Skip to main content

Our take —

This study demonstrates that the effectiveness of contact tracing may be limited, even in settings with very few cases. The model assumed that all cases and their contacts were successfully isolated and that there was minimal pre- and asymptomatic transmission, which we now know not to be the case. Thus, results should be interpreted as optimistic with respect to the potential effectiveness of contact tracing. Contact tracing in combination with other interventions may lower transmission rates, but only if it can be done quickly and cases and their contacts are effectively isolated.

Study design

Modeling/Simulation

Study population and setting

This modeling study simulated COVID-19 transmission from infected persons to their contacts and the potential impact of contact tracing in emerging epidemics (i.e., 40 cases or less in the population at the start of the outbreak) under a wide variety of scenarios, including varying levels of asymptomatic and pre-symptomatic transmission, reproductive numbers (range from 1.5-3.5), and contact tracing success levels. While the proportion of infected cases who were diagnosed varied across different scenarios, all individuals who tested positive were assumed to isolate and not transmit to any additional persons (i.e., 100% effectiveness of case isolation). Additionally, all traced contacts were assumed to immediately self-isolate and not infect any additional persons. Controlled outbreaks were defined as outbreaks with a final size of less than 5,000 cases.

Summary of Main Findings

Contact tracing was more likely to control a COVID-19 outbreak if: i)the number of cases to begin with was small (<20), ii)there were few pre- and asymptomatic cases as a proportion of all cases, and iii)there were shorter delays between symptom onset and case isolation. In moderate and high-transmission scenarios (Reproductive number, R0, equal to or greater than 2.5), the proportion of all contacts that need to be traced to control an outbreak exceeded 80%. Furthermore, even if we optimistically assume between five and 20 initial cases, the results show that large numbers of contacts (25-100 per week) would need to be identified and traced by public health departments in most outbreak scenarios.

Study Strengths

This study simulated the potential effectiveness of contact tracing of COVID-19 symptomatic cases under a wide variety of epidemic scenarios, demonstrating that contact tracing will be difficult even in settings with small numbers of cases. The authors clearly lay out all of the steps in the contact tracing process, and use the model to illustrate how gaps in each step might impact effectiveness.

Limitations

The study makes several assumptions that are unrealistic, including no onward transmission once cases are isolated. This model only applies to settings with very few cases (<40) and was published prior to learning that there is substantial pre-symptomatic transmission; thus, the results most likely overestimate the potential effectiveness of contact tracing to control COVID-19 spread.

Value added

This modeling study was among the first to show that contact tracing of COVID-19 symptomatic cases will have limited effectiveness if not done rapidly and in conjunction with other interventions.

Our take —

This study estimated the generation time (the time between the source and recipient infections) for 40 source-recipient pairs. Using data on symptom onset, authors estimate the proportion of transmission by a) asymptomatic transmission, b) pre-symptomatic transmission, c) symptomatic, and d) environmental transmission. A large proportion of transmission appears to occurs by pre-symptomatic individuals, suggesting that manual contact tracing may not be an effective intervention strategy alone. Technology-based approaches to contact tracing could help identify contacts faster, but could also create privacy concerns, and it is unclear how these tech solutions would link up with public health efforts to support isolation and quarantine.

Study design

Modeling/Simulation

Study population and setting

The authors used publicly available information on transmission pairs (from China, Germany, Italy, Singapore, South Korea, Taiwan, and Vietnam). Data were included if there was a high confidence of a true transmission event, and the timings of symptom onset was known for both the source and recipient infection.

