Estimating When New York Will Run Out of Hospital Beds

“The New York City hospitals are on the brink of being maxed out in terms of their available capacity,” warned former Food and Drug Administration commissioner Scott Gottlieb to kick off the trading day on CNBC. In between flashes of the futures markets and bumpers for Ben Bernanke’s upcoming appearance, the picture Gottlieb was painting suggested grave, imminent danger. “New York has another about five weeks to go for this between now and when they’re going to hit peak hospitalizations, so the fact that they’re stretched right now is worrisome.”

He confirmed the fear of physicians in the emergency rooms across New York shared in the Times ten days ago: America’s biggest city does not have the capacity for the COVID-19 epidemic. Mayor Bill de Blasio predicted in that piece that New York City would have 1,000 cases the following week. As of yesterday, it has 12,305.

While recovering myself from the virus, I started digging into the available public data to see if we could forecast when New York will run out of hospital beds. Thanks to my friend Abe, I learned about a public data source published by the city each week day on the number of emergency room visits based on the reported symptoms. Called EpiQuery, one of its many datasets are for surveillance of epidemic spread such as the one fueled by COVID-19.

Two of the five symptom sets fit the pandemic — respiratory and influenza-like illness. Combining ER visits reporting these symptoms, we can see a spike higher than any observed since EpiQuery began tracking in 2016. Yesterday was the highest number of ER visits in New York yet, a daily peak 16% higher than observed during 2018’s gnarly flu season and 237% higher than it would be on normal March day.

This trend clearly follows the news around the virus spread in the city over March. With a daily updated data source, we can take a stab at forecasting when New York’s available hospital capacity will be exhausted. To get a reasonable estimate, we’ll need to do three things:

1. Gather data on New York City’s current capacity and the hospitalization levels demonstrated by COVID-19 thus far.
2. Model the ER visits and produce a near-term forecast.
3. Combine the forecast and hospital data to estimate when New York’s capacity will be reached.

New York’s Hospital Beds and COVID-19’s Hospitalization Rate

New York State’s Department of Health publishes the total number of beds for every hospital in the state, including those in the New York metro area. The data are rich, breaking down the types of beds including the number of intensive care beds that are required for COVID-19 patients requiring ventilation. A quick web scraper can grab these values for us ultimately finding that DoH reports 57 hospitals in the metro area with 23,922 total beds.

Obviously, the total capacity is not an indication of the available capacity for New York — these 23k beds were not sitting empty waiting for a global pandemic to come along. News reports from last week indicated that ~65% of New York’s hospital beds are currently occupied which appears to jive from this report from 2018. This range does conflict with an analysis published by Harvard last week which suggests occupancy of 80% in The Bronx and Manhattan. From this research, I think we can assume the real availability is somewhere between the two with New York entering March with 8,373 available beds as the most optimistic case.

Finally, the regular media updates from the governor’s office gives us a sense of how many of these emergency room visits will result in hospitalization. As of Monday, 13% of the state’s COVID-19 cases required hospitalizing the patient with a quarter of those (3.2%) requiring intensive care.

Modeling The Outbreak In New York

With a sense of the total and available capacity as well as how many will be needed as a percentage of the infected population, we can shift our effort to modeling the outbreak in New York. Here we get some sorely needed good news this week: social distancing does seem to be slowing the spread of COVID-19 in New York.

Fitting a SARIMAX model to capture the seasonality of these symptom sets, we can forecast that daily ER visits with COVID-19 symptoms will flatten out over the next 45 days and could possibly decline.

This forecast seems to reflect what the governor’s office is estimating today. Hopefully, we’re both right.

That said, 4,900 ER visits a day will add up quickly and remains solidly higher than the previously observed peak two years ago. This peak period is also forecast to last twice as long as peak periods from previous years.

In short, more New Yorkers will be going to emergency rooms than we’ve seen before and they will be doing so continuously for the next month and a half.

Estimating When We Run Out

With a reasonable forecast of ER visits and the hospitalization rate, we can start to take a guess at when New York City will run out of available hospital beds. If we apply a 13% hospitalization rate and cumulatively add those patients over our forecast, New York City will likely run out of hospital beds by the middle of next week.

COVID-19 hospitalizations passed our optimistic estimate on available capacity this past weekend. However, New York hospitals prepared over the past two weeks to increase capacity and delayed elective procedures to help reduce occupancy. The arrival of federal resources should also improve the capacity picture, though the ready-to-use options still only adds 1,000 beds at a time. Assuming they were able to boost availability by 50% like the governor ordered, New York will likely run out of beds between 31 March and 2 April.

The story with the intensive care units is far more grim. COVID-19 patients requiring ventilation probably exceeded New York City’s capacity eight days ago.

When Governor Cuomo declares that New York needs ventilators now, he means yesterday. According to the best case forecast above, New York City will need 181 additional ventilators every day for the next 45 days.

Peeking Into Summer

Forecasting such an anomalous event far into the future is a low-confidence venture. But in the spirit of leaving the reader with some dim optimism, the model does suggest ER visits related to these symptoms can decline to normal flu season levels by August.

This would represent the absolute most optimistic forecast based on the data we’ve observed so far. But possibility is in short supply in New York this week and even the extreme lower bound of a confidence interval is welcome comfort during these times.

If you’re similarly data-inclined, I’ll work on publishing the Jupyter notebook and data soon — your feedback is warmly welcome.

Update [25 March 2020–7:28pm EDT]: The New York Times published a piece shortly after I pressed publish on this one with comprehensive reporting that multiple New York hospitals ran out of ventilators and other critical equipment. The video also suggests ER visits were twice their normal level and resulting in a higher number of hospitalizations.

Update [26 March 2020–9:09am EDT]: Another piece was published this morning in The New York Times covering the situation at Brooklyn Hospital Center. The chief of the emergency department indicated their intensive care unit was already at max capacity and she expects to run out of beds in the middle of next week.