Skip to main content

Our take —

While the puzzle is far from solved, this study, available as a preprint and thus not yet peer reviewed, provides evidence that the novel SARS-CoV-2 variant of concern 202012/01is associated with substantial increased transmissibility. No evidence was found to suggest that the new variant was associated with increased hospitalization or death rates. The study also reported that additional stricter mitigation measures along with high rates of vaccination will be necessary to control transmission due to the estimated increased transmissibility of the new variant and its increasing frequency in the population. Testing and sampling biases may partially explain the observed results, and founder and super spreading events cannot be fully ruled out as explanations for increasing frequency and transmissibility of the new variant.

Study design

Ecological, Modeling/Stimulation

Study population and setting

This study reports on the frequency and transmissibility of SARS-CoV-2 variant of concern (VOC) 202012/01 in England between September 28 and December 1, 2020. The frequency and proportion of the VOC in England was estimated from routine diagnostic PCR data. Diagnostic assays fail to detect the spike glycoprotein (S-gene) in the VOC, and so samples exhibiting S-gene target failure (i.e. S gene drop-out) that are positive for other SARS-CoV-2 genes, can be used as a proxy for VOC in the absence of whole-genome sequencing, and was shown to be 97% specific for the VOC in the UK during the analysis period. After estimating the proportion of all positive samples with S gene drop-out, the authors used a two-strain age and regionally structured transmission model fit to COVID-19 hospitalization and death data from seven regions to explain observed increases in COVID-19 diagnoses and frequency of the VOC during the observation period. Four hypotheses to explain the coincident rise in overall cases and the VOC were evaluated: 1) increased infectiousness of the VOC; 2) immune escape (i.e., reinfection of individuals who had been previously infected with a different variant by the VOC); 3) increased susceptibility to the VOC among children as compared to other variants, and 4) a shorter generation time. The model was also used to assess whether there were differences in odds of hospitalization or relative risk of death due to the VOC. The authors also assessed whether changes in contact patterns over time might explain observed changes in transmission using Google mobility data and age-specific contact data from the CoMix social contact survey. Lastly, the authors projected the course of the epidemic under various control strategies, including lockdowns, school closures, and mass vaccination, through the summer.

Summary of Main Findings

The authors found that the proportion of COVID-19 cases in England due to the VOC has rapidly increased since November. The increasing proportion of VOC cases was also associated with a significant increase in the effective reproductive number, Rt. Of the four hypotheses tested to explain the coincident rise in the VOC and Rt, the observed data were most consistent with the model including increased transmissibility of the VOC relative to pre-existing variants. The authors estimated that the VOC was 56% more transmissible (95% credible interval: 50-74%) than previously observed SARS-CoV-2 variants combined. However, there was no evidence that the VOC led to higher rates of hospitalizations or deaths. Changing contact patterns did not appear to explain the rise in the VOC. Lastly, the authors found that unless schools are closed and vaccinations increased to 2 million per week, the effective reproductive number is unlikely to fall below 1 as a consequence of increased frequency and transmissibility of the VOC.

Study Strengths

This study assessed multiple alternative mechanisms for the rapid rise in COVID-19 cases along with increasing frequency of the VOC in England, including increased transmissibility of the VOC, increased, susceptibility to the VOC among children (which has major implications for school re-openings/closures), and changes in social contact patterns.  Implications of the VOC on vaccination strategies were also explored.

Limitations

Limited information was reported on testing rates or genomic sampling across NHS regions or age-groups over the observation period which may impact interpretation of results. While unlikely, this study does not rule out the possibility that founder and super spreading events with subsequent spread of the VOC to other regions explain the observed data. Of note, there were apparent increases in social contacts among persons <18 years of age immediately prior to the rise in the VOC, which was not addressed in the limitations, despite reported outbreaks among schools over the same time frame. Further, estimates of Rt from contact data appear to fit the data relatively well, except for a brief time period between/during October and November.

Value added

This is the first study to estimate the relative transmissibility of the novel SARS-CoV-2 variant, VOC 202102/01, compared to pre-existing variants and its implications for future transmission dynamics and control.