Summary of Main Findings

This study estimated the generation time (the time between the source and recipient infections) for 40 source-recipient pairs that were chosen for their high confidence in direct transmission and the known times of symptom onset for both individuals. This was combined with information on the dates of symptom onset and intervals of exposure to estimate the proportion of transmission (measured by R0, the reproductive number) by a) asymptomatic individuals, b) pre-symptomatic (those who will become symptomatic but prior to showing symptoms), c) symptomatic, and d) environmental transmission. Using these estimates, the authors find that the majority of transmission occurs from the pre-symptomatic and symptomatic periods. The contribution to R0 by pre-symptomatic transmission alone was estimated to be 0.9 which is almost at the critical value (R0=1) that will lead to ongoing transmission. The authors then put these values in context of case isolation and contact quarantining which are key non-pharmaceutical interventions being deployed. These results suggest that manual contact tracing and quarantining will not be effective at containing the outbreak since a large proportion of transmission would occur prior to when case isolation and contact tracing could be achieved. These delays make traditional contact tracing less advantageous, and the authors recommend that additional avenues, such as technology-based contact tracing may be a more effective intervention. The authors provide a framework for this type of digital contact tracing and highlight limitations and opportunities for this type of intervention (such as privacy concerns, adoption for high risk groups, and the need for open and clear communication about the algorithm used).

Study Strengths

This study directly estimated the generation time from transmission pairs with a high confidence of direct transmission where the time of symptom onset for both the source and recipient were known. They highlight the importance of designing contact-tracing and quarantining interventions that take into account transmission that likely occurred prior to symptom onset.

Limitations

The framework used is reliant on known transmission pairs which may limit the generalizability of these results in areas with higher community transmission. Although the authors suggest that the use of a technology-based app that could quickly tell someone if they have recently been in contact with a case, they do not provide additional information on how that might be implemented or adopted by the general population, or how it would link up with other ongoing contact tracing efforts. The authors do not fully address concerns about adoption, privacy, and representativeness of this approach and the subsequent implications on the effectiveness.

Value added

This study provides estimates of the contribution of overall transmission from asymptomatic, pre-symptomatic, symptomatic, and environmental transmission using detailed data on source-recipient infections. These results suggest that interventions that depend on first identifying symptomatic individuals and their contacts will result in delays that greatly diminish its effectiveness.

Our take —

The severe lockdown measures put in place in Wuhan and all other cities in Hubei province in China, enacted early in the epidemic, slowed the exportation of COVID-19 from mainland China, but were insufficient to stop transmission globally. The author’s broad approach to modeling airport screening and contact tracing limit the utility of results, but symptom-based airport screening appears to be ineffective in reducing case exportation. These results are unlikely to be applicable after the very beginning of an epidemic, and other countries may not have the ability to implement city-wide lockdowns.

Study design

Modeling/Simulation

Study population and setting

China imposed complete lockdown on Wuhan city (January 23, 2020) and in 15 cities throughout Hubei province (January 24, 2020). The authors used daily case counts of COVID-19 in mainland China from December 8, 2019 to February 15, 2020 along with airline network data, to predict the volume of exported COVID-19 cases before and after the lockdowns, and under the counterfactual scenario of no restrictions. Using estimates of the incubation period and duration of the pre-isolation symptomatic period, the potential impacts of general airport screening (symptomatic or questionnaire-based) and contact tracing in China were also estimated from model-derived probabilities of COVID-19 cases traveling.

Summary of Main Findings

Travel lockdowns were estimated to reduce international exportation of COVID-19 cases by 71% (95% CI: 69%, 72%), from 779 to 230 cases. An estimated 82 additional exported cases from mainland China (95% CI: 72 to 95) were contained due to airport screening. Hypothetical quarantine following contact tracing in mainland China was estimated to have only a modest effect (25% reduction) on the probability of travel among infected cases.

Study Strengths

A strength of the study was the inclusion of airline network data in the analysis. Key model parameters (e.g., incubation period) were estimated using the best available data.

Limitations

The non-pharmaceutical interventions of airport screening, contact tracing, and quarantine are described and modeled in very general terms. Airline network data were outdated (from 2014). Authors address case under-reporting in China, but under-reporting of exportation events is also likely and thus affected model calibration.

