Community Mitigation Strategies, Mobility, and COVID-19 Incidence Across Three Waves in the United States in 2020

Abstract

Background:

Summarizing the impact of community-based mitigation strategies and mobility on COVID-19 infections throughout the pandemic is critical for informing responses and future infectious disease outbreaks. Here, we employed time-series analyses to empirically investigate the relationships between mitigation strategies and mobility on COVID-19 incident cases across US states during the first three waves of infections.

Methods:

We linked data on daily COVID-19 incidence by US state from March to December 2020 with the stringency index, a well-known index capturing the strictness of mitigation strategies, and the trip ratio, which measures the ratio of the number of trips taken per day compared with the same day in 2019. We utilized multilevel models to determine the relative impacts of policy stringency and the trip ratio on COVID-19 cumulative incidence and the effective reproduction number. We stratified analyses by three waves of infections.

Results:

Every five-point increase in the stringency index was associated with 2.89% (95% confidence interval = 1.52, 4.26%) and 5.01% (3.02, 6.95%) reductions in COVID-19 incidence for the first and third waves, respectively. Reducing the number of trips taken by 50% compared with the same time in 2019 was associated with a 16.2% (−0.07, 35.2%) decline in COVID-19 incidence at the state level during the second wave and 19.3% (2.30, 39.0%) during the third wave.

Conclusions:

Mitigation strategies and reductions in mobility are associated with marked health gains through the reduction of COVID-19 infections, but we estimate variable impacts depending on policy stringency and levels of adherence.

INTRODUCTION

The coronavirus disease 2019 (COVID-19) pandemic has been a serious public health emergency in the United States imposing unprecedented challenges on health systems, society, and the economy. In 2020 alone, SARS-CoV-2, the causative agent of COVID-19, caused more than 350 thousand deaths with over 20 million reported incident cases in the United States.¹ In response to the pandemic, US states imposed a myriad of community-based mitigation strategies such as social distancing, mask mandates, work closures, and shelter-in-place policies, with the goal of limiting opportunities for transmission and increasing public health preparedness to contain the pandemic.

Several studies have evaluated the effectiveness of community-based mitigation strategies on COVID-19 incidence, hospitalizations, and deaths. For instance, a modeling study found that early shelter-in-place policies in the pandemic averted 5 million confirmed infections in the United States.² Another study at the global level found that physical distancing interventions were associated with an overall 13% reduction in COVID-19 incidence.³ Other studies have reported similar results for several other mitigation strategies finding that they were effective in reducing the transmission and burden of COVID-19.^4–7 However, many of these studies were conducted early in the pandemic and may not capture the extent to which states and countries lifted restrictions and reinstated them due to the emergence of local outbreaks in their communities.⁸

Other complexities for accurately estimating the impact of mitigation strategies are that individual behavior plays an important role in transmission dynamics.⁹ Recent work has highlighted that adherence to interventions has decreased markedly since the pandemic started¹⁰ indicating that evaluation of the effect of mitigation strategies should consider changes in individual behavior. Mobility has been used as a proxy measure for capturing individuals’ behaviors and levels of social distancing adherence. Some studies have found that reductions in mobility are associated with lower COVID-19 transmission.^11–14 However, the need to reevaluate the mobility-COVID-19 relationship is highlighted in recent work revealing that the strength of association changes over time.^15,16 Analyses stratified by time periods may therefore reveal important trends in the effectiveness of community mitigation strategies.

The dynamic nature of the COVID-19 pandemic in the United States along with changes in social distancing measures and individuals’ behaviors offers an opportunity to extensively examine the impact of community-based mitigation measure levels on COVID-19 infections. In this study, we used time-series analyses to explore the relationships of mitigation measures and mobility on COVID-19 incident cases across US states by different time periods. We first examine the extent to which community mitigation strategies and mobility are associated with both COVID-19 incidence and the effective reproduction number. Then we examine compliance to migration strategies by identifying the association of community mitigation on levels of mobility. An examination of social distancing strategies and individual behavior as they changed greatly throughout the pandemic will help inform responses to the current pandemic and future outbreaks.

METHODS

Data Sources

We collated data on the number of daily reported COVID-19 infections by US states from the COVID-19 Data Repository at John Hopkins University.¹ We used data on the US population in 2019 from the US Census Bureau to compute daily incidence proportions Our observation period is restricted from March 1, 2020, to December 31, 2020, before widespread vaccine distribution.