Our take —

This study, available as a preprint and thus not yet peer reviewed, reviewed global sequence data (as of November 26, 2020) containing a two amino acid deletion in the SARS-CoV-2 spike protein, which is involved in viral entry into human cells. Authors described the primary lineages that contain this variant and the existing evidence that these lineages may be more infectious, including antibody experiments and the increasing frequency of sequences with these variants being reported in parts of Europe where viral sequencing is common. In vitro infectivity assays conducted as part of this study further corroborated their conclusion that the deletion may increase the ability of the virus to enter human cells, but additional experiments with actual SARS-CoV-2 virus lineages are needed. This study provides evidence suggesting that SARS-CoV-2 lineages with the H69/V70 deletion may be important for public health because they may be more transmissible, but the available evidence is limited by sampling bias and lack of epidemiological data. Additional sequencing and epidemiological analysis from around the globe are needed to determine the true frequency of B.1.1.7 and other deletion-containing variants, as well as the effect of these variants on COVID-19 transmission.

Study design

Ecological

Study population and setting

This study analyzed all publicly available SARS-CoV-2 sequence data as of November 26, 2020. After removing duplicate and low-quality sequences, a total of 194,265 SARS-CoV-2 sequences from across the globe were analyzed in this study. The aim of this study was to better understand the H69/V70 deletion in the spike protein, including the temporal and geographical distribution of sequences containing this deletion and what other spike protein mutations are observed alongside it. Of particular note are sequences belonging to the widely discussed B.1.1.7 lineage detected in the United Kingdom (which contains this deletion, alongside a number of other spike protein mutations).

Summary of Main Findings

After examining the repeated emergence of the H69/V70 deletion in the SARS-CoV-2 genome, both independently and alongside several other spike protein variants, the authors conclude that this deletion is a fitness-enhancing change that may stabilize other spike protein changes. Specifically, they found that the H69/V70 deletion occurred in many different global lineages (first observed April 2020), suggesting several independent acquisitions. They also note that the proportion of sequences with the N439K mutation (which is in the receptor binding domain of the spike protein) greatly increased after it was observed in combination with the H69/V70 deletion, and that this deletion was also observed in combination with Y453F and other mutations (the lineage associated with mink infections that have been shown to have reduced susceptibility to sera from recovered COVID-19 patients). The authors then address the much-noted B.1.1.7 lineage, in which the H69/V70 deletion occurs alongside N501Y and several other mutations. This lineage was first detected in the UK in September 2020 and is of concern because of its notable divergence from other lineages (which could suggest its evolution in a chronically infected host or emergence in a location with limited sequencing) and association with an increasingly high number of infections. Finally, the authors infected human cell lines with a pseudotyped virus containing the SARS-CoV-2 spike protein with and without the H69/V70 deletion. In these lab experiments, they found that virus with the deletion in the spike protein was better able to enter human cells in culture than virus without the deletion.

Study Strengths

This study analyzed data published through the end of November 2020, which allows the authors to comment on distribution of SARS-CoV-2 around the globe at the time of publication. The authors provided a comprehensive summary of all published genomes with the deletion in question, and described the primary lineages it is observed in.

Limitations

Geographic distribution of published genomes is heavily biased, making it difficult to assess exactly when and where particular lineages were circulating. For example, the B.1.1.7 has only been observed in European countries with extensive sequencing capacity, and the lack of sequencing data from other parts of the world means it is impossible to determine if this lineage is circulating elsewhere. Additionally, the authors did not discuss sampling and epidemiological factors that could contribute to the observed increase in frequency of the observed mutations, i.e., if the increase in frequency of the noted lineages is due to an increase in cases or sequencing in specific regions that already have the variant. Finally, the authors performed lab experiments to test infectivity of SARS-CoV-2 with and without the H69/V70 deletion, but the experiments used only psuedotyped virus, not complete SARS-CoV-2 (due to isolates not being available at the time). They also only used the spike protein with and without the H69/V70 deletion and not any of the other mutations commonly observed alongside the deletion, which limits the conclusions that can be drawn from this experiment.

Value added

This study describes all of the primary global SARS-CoV-2 lineages that contain the H69/V70 deletion in the spike protein and discusses the possible geographic and temporal origins of each. The infectivity assays performed suggest that the H69/V70 deletion increases the ability of the virus to infect human cells in vitro.