Value added

The study integrates empirical COVID-19 incidence data from mainland China with airline network data in an attempt to isolate effect estimates for multiple, co-occurring nonpharmaceutical interventions– travel lockdowns, airport screening, and contact investigations.

Our take —

In this important study, country-wide suppression measures in Europe (e.g., lockdowns and school closures) are estimated to have prevented tens of thousands of deaths and to have considerably reduced transmission, though not to levels that would extinguish the epidemic. Results suggest the need for suppression measures to continue for a long time, since susceptible individuals are still estimated to represent a large fraction of the European population.

Updated Review Available

This expert summary is for the non peer-reviewed preprint. We also summarized this paper after it underwent peer-review and was published in Nature on June 8, 2020.  You can find our updated review of the published article here.

Study design

Modeling/Simulation

Study population and setting

A semi-mechanistic Bayesian hierarchical model was fit to observed COVID-19 deaths in 11 European countries. The model estimated the impact of large-scale non-pharmaceutical interventions on the reproductive rate of the virus, with particular attention to whether reproductive number had been driven below 1, a level at which the outbreak would die out, and on the number of infections and deaths in these countries up to March, 31, 2020.

Summary of Main Findings

59,000 deaths (95% credible interval: 21,000 – 120,000) are estimated to have been averted by non-pharmaceutical interventions in 11 European countries up to March 31, and the reproductive number of the virus is estimated to have been reduced from 3.87 to 1.43 (averaged across countries). Many times more people are estimated to have been infected by SARS-CoV-2 than have been confirmed, with estimates of attack rate ranging from 0.41% (0.09%-3.2%) in Norway to 9.8% (3.2%-26%) in Italy. Lockdown was estimated to have the greatest impact, followed by school closure, but no intervention impacts were statistically significantly different from one another.

Study Strengths

The model is fit to observed deaths, which are likely to be more reliable than case counts or hospitalizations. The model reproduces observed data up to March 28th very well. Uncertainty from various sources is appropriately handled by the model explicitly and by the discussion implicitly. Prior distributions and parameter values are chosen based on current, best available data.

Limitations

There is considerable uncertainty over model parameters such as the infection fatality ratio. Interventions are assumed to have the same impact across countries and time; the timing of country-specific interventions makes it difficult to distinguish between the effects of specific interventions. The interval between infection and death means that effects of control measures are not yet as apparent in countries in an earlier phase of the epidemic; relatedly, the model is heavily influenced by countries with a high number of deaths that implemented interventions earlier. The model assumes that If there had been no intervention, the reproductive number would not have changed, which is unlikely as people often change behavior during large outbreaks, even if no formal policies are put into place.

Value added

This study is the most thorough model-based estimate of the impact of European country-wide non-pharmaceutical interventions at the time of review.

Our take —

This study used data from Italy, China, Korea, and the US to project the trajectory of the COVID-19 pandemic and subsequent demands on the US healthcare system. The statistical approach taken limits the model’s ability to capture complex dynamics, including slow decreases in deaths and resurgences. Estimates are continually updated as new data are incorporated; more recent reports using this model suggest healthcare demand will be lower than values projected here.

Study design

Modeling/Simulation

Study population and setting

This paper provides forecasts of the COVID-19 epidemic trajectory in each US State and demands on the healthcare system from March through July 2020. The authors used a statistical approach to estimate the expected daily number of deaths in each state by assuming that the daily death rate would roughly follow a normal distribution. To project the peak mortality level and the day when it would occur, the authors extrapolated from the relationship between mortality and the timing of key non-pharmaceutical interventions (e.g., lockdowns) from Wuhan, China. Assumptions about ICU utilization and age-specific death rates were based on data from Italy, China, Korea, and the US. Projected death rates are then used to estimate hospital service utilization using an individual-level microsimulation model, which attempts to describe how macro-level policies affect individuals.

