Systematic review of empiric studies on lockdowns, workplace closures, and other non-pharmaceutical interventions in non-healthcare workplaces during the initial year of the COVID-19 pandemic: benefits and selected unintended consequences

Coronavirus disease (COVID-19) is an infectious disease caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) that emerged in December 2019. COVID-19 has caused a global pandemic that resulted in long-term health problems as well as millions of deaths around the world [1]. The World Health Organization Director-General indicated that all countries must strike a fine balance between protecting health and minimizing economic and social disruption [2]. Several community-level containment and closure policies were implemented by government authorities to reduce the transmission of SARS-CoV-2 and avert overwhelming of healthcare systems. These policies included cancellation of public events, restrictions on gathering sizes, restrictions on internal movement and international travel, closure of public transport systems, school closures, closures of non-essential businesses, and lockdowns [3]. Governments provided fiscal support to varying extents to reduce financial hardship due to the COVID-19 pandemic and the interventions to reduce SARS-CoV-2 transmission [1, 4, 5].

About two-thirds of the global population over 15 years of age participate in the labor force [6]. SARS-CoV-2 transmission can occur in workplaces through respiratory droplets and aerosols generated by pre-symptomatic, asymptomatic, or symptomatic persons and through fomites [7, 8]. In 2020, employers were encouraged to implement several measures to prevent and reduce the transmission of SARS-CoV-2 within the workplace, including use of face masks or coverings, physical distancing to increase physical space between people and decrease the frequency of face-to-face contact (including teleworking), symptom and temperature screening, flexible leave policies to facilitate self-isolation of sick workers, environmental cleaning and disinfection, and engineering controls to improve air quality (Additional file 1: Appendix Table S1) [9‐11]. These measures could be used by essential businesses that were not subject to government-mandated closures and by all businesses when lockdowns were not in effect.

Research has primarily focused on preventing or reducing SARS-CoV-2 infection in healthcare workers, with non-healthcare workers receiving less attention [12]. A Cochrane systematic review on interventions in non-healthcare workplaces examined the effect of interventions introduced by researchers [12]. The review identified one study that met their inclusion criteria, which was a cluster-randomized non-inferiority trial that assigned staff working in schools to standard isolation after contact with a SARS-CoV-2-infected person or to daily COVID-19 testing and staying at work if the test was negative. Because randomizing employers or geographic regions to workplace-related non-pharmaceutical interventions (NPIs) may not be feasible or ethical during an outbreak, observational studies may provide the best available evidence. We conducted a systematic review to assess the benefits and unintended consequences of NPIs in non-healthcare workplaces that included observational studies. The objectives of our review were to evaluate the effects of NPIs within non-healthcare workplaces and community-level workplace closures and lockdowns, compared to no intervention, on the following outcomes in workers or the general population: 1) COVID-19 morbidity and mortality, 2) selected mental disorders, and 3) employment outcomes.

We registered our systematic review protocol on PROSPERO (ID # CRD42020182660) [13]. We reported the review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement (Appendix Table S2) [14].

Search of the databases yielded 15,529 studies. After screening titles and abstracts, we reviewed the full text of 853 studies for eligibility (Fig. 1). Among these studies, we excluded 806 that did not meet the inclusion criteria. The percent agreement between reviewers was 95% for title and abstract screening and 87% for full-text reviews. We identified 47 observational studies through database searching and 13 via other sources (i.e., examination of references of previous systematic reviews and authors’ knowledge), yielding a total of 60 observational studies that met the inclusion criteria. Forty studies reported on COVID-19 morbidity and mortality outcomes (Appendix Table S4) [30‐70], 15 assessed the effect on anxiety and depression symptoms (Appendix Table S5) [71‐85], and five assessed the effect on unemployment and labor force participation (Appendix Table S6) [5, 86‐89]. The studies were based on data from the first year of the pandemic, mostly covering the period March to July 2020. The domain-specific and overall risk of bias for each study are shown in Appendix Tables S7-S9. Studies that were excluded from the review are listed in Appendix Table S10.