Our take —

The city of Jena was the first in Germany to mandate mask use in all public spaces, and after 20 days, the number of new cases of COVID-19 per day had declined considerably. This analysis models a counterfactual scenario (using data from other German regions) in which Jena did not mandate masks, concluding that the mandate was effective in reducing new cases. The authors estimated similar but smaller effects in other German regions with mask mandates. This study did not directly measure the effectiveness of wearing a mask in preventing SARS-CoV-2 transmission, and unmeasured factors may have contributed to the decline in new cases, but the study adds to the evidence base suggesting that policies requiring mask use may be an important part of controlling transmission.

Study design

Ecological, Other

Study population and setting

This study was carried out in Germany and examined the effectiveness of compulsory public face mask requirements to reduce SARS-CoV-2 transmission. The authors first compared the experience of the city of Jena, which mandated mask use by the public on April 6, 2020, to a counterfactual scenario in which Jena did not implement a mask mandate. To construct the counterfactual, the authors employed a synthetic control approach, in which data from 401 regions in Germany were used to construct a “donor pool” of regions that were weighted to most closely estimate the pre-mandate cumulative case count in Jena and other regional characteristics. Observed cumulative cases after the mask mandate were compared to the modeled counterfactual. Second, the authors used the same approach for all regions in Germany that mandated mask use by April 22, 2020 (n=32). The authors also performed various sensitivity analyses and checks on how well the model assumptions were met.

Summary of Main Findings

There were 16 newly reported cases of COVID-19 in Jena between April 6 and April 26, 2020; in the counterfactual scenario of no mandate in Jena (the synthetic control), there were 62 new cases; this represents an estimated 74% reduction in new cases over the 20-day span. Placebo-in-space tests (estimating “effects” in other locations that did not actually have a mask mandate, and comparing the estimated effects in Jena to the distribution of these “placebo effects”) indicated statistically significant differences in Jena (p<0.10) beginning 13 days after the mask mandate. This indicates the results from Jena were unlikely to be due solely to chance. The lag was argued to be concordant with the incubation period of COVID-19, plus delays in testing and reporting. When considering all regions with mask mandates, the authors estimated that an average of 28 cases per region were prevented over a 20-day period after the intervention, corresponding to a 51% reduction in new cases during the post-intervention period.

Study Strengths

The authors performed several analyses to check the robustness of their findings to alternative assumptions, considered alternative explanations for the observed reduction in new cases during the study period (including anticipation of the mandate), and used an SIR model to estimate the lag period after which any effects of a mask mandate might be observed.

Limitations

Dates of mask mandates may be poor proxies for individual mask-wearing behavior. The study estimates the effectiveness of a mask mandate, not of mask wearing, and this difference was not always emphasized. There may be geographic heterogeneity in personal behavior that was associated with the timing of mandates. The authors considered cumulative case counts, demographic characteristics, and health care system characteristics as covariates, but did not consider any other non-pharmaceutical interventions that may have varied by region. The authors ruled out other interventions as an explanation for the results because other interventions in Jena were at least 10 days away from the mask mandate, but there may have been unmeasured confounding by other factors, including individual behavior change and/or epidemiologic characteristics of SARS-CoV-2 spread in Jena. There was little discussion of the makeup of the donor pool for the primary analysis, though the authors did perform some sensitivity analyses. Finally, results may not be generalizable to mask mandates during other time periods or locations, which may have different public responses to a mask mandate and different levels of general community transmission.

Value added

This study provides evidence of the effectiveness of mask mandates in reducing transmission of SARS-CoV-2 at the regional level in Germany.

Our take —

Following the introduction of a state-wide mask mandate on July 3, 2020, Kansas counties adopting the mandate witnessed a 6% reduction in COVID-19 incidence by August 23, while incidence doubled over the same period in counties opting out of the mask mandate. The introduction of a mask mandate alone is unlikely to account for these incidence differences, as counties with mask mandates also implemented complementary mitigation strategies to control COVID-19 transmission.

Study design

Ecological

Study population and setting

Using seven-day moving averages of COVID-19 incidence (number of cases per 100,000 persons) across 105 Kansas counties between June 1 and August 23, 2020, trends in COVID-19 incidence were compared in counties adopting a state-implemented mask mandate on July 3, 2020, to counties that opted out of the state mandate. A difference-in-differences analysis was used to estimate the impact of the mask mandates, comparing the rates of new COVID-19 cases before and after July 3, 2020 in counties implementing a mask mandate (n = 24) relative to those without such mandates (n = 81).