Summary of Main Findings

Authors estimated the peak of the US epidemic to occur in the second week of April 2020, during which time excess demand from COVID-19 was expected to be: 64,175 (95% uncertainty interval [UI]: 7,977-251,059) total beds; 17,380 (95% UI: 2,432-57,955) ICU beds; 19,481 (95% UI: 9,767-39,674) total ventilators. The projected date of peak excess health system demand varied across states from mid-April through May. In total, the authors estimated that there will be 81,114 (95% UI: 38,242-162,106) COVID-19 deaths from mid-March 2020 to mid-July 2020 in the US, with less than 10 deaths per day expected between May 31 and June 6.

Study Strengths

This model provides estimates across all US states and includes healthcare utilization forecasts, which are critical for planning and decision-making. Results for each state are accessible through an online visualization tool.

Limitations

Uncertainty in model estimates should increase the further out in time a model attempts to make predictions (e.g., months out); however, in this study, uncertainty shrinks, which suggests that there are many uncaptured elements. The authors rely on a purely statistical modeling approach with no epidemiological basis, and the assumed a priori shape of the mortality curve prevents the model from being able to capture complex and diverse disease dynamics experienced across states. The forecasts are largely driven by timing and intensity of social distancing measures, and do not account for adherence to social distancing measures, social norms, or underlying health characteristics of the population beyond age. Authors assume any US state implementing three of four interventions (school closures, closing non-essential services, shelter-in-place, and major travel restrictions) will see an epidemic trajectory similar to that reported in Wuhan, China. This is unlikely considering multiple additional interventions implemented in China, such as mandatory masks in public. Finally, the authors do not present any data on how well the model predicts (e.g., such as through running it using previously observed outcomes).

Value added

This study presents the first set of estimates of predicted health service utilization and deaths due to COVID-19 for each US state, assuming social distancing is maintained throughout the epidemic.

Our take —

Reanalysis of two controversial publications using additional data refutes the hypotheses that SARS-CoV-2 was engineered to contain pieces of HIV-1 and uses snakes as an intermediate host. Available evidence suggests the probable evolutionary origin of SARS-CoV-2 is bats, with the involvement of one or more mammalian intermediate hosts. Additionally, the study finds that comparing host and virus codon usage is not specific enough to determine intermediate hosts of coronaviruses.

Study design

Ecological, Modeling/simulation, Other

Study population and setting

The authors present a reanalysis of data from three recent publications: 1) similarities in genetic sequences within the spike protein of SARS-CoV-2 and human immunodeficiency virus (HIV-1) published by Pradhan et al. (https://www.biorxiv.org/content/10.1101/2020.01.30.927871v2, now withdrawn); 2) identification of potential intermediate hosts of SARS-CoV-2 by comparing the ways that the virus and animals use their genetic code to produce proteins, specifically their relative synonymous codon usage, published by Ji et al. (https://doi.org/10.1002/jmv.25682); and 3) assembly of a draft coronavirus genome from metagenomic reads from Malayan pangolins produced by multiple research groups.

Summary of Main Findings

The authors found that the genetic sequences within the spike protein share no significant similarity with HIV-1 (contradicting Pradhan et al.); rather, all four sequences were close matches to other viruses and three out of four matched exactly with sequences in a coronavirus from a bat. The reanalysis of codon usage between SARS-CoV-2 and potential intermediate hosts was performed using a more complete database than that used by Ji et al. and additional coronaviruses for comparison. The authors find that the most probable intermediate hosts for SARS-CoV-2, SARS-CoV, and MERS-CoV based on codon usage are frogs, which are not known to be involved in any way with the life history of these viruses, thus calling into question the biological validity of relying on codon usage for identifying intermediate hosts. Finally, they successfully put together all of the sequences of pangolin coronaviruses into a draft genome, with 73% coverage and 91% sequence identity (92% for the spike protein) compared to the SARS-CoV-2 genome.

Study Strengths

The authors make a clear and well-supported argument against the claims presented in the Pradhan et al. and Ji et al. studies. The authors reexamine the spike protein sequences with broader search parameters than the original Pradhan et al. study. For the codon usage analysis, a broader diversity of viruses (SARS-CoV and MERS-CoV) and potential intermediate hosts were considered, and the database of codon usage was updated more recently than the one used by Ji et al.

