I’m glad that I started paying attention to the dispute between the Imperial College and Oxford before watching this or I would have been confused as hell. But if, like me, you’ve been following that divergence, you get what she’s saying here. Bottom line: We still have no real idea how many people have this disease and aren’t showing symptoms, for whatever reason. It could be a small number. It could be an immense number. The death and hospitalization rates depend entirely on knowing, as does our national response. Pause here to watch the clip (or at least the first two minutes), then read on.

The second half of the clip finds her attempting to reassure Americans that there are still ventilators and ICU beds available in New York, which is nice, but for how much longer? No one’s going to feel reassured if she’s back on Monday to say, “Okay, update, every bed and machine is now taken.” It’s strange to see her urging everyone to chill out based on a snapshot at a particular moment in time when the rest of the daily message from her and Fauci is that we’re in this for the long haul and things are about to get very bad, and not just in New York.

The first half of the clip is about what the models are saying about COVID-19’s fatality rate. Right now, the dire Imperial College model and the much rosier Oxford model agree: There are likely to be tens of thousands of deaths in the UK, not the massive six- or seven-figure numbers that dire models projected. But they agree for drastically different reasons, and the consequences to the UK and the U.S. will be drastically different depending upon which is correct. The Oxford model speculates that there’s a gigantic number of people who’ve already been infected by coronavirus and didn’t realize it because they were asymptomatic. That’s what Birx is getting at in comparing the number of infected people to an iceberg, with the bulk of the iceberg still hidden from us under the water. We need antibody tests to have a better sense of how big it is. If it’s very big, that means that COVID-19 is no big deal for the vast majority of us; we’re racing towards herd immunity, approaching the peak of a very tall and sharp epidemic curve, and will soon be on the other side of it. This won’t be an 18-month problem. It’ll be a crisis of a few months, with only a tiny percentage of infected becoming very ill. Although that tiny percentage will still be a large number, tens of thousands of people, because the total infected population is so gigantic.

Our behavior doesn’t matter hugely to the Oxford model. Flattening the curve will help somewhat in lightening the load on hospitals, but the virus has already spread too far to slow it down dramatically. It’s been moving at lightning speed; now it’ll move at slightly less than lightning speed. And most of us will be just fine.

The Imperial College, whose model projected a few million deaths in the U.S. in a worst-case scenario, currently projects just 20,000 deaths or so in the UK. But that’s not because they’ve swung around towards the Oxford model’s assumption that there’s a giant iceberg of asymptomatic carriers, as I mistakenly initially thought yesterday. They think the iceberg is comparatively small, with a higher percentage of infected people developing severe symptoms. The reason they’ve lowered their death projections is because the U.S. and UK have locked down, imposing draconian social distancing practices to limit the spread of the disease. That’ll slow down the virus dramatically … until we lift those practices, when it’ll start spreading again. If we continue to turn social distancing “on” and “off” as needed over the next 18 months to contain sporadic outbreaks, then sure, we can hold down the death toll. But this won’t be over in a few months. And the people crowding into ERs right now aren’t extreme flukes relative to the total infected population. Your chances of needing hospitalization are waaaay higher in this model.

Needless to say, our behavior matters enormously in the Imperial College model. The whole strategy is predicated on the government’s and the public’s willingness and ability to orchestrate social distancing in intelligent ways. This is going to go on for a year, maybe more. Prepare accordingly.

What Birx was trying to do at the briefing, if I understand her correctly, is encourage people by noting that there’s evidence out there right now based on observed death rates that the iceberg of asymptomatic cases is bigger than we thought. Which makes sense, of course: No one but no one believes that there are “only” 100,000 positive cases in the U.S. Our ability to detect the disease is hamstrung by our inability to test widely enough. The question is simply how much bigger the iceberg is than we think. Are there a million actual cases of COVID-19 right now? Or are there 10 million? With each order of magnitude, the decimal point on the case fatality rate moves one place to the left. Conceivably, COVID-19 might be no more lethal than the flu across the population. Conceivably, it’s much more lethal.

Actual reality will end up somewhere between the Imperial College and Oxford studies once we do the blood tests we need to find out. But which one will we resemble more? A lot depends on the answer.

The strategy touted by Birx, Fauci, and others for dealing with a sustained outbreak a la the Imperial College model is to ramp up testing and contact tracing while the whole country is frozen at home right now, then have people gradually reemerge once those are in place. It’ll be a reboot of sorts now that we’ve slowed down the virus via lockdowns, with the U.S. suddenly shifting to South Korean practices of mitigation. But here’s the wrinkle, per Harvard epidemiologist Marc Lipsitch: There’s … little reason to think we’re capable of doing that. At least anytime soon.

If at the same time we ramp up testing capacity and the ability to trace contacts, in a very best-case scenario, we might conceivably be able to turn to a Korea- or Singapore-style mix of less intense distancing combined with super-intense contact tracing, isolation, quarantine

— Marc Lipsitch (@mlipsitch) March 27, 2020

But here's why it is a best-case, likely unattainable scenario. 1) We have not proven that US-style social distancing can produce R_effective<1 (declining case numbers). On this I'm hopeful, but it's a hope not a fact.

— Marc Lipsitch (@mlipsitch) March 27, 2020

Solving the testing problems will not be easy, despite heroic local and state efforts to make up for the feckless federal efforts on this front. Supply chains are delicate and it may not be possible to come from behind and establish Korea-level testing capacity.

— Marc Lipsitch (@mlipsitch) March 27, 2020

4) If we manage all of that, there will still be a world of COVID-19 transmission throwing sparks back at us. Like China today https://t.co/AXprlQAxC7 we will be in a long-term effort to prevent these sparks from starting new chains of transmission.

— Marc Lipsitch (@mlipsitch) March 27, 2020

So the scenario Dr. Birx is "assuring" us about is one in which we somehow escape Italy's problem of overloaded healthcare system despite the fact that social distancing is not really happening in large parts of the US https://t.co/aNnoduaR3O

— Marc Lipsitch (@mlipsitch) March 27, 2020

For the 12-18 months (best case under current models) till a vaccine. I desperately hope she is right, because much suffering will be avoided. But reassurance that this is likely, or even plausible, with the disorganized track record of the US response, is false reassurance.

— Marc Lipsitch (@mlipsitch) March 27, 2020

But it would be extremely naive to imagine that they are likely, and what do I know, but it seems like bad politics to "assure" the American people that they will come to pass when so many things could go wrong, any one of which leads to much worse outcomes.

— Marc Lipsitch (@mlipsitch) March 27, 2020

Fauci and Birx have been notably coy lately in talking about our prospects for testing and tracing. Testing continues to improve, but it would need to improve a lot — nationally — to believe we can catch new outbreaks anywhere they might arise. Aggressive contact tracing may simply be beyond our means, as Lipsitch says. If that’s true, then what’s the point of this “reboot”? We slow the virus down for a month, say, and then have a normal-ish month, and then … go back into lockdown for a month? On and off, normalcy versus self-quarantine, in alternating months indefinitely? We need to hear more about the post-lockdown mitigation strategies. Particularly whether they’re even feasible.