, 21 tweets, 10 min read
This thread on #COVID19 by an engineer is doing serious numbers. It's projecting millions of #CoronavirusUSA cases, but I'm very concerned it's a naive & uninformed take that will cause more harm than good. Here's why these kind of projections are likely poorly advised (1/n)
This thread, like similar armchair projections, are based on running with an assumption far beyond the domain where it is valid, typically unexamined application of exponential growth. Exponential growth occurs when something swells by a power of itself each step. Let's see:(2/n)
Exponential growth is found in many biological systems. Bacteria growth, for example, is expotential - the number of cells double with time, like that below. And MANY systems are like this; early stage cancer growth (my area of interest), and indeed epidemics too (3/n)
There are serious caveats to this. For example, early stage cancer growth is similar to the bacteria above - let's see what happens if you assume it goes on without limit. Here's some slides from a lecture I give on maths bio showing you'd quickly get pretty huge tumours..(4/n)
We actually don't get tumours bigger than countries, for good reason - there are other limiting factors to how biological systems can grow, and limitations to how they expand. In essence, assumption of unrestrained exponential growth too simple - what about for disease? (5/n)
In disease modelling, base reproduction rate R0 is often used to estimate how many secondary cases arise from a single case (typically at early stages of epidemic) - I wrote a little about it for Irish Times recently in context of #COVID19 (6/n) irishtimes.com/opinion/corona…
And like lots of biological system, growth is initially exponential in early stages. But take measles with R0=~15, & naively apply this for just 5 doubling generations...and you'd conclude that billions of times more people than live on earth would be infected🧐- (7/n)
This is precisely the issue - you can't just presume constant exponential growth, or extrapolate blindly from an R0. For example, as @GidMK points out, doing this would cause you to over-estimate cases in Wuhan by factor of 20x thus far (8/n)
Epidemiology is hard, and nuanced. All mathematical biology is. I'm just a mere cancer modeller and not qualified to wax lyrical but critical point is that models are only as good as the assumptions underpinning them. We also need to make a distinction here too.. (9/n)
..between mechanistic models and phenomenological models. This might seem academic, but the latter are "best-fit" to what we have to hand, whereas the former are rooted in biologically validated parameters. Image is from an old lecture I gave on subject, for those curious. (10/n)
Both are useful, but one has to be VERY careful over-extrapolating from empirical or phenomenological models in particular. You have to know WHY a models applies, and where it doesn't. And epidemics aren't simple to model. Nice primer here (11/n) sciencedirect.com/science/articl…
Anyway, I'm just a guy who models cancer stuff, not a mathematical epidemiologist. But it really annoys me when I see doctors and scientists who should know better spreading irresponsible tweets with numbers that are at best naive and at worst ignorant. It doesn't help (12/n)
My other area of expertise is, by coincidence, public understanding of science and how disinformation spreads. And as the @WHO make clear, we have an #covid19 "infodemic" problem where dubious information perpetuates widely, as I wrote for @IrishTimes 13/n irishtimes.com/opinion/corona…
..scientists & physicians are trusted in society. We have a responsibility to not panic people. I'm sure people don't mean harm, but semi-informed takes just muddy understanding. Please, can we just let public health bodies do their job? (14/n)
To conclude, #COVID19 is stressful enough already without sensationalised takes. Panic helps no one, and misinformation is toxic. A little less comment, more reflection might be the ideal - #coronavirus is scary, but there's good news too. Rant over theguardian.com/world/2020/mar… 15/15
PS: didn't even mention the benefits / limitations of SIR type models, but for lots of reasons assumption of well-mixed population breaks down quickly, & spatial factors start to matter. So too do counter measures! @Kit_Yates_Maths great on this.

mathworld.wolfram.com/SIRModel.html
Postscript: I keep getting same questions, so here's the responses.

(1) No, I'm not advocating being complacent - Coronavirus situation is extremely dangerous & could overwhelm many healthcare systems worldwide. But I don't think over-extrapolation of numbers helps anyone.
(2) China's cumulative rate is 56.1 cases per million people. Scaled to US population that'd be 18355 cases. Even with SK's poorer ratio, you'd expect about 47k cases in same interim. China's rate is slowing, which is good. Infection control very effective
(3) Finally, won't be discussing this further because I don't think either myself or original poster sufficiently expert to hold court. Despite best intentions, it's too easy to confound public understanding, & I have no intention of contributing to infodemic over this.
To conclude, please follow advice from public health bodies, wash your hands, isolate if you feel sick, practice good coughing etiquette, & be sceptical of claims you encounter online. Controlling rate of infection hugely important to keeping HC systems up
And finally, no - I don't have a soundcloud or anything. But if you can, donate to @WHO or a local medical charity! If you want to see more of my rants, I write often for @IrishTimes & @guardian, or you can read my book #TheIrrationalApe if you like amazon.co.uk/Irrational-Ape…
Missing some Tweet in this thread? You can try to force a refresh.

Keep Current with Dr. David Robert Grimes

Stay in touch and get notified when new unrolls are available from this author!

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

# Try unrolling a thread yourself!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" `@threadreaderapp unroll`

You can practice here first or read more on our help page!