Limitations

Regarding the analysis of pangolin coronavirus metagenomes, the phylogenetic distance of the pangolin coronavirus to SARS-CoV-2 is still too far to implicate pangolins as the intermediate hosts of the virus. More surveying must be done in bats, pangolins, and other mammals to identify the zoonotic source of SARS-CoV-2 in humans.

Value added

This study discredits two controversial hypotheses regarding the origin of SARS-CoV-2 that emerged early in the outbreak and generated significant media attention.

Our take —

This non peer-reviewed study found that social distancing drastically reduced transmission in Wuhan and Shanghai. Modeling suggests that school closures can significantly reduce overall transmission and disease burden, even though susceptibility to infection was lower among school-aged individuals. Consistent with other studies, the model assumed a reproductive number of 2.0–3.5 before interventions were implemented. More research is needed to refine age-specific estimates of susceptibility to infection and disease, including asymptomatic presentations. Consequently, results should be interpreted cautiously.

Updated Review Available

This expert summary is for the non peer-reviewed preprint. We also summarized this paper after it underwent peer-review and was published in Science on April 29, 2020.  You can find our updated review of the published article here.

Study design

Modeling/Simulation

Study population and setting

Authors evaluated the impact of social distancing on human mixing patterns using contact survey data from Wuhan (n=636) and Shanghai, China (n=557) before and during the outbreak, after cases had been identified and social distancing measures had been implemented in both cities. Survey participants in Wuhan were asked to describe their contact behavior on a regular weekday before the outbreak was recognized (late December 2019), and on the day before the interview (early February 2020, during the epidemic). Baseline contact behavior data for participants in Shanghai were obtained from a survey conducted in Shanghai in 2017–18 using similar methods and questions; participants in Shanghai were also asked to describe their contact behavior on the day before the interview (early February 2020, during the outbreak) using the same survey as Wuhan. Authors used contact tracing information from Hunan Provincial Centers for Disease Control and Prevention to calculate the age-specific relative risk of infection of 57 index cases, and estimate differences in susceptibility to infection and clinical disease by age. Using those estimates, authors modeled how SARS-CoV-2 population-level disease dynamics (e.g., transmission) are affected by differences in susceptibility by age and changes to social mixing patterns brought about by social distancing measures.

Summary of Main Findings

Wuhan was placed under total lockdown starting in late January 2020, including restricted travel, mass school closures, social distancing, and the closure of non-essential services and businesses. Although Shanghai implemented similar controls, as of February 17, 2020, it was only under semi-lockdown. This study found that the average daily number of contacts per participant was significantly reduced from 14.6 to 2.0 in Wuhan, and 20.6 to 2.3 in Shanghai. The largest number of contacts were recorded in school settings, and school closures eliminated contacts between school-aged individuals; contacts during the outbreak mostly occurred at home with household members. Individuals aged 0–14 years had a lower risk of infection during close contact with confirmed cases relative to individuals aged ≥65 years (OR=0.41; 95% CI: 0.18–0.93; p-value=0.03). Accounting for differences in susceptibility by age, authors then modeled the impact of various preemptive interventions (i.e., early in the epidemic) and showed that social distancing measures drastically reduced the R0 (reproductive number). Assuming a baseline R0 of 2.0–3.5, modeling suggests that school closures could reduce peak incidence by 64%, but would not completely stop transmission (i.e., reduce R0 to <1).

Study Strengths

Differential infection risks by age are considered when assessing the impact of school closures. Contact mixing in Shanghai was determined using data from a previously conducted survey, so is not subject to recall bias. The transmission model was calibrated against survey data. Authors used 5.1 days for the model’s serial interval (the time between symptom onset in the source and symptom onset in the recipient infections), which is longer than some earlier estimates but consistent with values used in newer studies.

