IHME sharply reduces projected hospitalizations as data shows NY's curve is flattening; Update: Deaths in NY flat, not rising; Update: Hospitalizations down

A sequel to yesterday’s post about glimmers of hope in New York State. The IHME model, which has influenced the White House’s own projections, has been close to the mark in projecting deaths in NYS thus far. (In fact, they may be conservative on deaths due to the likelihood that COVID-19 fatalities are being undercounted.) But they were massively off on projected hospitalizations due to the disease. Here’s what the model expected on April 4 in New York State as of a few days ago:

The reality on April 4, according to Andrew Cuomo, was 16,479 people hospitalized. The IHME model overstated the actual number by 400 percent.

We can only speculate as to why. IHME assumes widespread social distancing, as is being practiced in New York right now, so they weren’t expecting a worst-case scenario. Their model is more like a best-case scenario, yet they were still wildly off. It could be that therapeutic treatments in New York are leading to faster than expected recoveries. It could be that social distancing is even more effective in slowing the spread than IHME expected. It could be that the rate of severe cases among infected people is simply lower than anyone thought (although in that case we should expect the death projections to be wildly off too, and they aren’t). It could be, as a Twitter pal theorized, that hospitals in New York are so overloaded right now that patients who are struggling but not quite at death’s door are being sent home to recuperate even though they’d benefit from hospitalization if enough beds were available.

Whatever the answer, the new IHME projections for hospitalization in New York are waaaaaay down. This is what they’re now expecting for the peak, which should come in three days or so:

That wasn’t the only projection that changed overnight:

An important lingering question is whether the number of daily deaths will continue to increase this week in New York, as IHME projects (with a high of 878 expected on Thursday), or if the one-day decline noted by Cuomo in yesterday’s official numbers turns out to be the start of a downward trend. If in fact NYS has already peaked and is beginning to see a *decrease* in deaths at a moment when models expect deaths to increase for at least the next four days, we’ll have a mystery on our hands. Why did New York peak early? Did the draconian lockdown simply choke off new transmissions faster and more sharply than anyone believed possible?

A new paper published by Jeffrey Harris, an economist at MIT, points to growing evidence that New York has indeed flattened the curve:

Note the dates. Things start to flatten out around March 21, which is also around the time that the fever rate in New York City tracked by Kinsa’s “health weather” map began to reverse course and decline. Harris attributes that partly to Bill de Blasio’s belated order on March 15 to close NYC restaurants and bars. (Kinsa’s founder, Inder Singh, echoed that point in a post today based on his own reading of the company’s data. “Kinsa’s county-level illness data makes a clear point: when schools, restaurants and bars are closed, effectively forcing families into their homes, fevers drop.”) But Harris sees another factor: Simply put, the more aware locals are of a spreading epidemic, the more they’ll self-isolate without prompting.

There is, however, another strand in the economics literature suggesting that people voluntarily engage in avoidance behaviors once they fully perceive the risks of contagion. In a study of the time path of the H1N1 Swine Flu epidemic of 2009 in the U.S. – when there were no quarantines or lock downs – economists could see no other explanation for their finding that the incidence of new cases fell well below the exponential path predicted from epidemiologists’ classical models (Yoo, Kasajima, and Bhattacharya 2010). Similarly, the incidence of HIV reached a peak in 1983 well before the advent of the first antiviral medication in 1986 (Harris 1990a, Harris 1990b) and about a year before San Francisco’s Director of Public Health ordered the closure of fourteen bathhouses in October 1984. In the present context, ridership on New York City subways was already down 19 percent by March 12 and 60 percent by March 16 (Metropolitan Transportation Authority 2020). In all of these cases, the critical ingredient in the public policy mix may have been the successful communication of consistent, clear, accurate and timely information to millions of individuals, who responded by taking action without government coercion. Put bluntly, what flattened the curve was no more than the naked truth.

That explains why states like Iowa and Oklahoma, where no statewide stay-at-home orders have been issued, nonetheless show dramatic drops in fever rates over the last few weeks. During a global outbreak, most people don’t need the governor or the mayor to formally bar them from congregating with others to convince them that it’s a bad idea to do so. (Some do, like the spring-break idiots, which is why the stay-at-home orders have such an impact on fever rates.) The flip side of that, though, is that having Trump walk out onto the White House balcony and declare that it’s time to reopen for business isn’t going to move many people. They’re not looking to him, they’re looking to the experts and to the data. Convince people that the contagion is in hand and they’ll take the risk of going shopping again.

Lotta suspense now to see what Cuomo’s numbers look like today and the next few days. Did New York turn the corner early or is it still on its way up to the summit of a much-flattened curve? Either way, there are still thousands of deaths in New York to come. And the sort of benchmarks described in this piece for reopening the local economy — two weeks of declining cases, normal hospital capacity, testing everyone with symptoms, and monitoring contacts — are nowhere in sight. New York might meet the first three marks by early May, but I don’t know how they’re planning to meet the fourth.

Update: This poll is a bit outdated since Trump has backed off reopening the country on Easter, but here’s further evidence that most people aren’t coming out of their homes until it’s safe, not when some arbitrary deadline is imposed. Note that even a majority of Republicans declined to heed the guidance of reemerging by Easter back when Trump was still toying it. Partisan loyalty isn’t going to convince most people to risk their and their family’s health.

Update: And there’s the good news we’ve been waiting for. Yesterday Cuomo reported 594 deaths on Saturday; today he’s reporting 599 deaths on Sunday. Deaths aren’t continuing to skyrocket. Maybe New York really is peaking earlier, and lower, than everyone expects.

Update: This is even better news. Three straight days of declining hospitalizations.