Summary of Main Findings

Following the implementation of mask mandates, the seven-day moving average of COVID-19 incidence decreased by 6% in counties adopting mask mandates but doubled (100% increase) in counties opting out of these mask mandates. Approximately two-thirds (67%) of the state’s population resided in counties implementing mask mandates, which were dispersed throughout Kansas but highly clustered.

Study Strengths

The investigators used natural experiment methods (i.e., difference-in-difference estimation) to compare rates of COVID-19 before and after implementation of mask mandates across Kansas’ 105 counties. These mandates were put in place by the state during a time in which few other state-wide policies and factors were influencing the spread of COVID-19 differently between opt-in and opt-out counties. This makes it more plausible, but not certain, that a substantial proportion of the differences in COVID-19 cases between mask mandate opt-in vs opt-out counties were driven by the mask mandates and subsequently mask-wearing behavior. The authors also attempted to account for differences between city-mandated mask orders, some of which were nested in counties opting out of the state’s mask mandate, by reanalyzing their data, which produced similar (but smaller) effect size estimates compared to their published analysis.

Limitations

Because the authors compare COVID-19 incidence in counties with and without mask mandates, county-level differences in mask-wearing behaviors, mask ordinance implementation, and ordinance enforcement limit attribution of incidence reductions to policy implementation alone. Counties introducing a mask mandate would also be more likely to impose other restrictions (e.g., restaurant/business occupancy restrictions, suspension of mass gatherings) that would likely affect COVID-19 transmission dynamics. In fact, over half (54%) of counties with a mask mandate introduced at least one other COVID-19 mitigation strategy, while fewer than 10% of counties without a mask mandate implemented such control measures. In addition, other unmeasured factors, like behavior change in response to rising COVID-19 incidence, could have different impacts on COVID-19 incidence in opt-in versus opt-out counties. This limits the degree to which COVID-19 incidence differences can be attributed to the effectiveness of mask mandates.

Value added

This study is among the largest studies in the US to compare COVID-19 incidence in counties implementing a mask mandate at a single point in time to those opting out of a mask mandate, and is implemented in a scenario that makes assumptions more plausible than many other mask mandate policy studies.

Our take —

This study used four different analytic approaches to estimate the relative effectiveness of country- and US state-level non-pharmaceutical interventions (NPIs) early in the pandemic. The authors found that social distancing (e.g., cancelling small gatherings), school closures, and travel restrictions (e.g., border closures) were most effective, while environmental measures (e.g., disinfecting surfaces) were least effective. Though each statistical approach was distinct, all relied on a similar set of assumptions, and all were subject to biases. Assessing the effectiveness of NPIs is difficult for many reasons: many were implemented simultaneously, case reporting varies across regions and over time, and the success of a given NPI may heavily depend upon local context. This study is not definitive, but suggestive; it highlights important issues involved with NPI impact evaluation.

Study design

Ecological; Modeling/Simulation

Study population and setting

This study considered 6,068 non-pharmaceutical interventions (NPIs) in 79 territories (countries and US states) implemented in March and April of 2020. The authors ranked the effectiveness of these NPIs in reducing the effective reproduction number of SARS-CoV-2 (Rt) using eight broad themes, and 46 categories of NPIs that were implemented 5 or more times within those themes, established by the Complexity Science Hub COVID-19 Control Strategies List (CCCSL). Four analytic methods were used to rank effectiveness: 1) a case-control study (similar to a difference-in-differences approach), 2) LASSO time-series regression (an approach designed to prevent overfitting of models), 3) random forest regression (a a prediction and classification method) with NPIs ranked by measuring model performance with and without each NPI, and 4) Transformers, a machine learning technique for parallel processing of sequential data that can be applied to time series. All approaches used a range of lag periods (i.e., the timing between implementation of an intervention and its effect on the reproductive number, Rt), and considered possible confounding variables including the duration of the local epidemic, Rt at the time of a given NPI, population size, population density, and the previous number of NPIs implemented (total and within the same category as the given NPI). The results were applied to two external validation data sets.

