Latest COVID-19 Projections Point to Spring Peak

Data models show that late April could be the peak for new U.S. cases nationally and a low point of hospital critical care capacity

April 22, 2020

The latest data modeling projections by Columbia University Mailman School of Public Health scientists estimate that, nationally, new COVID-19 cases will begin to decline in the third week of April from a high of roughly 20,000 projected new cases daily, assuming a 30 percent reduction in social contact going forward. Under a 20 percent reduction in social contact, COVID-19 cases may not peak nationally until mid-May with approximately 30,000 new daily cases projected at the peak.

Three online tools allow users to visualize the updated weekly projections of new COVID-19 cases, new infections, and available critical care beds in states and counties across the United States under a variety of social distancing and hospital response scenarios over a six-week period: (1) a data visualization tool that graph projections over time, (2) a mapping tool that charts county-level projections, and (3) animated maps.

Columbia researchers are advising and providing projections to the White House Coronavirus Task Force, the U.S. Centers for Disease Control and Prevention, New York City, and various state governments.

The researchers’ projections account for varying degrees of social contact based on measures such as the closures of restaurants, bars, schools, playgrounds, and recommendations to wear a mask outside. Hospital surge response scenarios may include actions such as clearing non-COVID-19 patients from critical care beds, repurposing beds outside of ICUs for use in critical care, and the ability to use a single ventilator for multiple patients. Estimations do not reflect specific measures or levels of local preparedness.

Additional insights from projections generated on April 20 (50th percentile estimates) include:

  • Nationally, hospital critical care bed availability will be lowest at the end of April, assuming a 30 percent reduction in social contact going forward and a medium intensity hospital surge response. The availability of beds will vary widely across counties; areas projected to peak later will likely see critical bed shortages later, too.

  • Individual states will peak at varying times under a 30 percent reduction in social contact going forward. New York State may have already peaked, to be followed soon by California, Florida, Illinois, New Jersey, and other states that have already seen a high number of cases. Alaska, Alabama, Iowa, Kentucky, Maine, Mississippi, Oklahoma will not peak until May or later, although the number of cases there will remain relatively low.

  • Critical care bed demand will exceed capacity in Queens County, New York, until early May, even assuming a high-intensity surge response. Manhattan is projected to have sufficient critical care capacity to care for the cases arising among its residents.

“These projections provide policymakers with estimates of when their healthcare systems are at a high risk of being overwhelmed or severely stressed. It also reinforces the importance of social distancing in controlling the transmission of the virus,” says Andrew Rundle, DrPH, associate professor of epidemiology, who led the development of the data visualization and mapping tools.

In addition to Prof. Rundle, members of the research team include Charles Branas, PhD, professor and chair of epidemiology; Sen Pei, PhD, associate research scientist of environmental health sciences; and Jeffrey Shaman, PhD, professor of environmental health sciences. The data visualization tool was created by Tonguc Yaman, a doctoral student in the Department of Epidemiology.

The researchers use a computer model to project the spread of COVID-19 cases in the continental United States by accounting for transmission and case-fatality parameters, population movement, including by commuters, with U.S. Census Bureau data, and other constraints and variables. Updated projections posted on Github include data on hospitalizations, critical care admissions, and mortality under varying levels of social distancing and hospital surge response.

The projections contain uncertainties, such as the effects of various social distancing measures that may have been put in place in the last two weeks, as well as hospital capacity parameters, such as newly constructed field hospitals, that were not trackable in real-time and therefore not part of the models. The models do not account for potential shortages of medical staff who become infected, nor do they account for the ingenuity and perseverance of these staff, many of which are trained to work in crisis situations. Models also do not account for deaths from other illnesses and injuries that hospitals may not be able to service in the midst of the COVID-19 surge. As such, projections may change as the situation unfolds.