Limitations

Contact mixing patterns are self-reported and may be subject to several biases. For example, contact mixing patterns in Wuhan may be affected by recall bias; recall of up to seven days is generally assumed to provide acceptable accuracy, and participants were asked to recall up to two months. Differences in susceptibility to infection and symptom development by age were estimated using only 57 primary confirmed cases. Uncertainty regarding the susceptibility profile of infections still exists. Modeling results may not be generalizable to other locations where social distancing measures may differ in strategy or magnitude. Model does not include individual behavior modifications that may have occurred simultaneously (e.g., using masks or maintaining physical distance while in contact), which may have resulted in an underestimation of the effects of social distancing.

Value added

This study estimated the independent population-level effects of social distancing measures on population mixing patterns and subsequent transmission. Authors also estimated susceptibility differences by age using age-stratified relative risk, which contrasts with previous work in Shenzhen indicating no difference in susceptibility by age.

Our take —

Mobility out of Wuhan led to strong transmission chains in other parts of China. The strict ban on travel in or out of Wuhan reduced transmission in the early stage of the outbreak, when importation was the dominant source of new cases. Findings confirm existing knowledge and will not be helpful in settings where the epidemic is well beyond the early stages, or where such severe travel restrictions are impractical.

Study design

Ecological; Modeling/Simulation

Study population and setting

The Chinese government issued a strict ban on travel into or out of Wuhan on January 23, 2020 (the “cordon sanitaire”). Other travel restrictions and suppression measures, not enumerated here, were implemented in close succession. To assess the effect of the Wuhan travel ban on human movements and COVID-19 spread, the authors built a line list dataset of COVID-19 cases in China from December 1, 2019 to February 10, 2020, including information on travel history and demographic characteristics, by extracting individual-level data from official Chinese reports (provincial, municipal, national health governments). Proprietary indices of aggregated human mobility out of Wuhan, and the proportions going to different provinces, were extracted from the Baidu Qianxi web platform.

Summary of Main Findings

The mean incubation period was estimated to be 5.1 days (SD=3.0). The early case count (before February 10, 2020) was tightly associated with the volume of human movement out of Wuhan (R2=0.89). Cases exported from Wuhan before implementation of the cordon sanitaire measures appear to have contributed to initiating local chains of transmission in other provinces. Correlation between daily case counts and human mobility from Wuhan decreased once the cordon sanitaire was put in place, suggesting the increased relative importance of local transmission, and thus of local mitigation strategies.

Study Strengths

To validate the individual-level extracted data, authors conducted sensitivity analyses using reported case counts from the World Health Organization. Estimates of doubling time and incubation period are similar to those found in other studies.

Limitations

The multiple nearly simultaneous interventions make it difficult to single out the effects of the cordon sanitaire. Case definitions changed during the course of the epidemic in China. The human movement data collected from the Baidu Qianxi web platform do not represent the exact number of individual travelers; instead, they represent an index calculated through Baidu’s proprietary methods. Symptom onset date was only available for 667 cases; onset dates for the rest (n=31,436) were estimated using a linear regression model. China’s ability to implement draconian restriction measures may not be feasible in other settings. Findings may only be helpful for countries at early stages of the epidemic.

Value added

The conclusions are not novel, but the methods are strong and convincing.

Our take —

Using available genomes for SARS-CoV-2 and related coronaviruses from humans, bats, and pangolin, the study assesses the unique features of the novel virus and concludes that the features are likely of natural origin and not the result of laboratory manipulation. Whether these features evolved within an animal reservoir or in humans following initial spillover cannot be decided on genomes alone and will need confirmation from additional sampling.

Study design

Ecological, Modeling/Simulation, Other

Study population and setting

Strictly an analysis of existing SARS-related coronavirus genomes: SARS-CoV-2 genome from humans, RaTG13 SARS-related CoV from a bat, SARS-related CoV from a pangolin, SARS-CoV from humans, and two other SARS-related CoVs from bats.