Summary of Main Findings

There was general agreement among the four methods used to rank NPI effectiveness; of the eight broad NPI themes, social distancing and travel restrictions were the highest ranked, while environmental measures such as disinfecting surfaces were least effective. Six of the 46 categories showed statistically significant reductions in the reproductive number (Rt) with all four methods: canceling small gatherings (change in Rt: -0.22 to -0.35), closing educational institutions (-0.15 to -0.21), border restrictions (-0.06 to -0.23), increasing availability of personal protective equipment (-0.06 to -0.13), individual movement restrictions (-0.08 to -0.13), and national lockdown including U.S. state stay-at-home orders (-0.01 to -0.14). Environmental cleaning, public transport restrictions, and contact tracing measures were among the least effective NPIs across estimating methods. NPI effectiveness was highly heterogeneous across countries. Applying these methods to two external datasets produced broadly similar results. By artificially shifting NPIs to different times in relation to the age of the epidemic, the authors estimated that early adoption was beneficial for lockdown, small gathering cancellations, travel restrictions, and closure of educational institutions.

Study Strengths

This study considered a wide range of disaggregated NPIs, and compared their effectiveness using four distinct methods to examine the sensitivity of NPI ranking to the estimation method. Methods were tested on two external datasets.

Limitations

While the authors employed four different methods for estimating NPI effectiveness, the specification of models in each method was structurally similar, meaning that agreement across methods is less remarkable than it may appear. The estimation of Rt relies on reported cases; case reporting was highly heterogeneous across regions and over time during the study period, and was subject to discrete changes in case definitions and reporting standards. Estimates of changes in Rt may therefore be biased. Also, using this method makes it appear as if measures that increase case ascertainment (e.g., improving testing capacity, contact tracing, etc.) actually increase Rt. The timing of NPI implementation makes identification of impacts difficult, since so many were implemented roughly simultaneously– this issue is exacerbated because the expected lags between implementation and effects are not precisely known. The study is restricted to March and April of 2020, and so territories that were hit earlier by the pandemic are likely overrepresented. Although travel restrictions are ranked highly, they are likely to be effective only during the initial phases of an epidemic when local transmission represents a small share of new cases. In general, the interactions between the age of a local epidemic and NPI effectiveness are not handled well by the approaches here. Finally, although the authors adjusted for several territory-level variables, including the number of previous interventions implemented, residual confounding (e.g., country GDP) and unmeasured interactions (e.g., the success of contact tracing may depend on multiple other factors including public health infrastructure and testing capacity) are likely.

Value added

This is one of the largest attempts to quantify and rank the effectiveness of non-pharmaceutical interventions across multiple regions.  

Our take —

Following the implementation of multiple mitigation measures, specifically, the introduction of a statewide mask mandate on April 28, COVID-19 incidence, hospitalizations, and deaths per capita declined significantly through June 2020 in Delaware, US. The independent contribution of each of these mitigation measures to observed reductions in COVID-19 transmission and mortality, however, could not be determined.

Study design

Ecological

Study population and setting

Investigators aggregated laboratory-confirmed and probable COVID-19 cases in Delaware from March 11 to June 25, 2020. Four public health mitigation measures were implemented during this timeframe: case investigations (March 11), statewide shelter-in-place ordinance (March 24 to June 1), statewide mask ordinance (April 28), and contact tracing (May 12). Investigators assessed changes in COVID-19 incidence, hospitalizations, and deaths (per 10,000 Delaware residents per week, respectively) over the same period that these mitigation measures were introduced. Contact tracing outcomes were also described for all laboratory-confirmed COVID-19 cases.

Summary of Main Findings

A total of 9,762 COVID-19 cases were identified between March 11 and June 25, 2020. COVID-19 cases, hospitalizations, and deaths per capita increased throughout March, peaking in mid-April. Following the introduction of a statewide mask mandate on April 28, COVID-19 incidence, hospitalizations, and deaths declined at a rate of 82%, 88%, and 100%, respectively, through June. Among laboratory-confirmed COVID-19 cases, 67% were successfully contacted by health department staff and provided isolation instructions, of whom 83% either refused to name any close contacts or could not recall contacts. Among cases who named close contacts (mean = 2.5 close contacts), contact information was obtained from 66% of the 2,834 contacts identified, 47% of whom were successfully contacted and provided quarantine instructions. The median time from symptom onset in the case to contact elicitation was 8 days and from contact elicitation to contact interview was 2 days.