We used state-level data on daily community mitigation levels from the Oxford COVID-19 Government Response Tracker (OxCGRT) database.¹⁷ The OxCGRT database includes the stringency index (SI), which captures the strictness of closures and containment policies by day. We extracted state-level mobility data on the daily number of trips per day from the COVID-19 Impact Analysis Platform from the Maryland Transportation Institute.^14,18,19 To derive a ratio of the daily number of trips taken compared with the same time the past year (i.e., trip ratio), we divided the current days’ number of trips by the average number of trips in the same days of the identical month in 2019.

Statistical Analyses

We employed multilevel (mixed-effects) models to examine the associations of the SI and mobility with COVID-19 incidence proportion per 100,000 population. Inputs to the regression models included COVID-19 incidence proportions, the SI, and the trip ratio, which were all measured daily at the state-level. The regression models included random intercepts for US states and included a smooth function (fourth-order polynomial) of time since pandemic onset to characterize longitudinal trends. We lagged covariates by 11 days to provide sufficient time for policy changes and mobility to impact the number of new infections, as per a systematic review indicating that the average time from infection to case detection was approximately 11 days (95% CI = 8, 13 days).²⁰ We log-transformed COVID-19 incidence proportion to correct for the right-skewed distribution of the data and fit a linear model to evaluate the relative impact of predictors on incidence. Based on a priori theory-based conceptual model (eFigure 1, http://links.lww.com/EDE/B967), we regressed the SI in a univariate model to examine the total effect of the SI on COVID-19 incidence. Inferences for the trip ratio were drawn from a model that included the SI to account for potential confounding.

We conducted secondary analyses with the same methods to examine adherence to mitigation strategies by regressing the SI on log-transformed trip ratio. We also analyzed the effective reproduction number (R_t) as the outcome measure over incidence to further assess the impact of mitigation strategies and mobility on COVID-19 transmission, while accounting for the dynamically changing size of the susceptible population during an outbreak. We computed daily R_t values across US states with methods developed by Cori and colleagues,²¹ a commonly used Bayesian framework for deriving accurate estimates of instantaneous reproduction numbers²² in COVID-19 studies^23–26 using incidence time-series and serial transmission interval data as input. We assumed a serial interval of 5.2 days based on results from a systematic review.²⁷ The multilevel models for this analysis regresses the SI and trip ratio on log-transformed effective reproduction number.

Furthermore, we stratified all analyses by different time periods to account for the dynamic nature of the pandemic and the various “waves” of outbreaks in the United States. Given empirical evidence from surges observed in the trend of incident cases and the corresponding trends in community mitigation strategies and mobility, we disaggregated the pandemic in 2020 into the following three time periods such that they align with the waves observed in 2020: time period 1, from March 1, 2020, to June 18, 2020; time period 2, from June 19, 2020, to October 6, 2020; and time period 3, from October 7, 2020, to December 30, 2020. All regression analyses were stratified per these temporal periods, resulting in increased power for our analysis owing to the homogeneity of observations within the temporal periods. We conducted analyses at the state-level owing to a lack of reliable time-series data on community mitigation strategies for every US county. For all final models, we present the intraclass correlation coefficient to illustrate the fraction of the outcome variability that is accounted for by clustering. Complete model specifications are examined in detail in eAppendix p. 5–7, http://links.lww.com/EDE/B967.

To examine the robustness of our results based on the 11-day lag time applied to the SI and trip ratio, we conducted a one-way sensitivity analysis guided by the 95% confidence interval around the 11-day period.²⁰ We therefore repeated the primary analysis, where COVID-19 incidence proportion was the outcome using the same methods but changing the lag to values between 8 and 13 days. We further examined the appropriateness of the random effects specification for US states in our regression analysis by conducting a final sensitivity analysis where US states are used as fixed effects.

We conducted statistical analyses using R version 4.1.2 (R foundation for statistical computing) utilizing the EpiEstim package²¹ for computation of the effective reproduction number and the lme4 package²⁸ for multilevel modeling. Because all data used as input for analyses are publicly available and deidentified, this investigation was exempt from institutional review.

