key: cord-0767263-a1pbwdwx authors: Olney, Andrew M; Smith, Jesse; Sen, Saunak; Thomas, Fridtjof; Unwin, H Juliette T title: Estimating the Effect of Social Distancing Interventions on COVID-19 in the United States date: 2021-01-07 journal: Am J Epidemiol DOI: 10.1093/aje/kwaa293 sha: 47a0dabab013c6f58ca1d2ed754d2d41f36288ec doc_id: 767263 cord_uid: a1pbwdwx Since its global emergence in 2020, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused multiple epidemics in the United States. Because medical treatments for the virus are still emerging and a vaccine is not yet available, state and local governments have sought to limit its spread by enacting various social distancing interventions such as school closures and lockdown, but the effectiveness of these interventions is unknown. We applied an established, semi-mechanistic Bayesian hierarchical model of these interventions on SARS-CoV-2 spread in Europe to the United States, using case fatalities from February 29, 2020 up to April 25, 2020, when some states began reversing their interventions. We estimated the effect of interventions across all states, contrasted the estimated reproduction number, R(t), for each state before and after lockdown, and contrasted predicted future fatalities with actual fatalities as a check on the model’s validity. Overall, school closures and lockdown are the only interventions modeled that have a reliable impact on R(t), and lockdown appears to have played a key role in reducing R(t) below 1.0. We conclude that reversal of lockdown, without implementation of additional, equally effective interventions, will enable continued, sustained transmission of SARS-CoV-2 in the United States. CoV-2 is attributable to its transmissibility by aerosol and fomites 1,2 , by presymptomatic/asymptomatic carriers 3, 4 , and by the relatively mild clinical characteristics of symptomatic carriers, which often include fever, cough, and fatigue 5 . However, approximately 20% of confirmed cases develop severe or critical forms of COVID-19, including complications of respiratory failure, myocardial dysfunction, and acute kidney injury, with approximately 50% mortality for critically-ill patients 6 . As of July 2020, outbreaks or epidemics of SARS-CoV- 2 Sport is the banning of sporting events or public gatherings of more than 1000 persons. Public events is the banning of public gatherings of more than 100 participants. Finally, lockdown includes banning of non-essential gatherings or business operations, which is sometimes formalized as a stay-at-home or safer-at-home order. Notably some more restrictive interventions imply others, e.g., lockdown implies all other interventions, and public events implies sport. Table 2 (see also Web Figure 1 ). Of primary interest are R t estimates before and after lockdown and corresponding forecasted death counts 2 weeks into the future. Across states, the mean R t before lockdown was 1.86 (SD=0.56, range: 1.00-3.37) and the mean R t after lockdown was 0.88 (SD=0.25, range: 0.50-1.41). Notably, no state had a mean R t below 1.0 pre-lockdown, but 29 states had a R t below 1.0 after lockdown. While lockdown was associated with reduced R t in all states that underwent lockdown (a 54.4% reduction, see Table 1 ), in these 29 states, lockdown appears to have been the single critical intervention allowing containment of the disease. In the remaining states, pre-lockdown R t was too high (i.e., greater than 2.2) for lockdown to bring R t below 1.0. Predicted deaths vs. actual deaths two weeks into the future in each state serve as a validity check on the model's estimates of intervention effects (see also Web Figure 2 ). Fortyfive states (90%) had actual deaths that were within the 95% CI of predicted deaths. Notably, the mean predicted deaths were well above actual (>100 deaths) for Connecticut, New Jersey, Massachusetts, and New York. The mean absolute error of mean predicted deaths was 50.80, and without these four states the mean absolute error was 10.08. As expected, the model fit to actual deaths was even closer on the observed data, with mean absolute error at 5.90 (N=2951). Our study has several limitations. First, the assumption that all interventions have the same implementation and effect in all states is a strong assumption. For example, the public events intervention banning gatherings of 100 persons or more could be met by a ban on 10 persons or more or 50 persons or more; it is unlikely that such bans are truly equivalent. Schools or universities closing treats primary, secondary, and higher education the same, though emerging evidence suggests that younger children may be less effective at spreading the virus than adults 17 . This limitation has since been partially addressed in the European model by allowing random effects for lockdown only. Second, the assumption that interventions are binary, instantaneous, and non-harmful are strong assumptions and oversimplifications that do not account for time-varying compliance with intervention or unintended consequences. Using mobility data as a measure of population mixing 14, 18, 19 partially addresses this. Third, the parameters of the model are estimated using reasonable, but still uncertain, assumptions about prior distributions. We have used the same assumptions as in the European model, but these assumptions may be contradicted by future empirical work. Modeling of SARS-CoV-2 is emerging and rapidly diversifying, including classical SEIR models and derivatives 20 , deep learning 21 , and piecewise models for sub-exponential growth 22 . State and local governments are likewise rapidly adjusting policy decisions regarding interventions based on case data and economic concerns. As the United States adopts an increasingly fragmented response to SARS-CoV-2, modeling approaches like ours that focus on shared interventions may not be tenable. While our results give valuable insights into which interventions did and which did not change the transmission rate substantially, we recommend Aerosol and surface stability of SARS-CoV-2 as compared with SARS-CoV-1 Turbulent gas clouds and respiratory pathogen emissions: potential implications for reducing transmission of COVID-19 Presumed asymptomatic carrier transmission of COVID-19 Evidence supporting transmission of severe acute respiratory syndrome coronavirus 2 while presymptomatic or asymptomatic. Emerg Infect Dis Clinical characteristics of coronavirus disease 2019 in China Intensive care management of coronavirus disease 2019 (COVID-19): challenges and recommendations Estimating the number of infections and the impact of non-pharmaceutical interventions on COVID-19 in European countries: technical description update Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe US state social distancing intervention dates Estimates of the severity of COVID-19 disease Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand. Imperial College London American Community Survey Demographic and Housing Estimates, Table DP05 The New York Times. Nytimes/Covid-19-Data. The New York Times State-level tracking of COVID-19 in the United States Infectious disease epidemiology A probabilistic programming language Contact tracing during coronavirus disease outbreak Projections for first-wave COVID-19 deaths across the US using social-distancing measures derived from mobile phones. medRxiv Forecasting the impact of the first wave of the COVID-19 pandemic on hospital demand and deaths for the USA and European Economic Area countries Initial simulation of SARS-CoV2 spread and intervention effects in the continental US Forecasting the COVID-19 trajectory Reproductive number of the COVID-19 epidemic in Switzerland with a focus on the cantons of Basel-Stadt and Basel-Landschaft