Thread by @jasoncrawford

A new paper from @imperialcollege's #COVID19 Response Team models different interventions—from “stay home when sick” to “close the schools”—and estimates their effect on cases and fatalities.

I read the paper and will summarize it here. It is sobering.

imperial.ac.uk/news/196234/co…

1-tweet summary:

• Millions of lives are at stake
• Intermittent but ongoing society-wide measures needed to suppress the outbreak & avoid swamping ICUs
• Merely slowing the spread without reducing # of cases can only reduce deaths by ~50%
• “Stay home when sick” not enough

Any model necessarily makes assumptions and leaves out some factors, and I think there are some critical ones left out here, which I'll note.

But the paper is important because it makes clear the magnitude of the threat.

The red line at the *bottom* is ICU bed capacity:

This risk of swamping the health care system—and of an unconscionable number of deaths—was already clear from napkin math, IMO. But not everyone believed it.

Now with an academic paper and a more sophisticated model, maybe more people will believe it.

https://twitter.com/jasoncrawford/status/1238689938696073216

I'll give the paper's conclusions first, then ask how much we can trust the model; including what I think the model is missing (some significant things), and how much that matters (not much, for the bottom line).

First, what's the worst case—if we all, stupidly, just ignored this completely?

The paper estimates that, were the virus allowed to spread uncontrolled, it would infect over 80% of the population—resulting in over 500,000 deaths in the UK and over 2 million in the US.

In this worst-case scenario, “critical care bed capacity would be exceeded as early as the second week in April, with an eventual peak in ICU or critical care bed demand that is over 30x greater than the maximum supply” in both the US and UK.

This is what's at stake—millions of lives in the US and UK alone.

Fortunately, we're already starting to act, so this worst case won't come true.

But how much of it might come true? How much will interventions help, and which ones do we have to take?

Here are the interventions considered by the model:

Case Isolation: Stay home if sick

Home Quarantine: Whole family stays home if one gets sick

Social Distancing: Go to school/work, but stay home more, reduce other contact (by all or just those over 70)

Close Schools: obvious

The paper considers two strategies: (1) “suppression”, where we prevent the virus from infecting everyone, and (2) “mitigation”, where we let it infect everyone, but slowly.

(Note that the usage of these terms may differ from technical usage among at least some epidemiologists, who use “mitigation” vs. “containment” to mean somewhat different things. In this thread I'll stick with the paper's usage of terms.)

https://twitter.com/jasoncrawford/status/1239352355054706688

The advantage of mitigation is that you build herd immunity, and once the disease has burned through the population, it's over—but many more people get infected.

With suppression, you save lives, but you have to keep up the measures indefinitely, until a vaccine or treatment.

Of the interventions the paper considers, the *best-case* scenario for the “mitigation” strategy is to save *half* the lives.

In other words, still over *1 million* deaths in the US—making it by far the top killer of 2020.

(Calculations and table from nytimes.com/interactive/20…)

The “suppression” strategy can do much better.

Because intense measures can't be sustained indefinitely, the paper considers a “trigger” model: broad interventions start when the # of ICU cases hits a threshold, and last until it is below another threshold.

Here's what it looks like. The orange line is the number of cases in the ICU; the blue rectangles show when the broad interventions are in place. (Case isolation and home quarantine are in effect the whole time.)

This strategy can reduce the number of deaths by 90% or more.

“We therefore conclude that epidemic suppression is the only viable strategy at the current time.”

Until a vaccine is available, or until the number of cases is low enough that we can “adopt intensive testing, contact tracing and quarantine measures” like South Korea.

The model also shows that, no matter what strategy we choose, no one of these interventions is sufficient. We have to use a combination of multiple interventions to have a large impact.

How much can we trust the model?

It is based on real data wherever possible, e.g., for population density, age distribution, and the R0 or spreading factor.

They also test sensitivity to various parameters, such as R0, the trigger threshold, and the case-fatality rate.

To my mind, the most significant thing they *don't* model, as far as I can tell, is the increase in the case-fatality rate when we exceed ICU capacity. (!)

But this only makes the bottom line *stronger*, which is that moderate measures will cost millions of lives.

On the optimistic side, they also don't consider the potential to ramp up ICU capacity, which IMO we *must* do.

But again the conclusion remains: strong action is needed.

Some other things that aren't considered:

• Seasonality: is it less virulent in summer?
• Re-infection (they assume lifetime immunity)
• Other interventions, such as masks, or working from home

None of this is meant as criticism—any model necessarily makes simplifying assumptions. But these factors might be useful to model in followup work.

I hope this makes clear to everyone why extreme measures are needed, such as canceling events, or closing schools and restaurants. The situation is dire. These reactions are not “panic”—they are rational, like evacuating a city during a hurricane.

My (evolving) view of the solution is something like:

• Short term: Broad-based isolation; increase ICU capacity

• Medium term: Fast, cheap, ubiquitous testing; case isolation and contact tracing (end/relax broad-based isolation)

• Long term: Vaccine and/or pharmaceuticals

We will defeat this scourge. We will emerge victorious—stronger, even. But not if we indulge in blind optimism. We must face the stark reality of the danger we're facing and meet with swift, bold, decisive action.

And now one of the authors of the paper may have COVID-19. Best wishes for a speedy recovery @neil_ferguson!

https://twitter.com/neil_ferguson/status/1240171876695117824?s=21

https://twitter.com/neil_ferguson/status/1240171876695117824