RESULTS

In 2020, there were approximately 20.1 million reported infections of COVID-19 in the United States, yielding a cumulative incidence proportion of 6,120 per 100,000 population. The number of reported infections were 2.20, 5.31, and 12.5 million in time period 1, time period 2, and time period 3, respectively. The temporal trend for the COVID-19 incidence exhibits three peaks in infections at 10.9 new cases per 100,000 population in time period 1, 23.1 per 100,000 population in time period 2, and 76.3 per 100,000 population in time period 3 (Figure). For the COVID-19 R_t, the value decreases sharply in the beginning of the time period 1 followed by an interval below 1 starting in mid-May. The number increases in the beginning in time period 2 but stays near 1 afterwards. R_t stays over 1 for almost all of time period 3. For the stringency index (SI), the index increases sharply from March to April in time period 1 and peaks at 72.4 in the mid-April. The SI steadily decreases after the peak until a small increase in November. Similarly, the trip ratio decreased quickly from March, until reaching its lowest value of 0.51 in April, and then gradually increased again in time period 1. The ratio began to decrease again in time period 2 and then slightly increased in time period 3 before leveling off.

FIGURE.

Temporal trends of (A) COVID-119 incidence proportion per 100,000 population, (B) COVID-19 effective reproduction number, (C) the stringency index, and (D) the trip ratio for the United States, March 1, 2020–December 31, 2020.

There was variation in these temporal trends across US regions (eFigure 3, http://links.lww.com/EDE/B967). COVID-19 incidence trends during time period 1 were mainly driven by the Northeast region, in time period 2 by the South and West regions, and in time period 3 by all regions though the Midwest region had the greatest daily incidence proportions in this time period. Trends in the SI were similar across the regions but the level varied by region. Generally, the Midwest and South regions had the lowest values for the SI throughout the entire study period. These two regions also had trip ratios closer to 1 compared with the West and Northeast regions.

Associations Between Stringency Index and Mobility on COVID-19 Incidence Proportion

In regression analyses between SI and COVID-19 incidence proportion with random intercepts for US states, the intraclass correlation coefficients (ICCs) for time period 1, time period 2, and time period 3 were the following: 0.469, 0.672, and 0.423 (Table 1). The ICCs were similar for models assessing the trip ratio-COVID-19 relationships. In time period 1, the SI was negatively associated with COVID-19 incidence proportion (β = 0.994, 95% CI = 0.991, 0.997), whereas the trip ratio was not related to COVID-19 incidence (β = 0.908, 95% CI = 0.608, 1.36). Specifically, we found that every five-point increase in the SI for time period 1 was associated with a 2.89% (95% CI = 1.52, 4.26%) reduction in COVID-19 incidence proportion.

TABLE 1.

Mixed-effects Linear Regression With Random Intercepts for US State Results for COVID-19 Incidence Proportion per 100,000 Population Stratified by Temporal Period

Parameter	Time Period 1 Model	Time Period 2 Model	Time Period 3 Model
	β (95% CI)	β (95% CI)	β (95% CI)
Stringency index model
Stringency index	0.994 (0.991, 0.997)	1.00 (0.998, 1.01)	0.990 (0.986, 0.994)
Within US state variance	0.682	0.266	0.312
Between US state variance	0.603	0. 543	0.223
ICC	0.469	0.672	0.423
Trip ratio model
Trip ratio	0.908 (0.608, 1.36)	1.35 (0.999, 1.83)	1.42 (1.04, 1.93)
Within US state variance	0.682	0.265	0.311
Between US state variance	0.605	0.55	0.235
ICC	0.47	0.674	0.43

Text in bold highlights regression coefficients. Although not shown here, models included a smooth function of time since pandemic onset with 4 degrees of freedom. The stringency index model does not include any other covariates, whereas the trip ratio model includes the stringency index as a covariate.

Time period 1: March 1, 2020, to June 18, 2020; time period 2: June 19, 2020, to October 6, 2020; time period 3: October 7, 2020 to December 31, 2020. CI indicates confidence interval; ICC, intraclass correlation coefficient.

For time period 2, the SI was no longer associated with COVID-19 incidence proportion (β = 1.00, 95% CI = 0.998, 1.01) but the trip ratio-COVID-19 incidence relationship trended in the positive direction (β = 1.35, 95% CI = 0.999, 1.83). The results indicate that after adjusting for the SI, decreasing the number of trips at the state-level compared with the same time in 2019 by 50% was associated with a 16.2% (95% CI = −0.07, 35.2%) reduction in COVID-19 incidence proportion for time period 2.