Study Strengths

Investigators assessed trends in COVID-19 incidence as well as hospitalizations and deaths throughout a period when multiple public health mitigation measures were enacted in Delaware.

Limitations

Given the absence of suitable comparison groups in analysis, observed reductions in COVID-19 incidence, hospitalizations, and deaths cannot be exclusively attributed to the implementation of public health control measures. Furthermore, this study did not include any individual-level behavioral measures on compliance or adherence to public health mitigation effort.

Value added

This is among the first studies to assess the potential effect of multiple mitigation measures on health indicators across the COVID-19 illness continuum (infection, hospitalization, and death).

Our take —

This ecological study used smartphone data to assess reasons for comparatively lower levels of physical distancing previously observed in lower versus higher income census block groups in the United States. The authors demonstrate that work outside the home was the most likely explanation for disparities in physical distancing between income groups.

Study design

Ecological

Study population and setting

This study used aggregated, anonymized smartphone data from January 6 to May 3, 2020 to test the hypothesis that differences in physical distancing previously observed between higher and lower income neighborhoods in the US were explained by work-related demands outside the household. The authors analyzed daily mobility data from approximately 19 million smartphones in all 50 states. The primary outcome was the proportion of smartphone users who spent all day at home, where home was defined as the place where a user spent the most nights over the last 6 weeks. A secondary outcome was the proportion of users who went to work on a given day. Leaving the home for work was defined as stopping at a single location for three or more hours between the hours of 8 am to 6 pm or going to four or more locations for less than 20 minutes each (e.g., delivery work). Travels to points of interest outside of work were also analyzed as a secondary outcome, including visits to parks, playgrounds, carryout restaurants, places of worship, convenience stores, and liquor stores. The primary exposure was median household income in census block groups (CBGs) categorized into quintiles. The impacts of state-level physical distancing policies (stay at home orders and non-essential business closures) on the primary and secondary outcomes by income quintile were analyzed using difference-in-difference linear regression models with fixed effects for state and calendar date.

Summary of Main Findings

There were a total of 210,119 census block groups and points of interest included in the analysis. Increases in physical distancing were observed over the analysis period for all income groups, although increases were more pronounced in higher income neighborhoods. The proportion of days spent at home increased by 11 percentage points in the lowest income group, while it increased by 27 percentage points in the highest income group (p<0.0001; 95%CI: 16.0-16.1). Changes in work-related mobility appeared to explain much of this difference: the highest income group went from working the most days outside the home to the fewest after physical distancing policies were implemented. Visits to all POIs declined in all income groups. Visits to places of worship were the only point of interest with a clear income gradient associated with a decline in visits, with fewer visits in higher income groups.

Study Strengths

This study used a large, well-characterized mobile phone dataset from diverse geographic locations in the United States. Outcomes and exposures were clearly defined, and analyses were conducted using standard, easily interpretable analytic techniques.

Limitations

The number of smartphones per census block group was strongly associated with the median household income level, with fewer smartphone users in lower income census blocks. This may have biased results, particularly if those with smartphones in lower income census blocks were substantially more likely to work outside the home than those without smart phones. Younger smartphone users may have had changes in education-related mobility that were counted as work-related; if those from wealthier areas stayed home from school more frequently, the relationship between income and work-related mobility would be overestimated. Working outside the home was crudely defined as time spent outside the household during daytime hours, and did not include individuals who worked outside the home after 6 pm and before 8 am. Smartphone locations were measured at irregular intervals; aggregated mobility data could systematically misrepresent work-related or home-related behavior.

Value added

This study is the first to assess whether disparities in physical distancing observed between higher and lower income neighborhoods is explained by work outside the home.

Our take —

In San Francisco, California, only half of reachable COVID-19 index cases between April 13 and June 5, 2020, reported >1 close contacts, 84% of whom were notified of their exposure. Of notified contacts, 45% were tested for SARS-CoV-2 infection; 26% of those tested were positive. These results, and the average 6-day delay between case symptom onset and contact notification, provide reasons for concern about the effectiveness of contact tracing for transmission control in San Francisco, at least during the study period. Because contact tracing outcomes were estimated during a municipal shelter-in-place ordinance, these outcomes may not reflect performance of a contact tracing program after restrictions were lifted.

