1. A thread on why we want to #FlattenTheCurve.

Elsewhere, I've strongly endorsed the movement to flatten the curve, and with help from @k_thos have developed data graphics that illustrate the basic principle, including the one below.

2. Why did I think this was an important public health health message a couple of weeks ago when I started pushing it? And why do I think it is still a good message?

3. Before I answer this, I want to be clear about what I am NOT advocating. There is a common misperception that #FlattenTheCurve means letting the epidemic spread through the population until herd immunity ( https://en.wikipedia.org/wiki/Herd_immunity) is achieved.

4. This has never been my intent, and I feel strongly that we cannot allow this to happen.

The cost of allowing 20%-80% of the population to become infected is far too high to accept.

5. A study from top disease modelers out of Imperial college, WHO, and elsewhere was released today: https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf

This study, summarized in a thread from @kakape (below), uses mathematical models to look at the impact of that approach. https://twitter.com/kakape/status/1239694176066965505

6. In short? It's a disaster.

In the US, two million people die.

This is neither a controversial claim nor a surprise to anyone who has been thinking about it. @Mlipsitch explained in Feb that 40-70% of the population would be infected—and death rate estimates were around 1%.

7. Moreover, the Imperial college model (and I think just about any reasonable model) produces enough concurrent cases that the healthcare system is overwhelmed.

If you let the epidemic go until checked be herd immunity, you end up WAY over the dotted line on the diagram.

8. And despite your best efforts to minimize the rate of spread, the trajectory is likely to look much more like the tall curve than the shallow one.

9. If we are not going to let the pandemic sweep through the entire population, why talk about flattening the curve at all?

For me, the key is this:

10. When we ask people to #FlattenTheCurve for #COVID19 #coronavirus, what we are asking is for them to take aggressive, proactive steps to slow down the spread of the virus *before* the epidemic reaches a crisis point in our hospitals. Simple as that.

11. Here is why that matters. Below is a very approximate schematic for the time course of disease for a patient requiring hospitalization. 5 days of incubation are followed by 5 days of symptoms followed by a hospital stay of 8-16 days depending on ICU needs.

12. Meanwhile, the generation time of the disease is around 6 or 6.5 days. The exact numbers don't matter a lot, but assume that each infection initially generates 2.4 new infections (R0=2.4).

This puts the doubling time at around five days.

13. Now suppose you have the option of enacting strong non-pharmaceutical interventions (NPIs): school and business closures, social distancing, etc. The epidemic is doubling every 5 days.

Suppose you wait until the crisis point where hospital demand reaches hospital capacity.

14. You institute NPIs; new transmissions slow.

But think back to the timeline I drew. Even if your NPIs slow down transmissions enough that each case creates R0<1 additional cases, you are still in big trouble—there are a lot of not-yet-hospitalized cases in the timeline.

15. In fact, you've got 10 days of cases not yet hospitalized but infected before your NPIs were instituted. Doubling time is 5 days. So 10 days out there will be 4x as many people needing hospitalization each day as there were when you reached crisis. No way to handle that load.

16. That's what happens if you don't #FlattenTheCurve, and wait until crisis before taking strong action.

If you flatten the curve, you have more time to respond once you hit the crisis point.

17. Suppose you proactively reduce R0 to 1.1. When you hit crisis, you implement further NPIs. You've got 10 days of cases in the timeline, but with R0=1.1 and a 6 day generation time, you're looking at only about 20% increase in hospital demand by the end of those 10 days.

18. My view is, and always has been, that it will be difficult to muster the political and social will to implement strong enough NPIs to reduce R0 below 1.

Only once we see a true crisis, with hospitals reaching or exceeding capacity, will this be feasible.

19. If we fail to flatten the curve in advance, we will overshoot capacity manyfold, with horrible consequences and forcing dreadful triage decisions.

If we flatten the curve in advance, we will only slightly overshoot capacity even if we wait until crisis to act decisively.

20. Why not just institute strong NPIs well in advance of reaching hospital capacity? In a perfect world, we would. But this isn't a such world.

People aren't willing to shoulder the burden of extraordinary inconvenience that NPIs require until they SEE the threat firsthand.

21. Given that, we have to make sure that we can correct course with a minimum of additional strain on the healthcare system. And that means encouraging people to take strong social distancing measures in advance of the crisis. In other words, to #FlattenTheCurve.

/fin

Addendum: this was a long and somewhat technical explanation. Simple metaphors can be useful. The thread below is a first attempt a very simple metaphor. https://twitter.com/CT_Bergstrom/status/1239843765814091777