For the final time period, both the SI (β = 0.990, 95% CI = 0.986, 0.994) and trip ratio (β = 1.42, 95% CI = 1.04, 1.93) were associated with COVID-19 incidence proportion. Specifically, we found that every five-point increase in the SI during time period 3 was associated with a 5.01% (95% CI = 3.02, 6.95%) decrease in COVID-19 incidence proportion. In the same time period, we found that reducing the number of trips compared with the same time in 2019 by 50% at the state-level was associated with a 19.3% (95% CI: 2.30, 39.0%) reduction in COVID-19 incidence proportion per 100,000 population after adjusting for the SI.

Associations Between Stringency Index and Mobility

Results for regression analyses with random intercepts for location where the trip ratio is the outcome are in Table 2. The ICCs for time period 1, time period 2, and time period 3 were the following: 0.425, 0.729, and 0.581. In time period 1, we found that the SI (β = 0.999, 95% CI = 0.998, 0.999) was negatively associated with the trip ratio such that every five-point increase in the index was associated with a 0.69% (95% CI = 0.56, 0.83%) decrease in the trip ratio. The SI was no longer associated with the trip ratio in time period 2 (β = 1.00, 95% CI = 0.999, 1.00). However, the relationship reappeared in time period 3 (β = 0.998, 95% CI = 0.998, 0.999) with results illustrating that every five-point increase in the index was associated with a 0.93% (95% CI = 0.64, 1.22%) decrease in the trip ratio.

TABLE 2.

Mixed-effects Linear Regression With Random Intercepts for US State Results for the Trip Ratio Stratified By Temporal Period

Parameter	Time Period 1 Model	Time Period 2 Model	Time Period 3 Model
	β (95% CI)	β (95% CI)	β (95% CI)
Stringency index	0.999 (0.998, 0.999)	1.00 (0.998, 1.00)	0.998 (0.998, 0.999)
Within US state variance	0.007	0.002	0.011
Between US state variance	0.005	0.006	0.015
ICC	0.425	0.729	0.581

Although not shown here, models included a smooth function of time since pandemic onset with 4 degrees of freedom. The stringency index model does not include any other covariates.

Time period 1: March 1, 2020, to June 18, 2020; time period 2: June 19, 2020 to October 6, 2020; time period 3: October 7, 2020, to December 31, 2020. CI indicates confidence interval; ICC, intraclass correlation coefficient.

Associations Between Stringency Index and Mobility on the COVID-19 R_t

In regression analyses for the SI using random intercepts for US states where the COVID-19 (R_t) was the outcome (Table 3), the respective ICCs for time period 1, time period 2, and time period 3 were the following: 0.032, 0.067, and 0.128. The ICCs were consistent in the trip ratio models. We observed, in time period 1, that the SI was negatively associated with COVID-19 R_t (β= 0.992, 95% CI = 0.992, 0.994), whereas the trip ratio was positively associated with R_t (β= 1.93, 95% CI = 1.71, 2.19). The results indicate that every five-point increase in the SI was associated with a 3.46% (95% CI = 2.99, 3.92%) decrease in R_t. The trip ratio model displayed that reducing the number of trips compared with the same time in 2019 by 50% at the state-level was associated with a 39.1% (95% CI = 30.6, 48.1%) drop in COVID-19 R_t.

TABLE 3.

Mixed-effects Linear Regression With Random Intercepts for US State Results for Effective Reproduction Number Stratified by Temporal Period

Parameter	Time Period 1 Model	Time Period 2 Model	Time Period 3 Model
	β (95% CI)	β (95% CI)	β (95% CI)
Stringency index model
Stringency index	0.992 (0.992, 0.994)	0.997 (0.996, 0.997)	0.997 (0.996, 0.998)
Within US state variance	0.136	0.034	0.027
Between US state variance	0.005	0.002	0.004
ICC	0.032	0.067	0.128
Trip ratio model
Trip ratio	1.93 (1.71, 2.19)	0.909 (0.829, 0.996)	1.44 (1.33, 1.57)
Within US state variance	0.133	0.034	0.026
Between US state variance	0.004	0.003	0.006
ICC	0.032	0.068	0.181

Text in bold highlights regression coefficients. Although not shown here, models included a smooth function of time since pandemic onset with 2 degrees of freedom. The stringency index model does not include any other covariates, whereas the trip ratio model includes the stringency index as a covariate.

Time period 1: March 1, 2020, to June 18, 2020; time period 2: June 19, 2020, to October 6, 2020; time period 3: October 7, 2020, to December 31, 2020. CI indicates confidence interval; ICC, intraclass correlation coefficient.

