling average (7DRA), the latter being the smoothest average of them all. Using a rolling average is a good predictor when used as a multiplier in any model. 30/

I developed these extra FEM models after I saw FEM2 was insufficient and unrealistic in some ways. The purple curve is FEM2D, a modified FEM algorithm which now shows that we will max out with over 180,000 cases *per day* by the end of November at the rate things are going. 31/

It tracks much more closely with the real number, which you can see as the noisy/jagged portion in the lower left (the actual numbers of cases per day). 32/

Anyone who tells you the number of cases per day is decreasing is simply wrong: Here's our trend for the past month. These are daily actuals. There is a 4- or 5-day cyclical behavior to them that I am curious to represent mathematically. 33/

Projections for new cases/day using FORECAST and various linear methods. It's not pretty. 34/

New chart: Here are the published values for cumulative/total hospitalizations. Choose either site: They only differ by approximately 300, the explanation for which I do not know. 35/

Thinking about taking a scenic trip to a Florida ICU bed? Book early!

Grim humor aside, this is an interesting math problem, which could be represented by a linear, homogeneous, first-order differential equation: 36/

The number of beds available at any time equals the total number of beds, minus those occupied, plus new beds being constructed (if any), plus those cleared by patients who recover or die. Right now, this is just a static plot tracking the actual reported available ICU beds. 37/

New graph: The death toll has risen from 3,100 to 5,000 in a month. 2,000 deaths a month. Is that acceptable to send people back to work and school? 38/

Chart showing the averages for deaths/day: daily, cumulative average (since the start of my modeling), and a 7DRA figure (the smoothest curve). Beware anyone who says the death *rate* is decreasing, because it is not: We are averaging nearly 60 deaths per week right now. 39/

FIN.

Thanks to @GeoRebekah and her team for her tireless and indispensable work, whose data I use extensively in my models. 40/40