Summary of Main Findings

The genome of SARS-CoV-2 has two unique features that distinguish it from other SARS-related coronaviruses: 1) a set of key changes in the amino acids within the region of the spike protein that the virus uses to bind and enter human cells and 2) a unique insertion of amino acids between the two subunits of the spike protein. The presence of similar spike protein changes in bat and pangolin coronaviruses and analogous amino acid insertions in MERS coronavirus and avian influenza viruses suggest that SARS-CoV-2 is of natural origin.

Study Strengths

Frames the hypotheses regarding the origin of the SARS-CoV-2 and evaluates the parsimony of each hypothesis against the available evidence.

Limitations

There is no way to conclusively discern based on genomic data alone whether unique features of SARS-CoV-2 occurred in animal reservoirs prior to spillover or during an initial human-to-human transmission phase prior to the recognized outbreak of cases. Additional genetic sequencing of coronaviruses in wildlife and serological surveys of SARS-CoV-2 in human populations will be necessary.

Value added

The paper dismisses speculation that SARS-CoV-2 was an engineered virus. The paper also provides testable hypotheses about whether the genomic features were acquired within animal reservoirs prior to spillover into humans or within humans during initial human-to-human transmission of a low-pathogenic zoonotic infection prior to the recognition of the outbreak.

Our take —

This modeling study used data from China, Italy, and the UK to determine the potential impact of non-pharmaceutical interventions on the epidemic in the US and UK. Population-wide social distancing was found to be the most impactful intervention, and models suggest that full suppression of disease transmission may be the only way to avoid exceeding healthcare facility surge capacity until another broad-scale, effective intervention like vaccination can be put in place. Uncertainties around model parameters, such as the efficiency of disease transmission, should be considered and results should be interpreted cautiously when applied to other settings.

Study design

Modeling/Simulation

Study population and setting

The study deployed an existing pandemic influenza planning model in Great Britain to explore the potential impact of five non-pharmaceutical interventions (NPIs) implemented in Great Britain and the United States. This model used Census data such as the population density and age distribution, household structure, workplace and school size, and commuting data, to create a synthetic, yet realistic, population similar to that of Great Britain. Authors modeled disease-causing contact patterns by relying on previously-collected surveys of social mixing. The model evaluated the following NPIs: case isolation in the home, voluntary home quarantine, social distancing of the elderly, social distancing of the entire population, and school and university closure.

Summary of Main Findings

Population-wide social distancing would have the largest impact; in combination with other interventions – notably home isolation of cases and school and university closure. NPI strategies enduring for three months, which the authors termed “mitigation strategies,” were unlikely to enable the US and UK healthcare systems to remain below emergency surge capacity limits. Population-wide social distancing, when applied for extended periods of time (e.g. five months), suppressed disease transmission to a level where R (the reproductive number) fell below 1 (on average, a single case infects less than one other individual).

Study Strengths

Incorporates information from China, as well as new (at the time) data from Italy and the UK. Study applies these data to specific population and geographical information, allowing the trajectory of the epidemic to be modeled for both the UK and US.

Limitations

These models provide useful insights on the relative impact of NPIs, but due to large uncertainties in disease features like the efficiency of virus transmission and the proportion of asymptomatic infection and behavioral compliance of NPIs over extended periods of time, it is difficult to evaluate the accuracy of the absolute magnitude of health impact from these models. Ethical and economic implications of these strategies were not considered, which further breeds uncertainty in behaviors that may affect disease transmission. Results may not be applicable to low- and middle-income countries.

Value added

This is one of the first studies to estimate the impact of NPIs on the course of the epidemic in high-income countries. Compares strategies that aim to suppress disease transmission to those that aim to slow disease spread, thus mitigating the impact on surge capacity. Considers feasibility and the health impact when triggering NPIs with epidemic outcomes (e.g., using weekly ICU cases to turn an NPI on or off in the model) for multiple NPI measures.