In time period 2, the SI remained negatively associated with COVID-19 R_t (β = 0.997, 95% CI = 0.996, 0.997) such that every five-point increase in the SI was associated with a 1.72% (95% CI = 1.25, 2.18%) decrease in R_t. The trip ratio trended in the negative direction with COVID-19 R_t (β = 0.909, 95% CI = 0.829, 0.996). In time period 3, both the SI (β = 0.997, 95% CI = 0.996, 0.998) and trip ratio (β = 1.44, 95% CI = 1.33, 1.57) were found to be associated with the effective reproduction number. Every five-point increase in the SI was associated with a 1.33% (95% CI = 0.85, 1.80%) decrease in R_t. In the same time period, reducing the number of trips compared with the same time in 2019 by 50% at the state-level was associated with a 20.1% (95% CI = 15.1, 25.2%) drop in COVID-19 R_t.

One-way Sensitivity Analyses

Results for one-way sensitivity analyses utilizing different lags for the primary analysis where COVID-19 incidence proportion was the outcome are shown in eFigure 4, http://links.lww.com/EDE/B967. Briefly, using lags between 8 and 13 days did not substantially change regression coefficients for the SI or trip ratio during time periods 2 and 3. However, results appeared slightly more sensitive to the different lags during time period 1, particularly for the SI, such that the direction of the relationships changed at the earliest lag values. The final one-way sensitivity using fixed effects for US states over random effects yielded very similar results across the temporal periods for the primary analysis (eTable 2, http://links.lww.com/EDE/B967).

DISCUSSION

This investigation represents one of the first studies to empirically summarize the impact of community-based mitigation strategies and mobility on COVID-19 infections throughout the course of the pandemic in the United States in 2020. We observed that mitigation strategies and reductions in mobility have marked health gains through the reduction of COVID-19 infections but their impacts vary depending on the three observed “waves” of infections. More stringent policies were associated with declines in COVID-19 infections during the first and third waves but these policies did not affect infections during the second wave. Similarly, fewer trips taken per day were also associated with reduced COVID-19 infections but this relationship was only observed in the second and third waves in 2020.

Although our results indicate that every five-point increase in the level of stringency in mitigation policies was associated with an approximate 3% and 5% reduction in COVID-19 incidence during wave 1 and wave 3, respectively, we did not observe this association in the second wave. The lack of an association may be owing to poor compliance in mitigation strategies as the only instance of when the relationship between policy stringency and mobility was absent occurred in the second wave. These results align with previous work indicating that adherence to mitigation policies was the lowest in the summer months in the United States,¹⁰ but that adherence generally rebounds after substantial drops.²⁹ We found a similar phenomenon here as the policy stringency-mobility and policy stringency-COVID-19 incidence relationships reappeared during the third wave.

Similarly, the impact of mobility on COVID-19 infections varied over time as we found that cutting the number of trips taken in half compared with the same time in 2019 was associated with approximately 16% and 19% declines in the COVID-19 infection rate at the state-level in the second and third waves, respectively. However, this association was not present during the first wave. This finding is consistent with some studies finding that mobility was not associated with COVID-19 infections during the first wave of the pandemic.^15,30 Mobility may potentially have a less important role in COVID-19 transmission during the first “wave” compared with other adopted mitigation strategies and behavioral changes such as mask mandates, restrictions on gatherings, school closures, and maintaining physical distance.

Although we observed that the relationships between mitigation strategies and reductions in mobility on COVID-19 incidence changed across waves, there was less variation in these relationships when the COVID-19 R_t was evaluated. Both mitigation strategies and mobility were consistently associated with R_t across the waves. The second wave was the only exception where the relationship between and R_t changed. The difference in results is likely owing to incidence and the R_t characterizing different components of the pandemic; incidence measures daily levels of infections in real-time, R_t incorporates short-term information on the susceptible population. Although R_t provides an important measure of transmission for policymaking, incidence supplements decision-making as knowledge of the number of infected people in the population assists in how policies should change. For example, despite the South region having the largest daily incidence proportions in the second wave, we found that the region did not have the highest COVID-19 R_t. The additional information on case counts can provide further context on how mitigation strategies should adjust during waves.³¹

Previous studies evaluating the impact of COVID-19 mitigation measures have used mobility as a proxy measure of adherence on the basis that populations will make fewer trips and physically interact less when policies are in place.^11,15,32,33 Studies have shown that mobility declined once the first mitigation policies were implemented in March 2020³⁴ and that mobility increased shortly after policies were lifted.⁹ We therefore theorized that mobility, as a marker for compliance, is a possible mechanism through which COVID-19 mitigation strategies influences cases. Our results provider further evidence that mobility is an effective marker of adherence as our results evaluating the policy stringency-mobility relationship are aligned with evidence assessing compliance through self-report.¹⁰

