COVID-19 Projections: Delayed Response to Rebound Would Cost Lives
New projections from Columbia University Mailman School of Public Health scientists find that delays in reimplementing social distancing following a relaxation of control measures could result in a stronger rebound of COVID-19 infections and deaths. In a retrospective counterfactual analysis, they find that if social distancing and other control measures had been in place in major U.S. metropolitan areas one to two weeks earlier, more than half of the infections and deaths seen to date could have been prevented.
Currently, daily confirmed cases are in decline after the easing of social distancing measures, due in large part to the impact of measures in place before May 4. This decreasing trend, coupled with the lag between infection and diagnosis, conveys a “false signal” that the pandemic is under control. However, due to the high remaining population susceptibility, a large resurgence of both cases and deaths will likely occur, peaking in early- and mid-June, even with the possible reimplementation of social distancing measures.
Results are published in the preprint server MedRxiv. [November 6 Update: the study is now published in the peer-reviewed journal Science Advances. The numbers in the Counterfactual section below have been updated below.]
Hypothetical: How Quickly Will Authorities Respond to a Rebound?
The scientists compared two hypothetical scenarios in which social distancing measures are reimplemented two to three weeks after being relaxed nationally beginning May 4. The difference between the two- and three-week delays could be 214,545 additional confirmed cases and 23,110 deaths nationally by July 1. In the case of a two-week delay, they estimate a daily peak of 35,288 new cases and 3,392 new deaths by July 1; in the case of a three-week delay, they estimate a daily peak of 42,560 new cases and 4,166 new deaths by July 1.
Study results indicate that the rapid detection of increasing case numbers and fast re-implementation of control measures are needed to control a rebound of outbreaks of COVID-19. The models assume an ability to re-implement a 25 percent weekly reduction of transmission rates nationwide. Yet, due to fatigue in the general public towards social distancing and consequent reduction in compliance, researchers say this assumed reduction may be overly optimistic.
“Efforts raising public awareness of the ongoing high transmissibility and explosive growth potential of COVID-19 are still needed at this critical time,” says lead researcher Jeffrey Shaman, PhD, professor of environmental health sciences at the Columbia Mailman School of Public Health. “Our results also indicate that without sufficient broader testing and contact tracing capacity, the long lag between infection acquisition and case confirmation will mask the rebound and exponential growth of COVID-19 until it is well underway.”
Counterfactual: What If Social Distancing Measures Were Implemented Earlier?
In a retrospective analysis, the researchers find that, nationwide, 601,667 confirmed cases (53%) and 32,335 deaths (49%) of reported deaths up to May 3 would have been avoided if observed control measures had been adopted one week earlier—on March 8 instead of March 15. In the New York metropolitan area, 191,356 (73%) of confirmed cases and 16,950 (78%) of deaths would have been avoided if the same sequence of interventions had been applied one week earlier. Had the sequence of control measures occurred two weeks earlier, the nation would have seen a reduction of 1,041,261 (91%) cases and 59,351 (91%) deaths, and a reduction of 254,087 cases (97%) and 21,175 deaths (97%) in the New York metropolitan area.
The researchers note that their counterfactual experiments are based on idealized hypothetical assumptions. In practice, initiating and implementing interventions earlier during an outbreak is complicated by factors such as general uncertainty, economic concerns, logistics and the administrative decision process. Public compliance with social distancing rules may also lag due to sub-optimal awareness of infection risk.
About the Model and Its Uncertainties
The researchers used computer models to simulate the spread of COVID-19 cases by estimating transmission rates and other key parameters using population movement, observed cases, and deaths. There are a number of uncertainties, including the fact that their model may not account for changes in social distancing and contact patterns over the last two weeks. The impact of increased activity on contact patterns, the transmission of SARS-CoV-2 and COVID-19 incidence remains highly uncertain, as levels of compliance with social distancing, return to work, and consumer willingness to frequent businesses are unknown.
The current study focuses on the transmission dynamics in metropolitan areas with dense populations and abundant observations—New York, New Orleans, Los Angeles, Chicago, Boston, and Miami, among others—while also accounting for the spread of COVID-19 in less populated areas.
Co-Authors, Funding, Disclosures
Additional co-authors of the paper include Sen Pei and Sasikiran Kandula, both research scientists at Columbia Mailman School of Public Health. Shaman and Columbia University are partial owners of SK Analytics, an infectious disease forecasting company. Shaman also discloses consulting for Business Network International. This study was supported by funding from the National Institutes of Health (GM110748) and the National Science Foundation (DMS-2027369), as well as a gift from the Morris-Singer Foundation. The funders had no role in the design, data collection and analysis, decision to publish, or preparation of the manuscript.