Study design

Ecological

Study population and setting

The authors aggregated all cases of confirmed COVID-19 in San Francisco, California, during the shelter-in-place order from April 13 to June 5, 2020. Close contacts (both household and non-household) were identified from index cases and from May 5 onwards and were referred for COVID-19 testing irrespective of symptoms. The authors quantified outcomes along the contact tracing continuum using descriptive statistics.

Study Strengths

The authors enumerated specific outcomes in the contact tracing pipeline (e.g., number of contacts identified, number of close contacts notified/tested, median time to contact notification) to assess the effectiveness of contact tracing strategies in San Francisco. Additionally, the authors disaggregated contact tracing outcomes by contact status (household versus non-household contacts) and compared summary statistics across racial/ethnic identities.

Limitations

The authors did not describe in detail procedures for reaching close contacts, including communication modalities used or number of contact attempts before close contact was deemed lost-to-follow-up. Because close contacts were identified during a shelter-in-place ordinance, index cases may have been reluctant to report non-household contacts to the health department; the number of non-household contacts was, therefore, likely underestimated. Data collection during a shelter-in-place ordinance may also not be reflective of health system performance or capacity once these restrictions are lifted. Lastly, because the authors did not enumerate any health system performance metrics (e.g., number of contact tracing staff), presented contact investigation outcomes cannot be fully attributed to health system performance or public health workforce capacity.

Value added

This is among the first studies in the United States to describe outcomes related to COVID-19 contact tracing efforts during the early months of the U.S. epidemic in a large metropolitan area.

Our take —

The US public response to COVID-19 is widely perceived to be a partisan issue. This study examined changes in mobility from March to May 2020, and found that political partisanship, as measured by county level vote margins in the 2016 presidential election, was associated with physical distancing behavior at the county level. Republican counties reduced mobility and visits to non-essential services less than Democratic counties, and this partisan difference increased as the pandemic progressed. Multiple difficulties in measuring partisanship, physical distancing behavior, and other important variables mean that the results should be interpreted cautiously.

Study design

Ecological

Study population and setting

The authors analyzed daily GPS location data from approximately 15 million smartphones to estimate differences in physical distancing (using general movement and visits to non-essential services) at the county level among 3,025 counties in the United States from March 9 to May 29, 2020, compared to the period before March 9, 2020. Associations among partisan identity, physical distancing, and COVID-19 infection and death rates were investigated. Partisanship was measured by county-level vote margins between the Democratic candidate (Hillary Clinton) and the Republican candidate (Donald Trump) in the 2016 presidential election. The authors used mixed effects models to estimate the association between partisanship and physical distancing. Viewership of Republican-leaning and Democratic-leaning media was measured with data from SimplyAnalytics, and tested for associations with physical distancing. Mediation models were utilized to test associations between partisanship-related physical distancing and infection and fatality growth rates.

Summary of Main Findings

Average reductions of 21% in general movement and 31% in visits to non-essential services were observed at the county level. Trump-leaning counties exhibited an average 24% decrease in physical distancing metrics, while Clinton-leaning counties exhibited an average 38% decrease during the same period. From mixed effect models, there was an estimated 0.11% reduction in physical distancing for each percentage increase in the vote margin favoring Trump. Physical distancing was highest from March to mid-April and declined afterwards in both Trump-leaning and Clinton-leaning counties. Differences in physical distancing by partisanship widened as the pandemic progressed. The association between partisanship and physical distancing remained in subsequent analyses when counties were collapsed into dichotomous partisan categories and when adjusted for a number of covariates including race, age, income, cases per capita, commute time, type of employment and state-specific policies among others. Viewership of Republican-leaning media (e.g., Fox News) was associated with reduction in physical distancing over viewership of Democratic-leaning media (e.g., MSNBC); this association strengthened as the pandemic progressed and was observed even when accounting for 2016 voting partisanship. Mediation analyses provided some support for the hypothesis that partisan differences in physical distancing affected infection and fatality growth rates.

Study Strengths

The authors considered a wide range of possible confounding variables at the county level in analyses. Viewership of partisan media was considered separately from the primary measure of partisanship.