Overall, the findings from this investigation have important implications for outbreaks of novel infectious diseases as we have observed that community mitigation strategies and reductions in mobility are effective in reducing COVID-19 infections throughout the pandemic in the absence of vaccines. Though the effects of mitigation strategies and mobility fluctuate over time, the findings underscore a need for policies to limit interactions with strategies to ensure adherence to mitigation policies. Decision makers may consider adopting comprehensive communication strategies to inform populations about COVID-19 risks, adopted interventions, and why adherence to mitigation strategies is critical, which have been shown to be effective in increasing compliance across various settings.^35,36

Although not examined here, political disposition has played an important role in both implementation of COVID-19 mitigation measures^37–39 and their compliance.^40–42 This body of evidence has shown that conservatives and liberals respond differently to the COVID-19 pandemic such that conservatives are less likely to adopt recommended behavior changes. Some evidence indicates that these partisan differences may be a function of varying levels of reported knowledge of COVID-19, perceived accuracy of media reports on the pandemic, and perceived health risks associated with COVID-19.⁴³ Effective approaches for reducing ideological effects on compliance are having tailored messaging, where residents are more likely to identify as conservative,⁴² messaging expressed from government officials of the same political party,⁴⁴ and increasing trust in science.⁴⁵ Put together, interventions aiming to increase adherence to COVID-19 mitigation strategies should consider the effects of political ideology when developing effective communication campaigns.

Although multiple reports have investigated the link between individual state-level social distancing policies,⁸ changes in mobility,⁴⁶ and COVID-19 transmission, many have not examined how these relationships are modified as function of epidemic waves as presented here. Further, the inclusion of daily social distancing policies and mobility patterns in the same time-series models improves inferences for assessing the dose-response effect of mobility on COVID-19. We were also able to incorporate multiple COVID-19 outcomes, incident cases and R_t, into our analyses, illustrating how mobility patterns and community mitigation strategies impact different aspects of COVID-19 dynamics.

Strengths and Limitations

This study has several strengths, including the ability to examine the impact of community mitigation strategies and mobility across the entire pandemic for the United States in 2020. We were also able to investigate how the relationships change depending on the three waves that were observed in the United States. In addition, our mobility measure, the trip ratio, was able to control for seasonality of movement by linking the current days’ mobility patterns to the patterns experienced in the identical days of the same month before the COVID-19 pandemic. This is in contrast to other well-known mobility indices produced by Google, Apple, and Facebook that always compare current mobility patterns to patterns in early 2020. When mobility is compared with patterns in January and February of 2020, changes in mobility occurring in mid to late 2020 may be a function of differences in general mobility patterns (e.g., changes in weather may increase mobility) rather than factors related to the pandemic.

However, our analyses should be interpreted in the context of their limitations. First, the ecological nature of the data prevents us from making inferences regarding mobility and COVID-19 infections at the individual-level. Second, given the comprehensive nature of the stringency index, we were not able to assess the effects of individual interventions. Rather, our results reflect the cumulative impacts of various mitigation strategies on COVID-19 infections. Third, there is potential exposure misclassification associated with the stringency index owing to individual metropolitan areas and counties deviating from state-level policies according to local behaviors and infection levels. We believe misclassification was partially minimized as the stringency index took into account the scope and intensity of individual policies throughout the state. Fourth, the COVID-19 infection data used in this analysis reflects reported diagnoses and thus does not represent the true burden of COVID-19 owing to underreporting. Finally, there are likely other important factors influencing COVID-19 infections (e.g., smoking prevalence, population age-sex structure, mask wearing prevalence, population density) that this study could not take into account owing to a lack of complete time-series data.

CONCLUSION

Overall, this investigation empirically demonstrates that community-based mitigation strategies and reductions in mobility were associated with fewer COVID-19 infections, consistent with the hypothesis that these strategies helped contain the COVID-19 pandemic in the absence of safe and effective vaccines. The variation in the strength of associations between mitigation strategies and mobility on COVID-19 incidence across the “waves” of infections demonstrate that compliance to interventions is likely important to the success of mitigation strategies and reductions in mobility may be particularly effective as mitigation policies are relaxed. Further research is needed to empirically investigate how mitigation strategies and mobility levels interact with mass vaccination strategies in managing pandemics.