Limitations

Aggregated mobility data are not perfect proxies for physical distancing behaviors. Data at the county level may mask individual-level variation; the results do not necessarily imply that individual Republicans are less likely to comply with stay-at-home orders and physical distancing compared to their Democratic counterparts. County-level variables included as possible confounders in models were crude and likely did not capture all salient differences between Republican and Democratic counties. Models only included state-level social distancing mandates; regional and municipal regulations and recommendations may have influenced behavior. Compliance with other public health measures such as mask-wearing or staying 6 feet apart were not taken into consideration. Adjusted variables such as employment did not necessarily account for the variability in physical distancing within different employment settings. Differences in counties may exist such as the need to travel further to access essential services or goods. Finally, the construct of partisanship was not well defined; other characteristics of county residents (e.g., levels of distrust for authority, social alienation) might influence both the partisanship measure and physical distancing behavior.

Value added

This is one of the most comprehensive analyses to date of the relationship between partisanship and physical distancing in the United States.

Our take —

This study provides reasonably strong evidence that Trump 2020 campaign rallies caused a substantial number of COVID-19 cases in the US counties where they took place, although the precise number of cases caused is difficult to estimate. The methods were broadly appropriate to disentangle the effects of the rallies from non-rally-related COVID-19 epidemic trends. However, some results are not reported in sufficient detail for readers to assess how well key assumptions underlying the methods were satisfied by the data. Further, the translation of additional cases to deaths is subject to considerable uncertainty and should be considered cautiously.

Study design

Ecological

Study population and setting

This study considered 21 US counties that hosted rallies in support of Donald Trump’s 2020 presidential campaign. The authors compared the change in incident COVID-19 cases before and after the rallies with a prediction of what would have occurred in those counties had the rally not occurred. The prediction was based on incident COVID-19 cases in a matched set of counties in which there was no rally, adjusted for a set of variables describing important differences between rally and non-rally regions. Separate analyses were conducted for each event, and combined afterwards. In order to explore the degree to which the methods might have affected results, the authors used a variety of procedures, matching variables, and adjustment schemes. These variables included county-level demographics, urbanicity, political preferences, COVID-19 policies, and socioeconomic status. COVID-19 case fatality rates (the proportion of cases who died) were estimated for each county by dividing deaths by incident cases after a lag. The number of additional cases attributable to the rally was multiplied by the estimated case fatality rate to estimate the number of deaths attributable to the rallies.

Summary of Main Findings

The primary analysis (matching counties was based on cases per capita) found that rallies increased the number of cases in rally counties by 332 per 100,000 residents (95% CI: 156 to 508), for a total of 38,697 additional cases and 775 additional deaths. In an alternate model that matched counties based on additional demographic variables (total population, % college-educated, and 2016 Trump vote %), the estimated effect was slightly smaller, with an additional 261 cases per 100,000 residents (95% CI: 62 to 460) and 608 additional deaths. Additional model variants produced broadly similar results.

Study Strengths

The methods, and the in-depth treatment of alternative sets of assumptions and models, are well designed and appropriate for examining the impact of rallies on local cases. The implied difference-in-difference strategy using a matching scheme with adjustment is very likely to thoroughly address differences between rallies, and provides a strong foundation for identifying the impact of the rallies on cases. The authors directly address many of the issues that make this kind of estimation difficult through individualized statistical modeling for each region, careful selection of comparators, and in-depth examination of the effects of alternative assumptions, models, and procedures.

Limitations

The most substantial limitation is the weakness of the method for estimating deaths conditional on additional cases. Estimates of county-specific case fatality rates are highly uncertain, since cases likely to be attributable to the rallies are likely in different kinds of people than would have occurred otherwise, and the case fatality rate may change over time. There is also little detail and visual representation of the county-specific models themselves, so it is difficult to evaluate the validity of key assumptions, particularly regarding the rates of COVID-19 in counties used as controls. Visual representation of the case trends in counties with and without rallies would have enabled evaluation of model assumptions.

Value added

While other studies have explored the impact of mass gatherings or found associations between political alignment and COVID-19 cases, this study provides direct evidence of how a pattern of mass political gatherings directly impacted COVID-19 cases. Further, the design of the study can serve as an example for future research on similar topics and/or using similar methods.