Of the 40 studies on COVID-19 morbidity and mortality, 16 were based on data from the USA, and 13 studies analyzed data from multiple countries, ranging from 2 to 202 countries (Appendix Table S4). Other studies included data from countries in Europe (Spain, Italy, Germany), Asia (India, China), Africa (South Africa), and Australia. The median study period over which outcome data were collected was 10 weeks (interquartile range: 8 weeks, 17 weeks). The overall risk of bias was moderate for 25 studies, serious for 14 studies, and critical for one study (Appendix Table S7). All studies had at least a moderate risk of confounding, and most studies had a low risk of bias for the other six domains. Although studies on physical distancing (teleworking) [41] and physical barriers [45] reported a significant decrease in COVID-19 incidence in workers, and studies on initial business closures (i.e., restaurant or entertainment business closures) reported a significant decrease in COVID-19 case growth rate and epidemic doubling time in the general population [35, 67], the sign tests were not significant (Table 1). Studies on workplace closures reported a decrease in COVID-19 incidence or case growth rate (six of seven studies) and reproduction number (four studies) in the general population, but the sign tests were not significant (Table 1). Studies showed that lockdown significantly decreased COVID-19 incidence or case growth rate (23 studies, p < 0.001), reproduction number (11 studies, p < 0.001), and COVID-19 mortality or death growth rate (seven studies, p < 0.05) in the general population (Table 1). The 23 studies on the effect of lockdown on COVID-19 incidence or case growth rate reported a variety of effect measures, with seven studies reporting percentage decrease in daily case growth rate (median: 6 percentage decrease) [33‐35, 42, 50, 56, 65], and six studies reporting the growth rate before and after lockdown (median growth rate before lockdown: 18.0 percentage increase; median growth rate after lockdown: 3.8 percentage increase) [32, 60, 61, 66, 68, 70].

Among the 15 studies on anxiety and depression symptoms, 10 were conducted in European countries (Spain, Italy, Germany, Ireland, United Kingdom) and two were conducted in the USA (Appendix Table S5). All studies reported on the effect of lockdown, with the median interval between the baseline and follow-up outcome measurements being 6 weeks. Several instruments were used for assessing anxiety symptoms, including the Generalized Anxiety Disorder Scale [76‐79, 82], the Brief Symptom Inventory [74, 81], and the Depression Anxiety and Stress Scale [83, 84]. Instruments for assessing depression symptoms included the Patients Health Questionnaire [71, 76‐79, 82], the Brief Symptom Inventory [74, 81], and the Depression Anxiety and Stress Scale [83, 84]. The overall risk of bias was serious for five studies and critical for 10 studies (Appendix Table S8). This was mainly because of risk of bias in the confounding, selection of participants, and missing data domains. Fourteen studies reported the effect of lockdown on anxiety and/or depression symptoms in the general adult population and one study reported the effect on depression symptoms in workers. For the effect of lockdown on anxiety symptoms, the pooled standardized mean difference was -0.02 (95% CI: -0.06, 0.02) (Fig. 2a). For the effect of lockdown on depression symptoms, the pooled standardized mean difference was 0.16 (95% CI: 0.10, 0.21) (Fig. 2b). The funnel plot for depression symptoms showed some asymmetry in the distribution of studies about the pooled standardized mean difference (Fig. 3b), and the Trim and Fill adjustment indicated that publication bias could account for the observed effect (adjusted pooled standardized mean difference = 0.001, 95% CI: -0.04, 0.02).

Among the five studies on unemployment and labor force participation, three were from the USA, one from Mexico, and one from Australia (Appendix Table S6). The median interval between the baseline and follow-up outcome measurements was 3 months. The overall risk of bias was moderate in two studies and serious in three studies (Appendix Table S9). All studies had a moderate or serious risk of confounding, and one study had a serious risk of bias because of missing data. The studies showed that lockdown increased unemployment (pooled mean difference = 4.48 percentage points, 95% CI: 1.79, 7.17) (Fig. 4a) and decreased labor force participation (pooled mean difference = -2.46 percentage points, 95% CI: -3.16, -1.77) (Fig. 4b).

Empiric studies showed that lockdown reduced COVID-19 incidence or case growth rate, reproduction number, and COVID-19 mortality or death growth rate in the general population during the initial year of COVID-19 pandemic. We found few studies on the effect of NPIs other than lockdown on COVID-19 morbidity and mortality outcomes. Lockdown increased unemployment and decreased labor force participation, but no effect was observed on anxiety symptoms. Lockdown had a small effect on increasing depression symptoms, but publication bias could account for the observed effect. The risk of bias for most of the studies on COVID-19 and employment outcomes was moderate or serious, and that for the studies on anxiety and depression symptoms was serious or critical.

Non-pharmaceutical measures can reduce SARS-CoV-2 transmission by reducing the likelihood of transmission per contact and by reducing contacts between infectious and healthy persons [90]. Studies published in 2023 found that employed adults who had telework experience before illness onset were less likely to work onsite while ill during COVID-19 and other acute respiratory illnesses than persons without telework experience, suggesting that telework may reduce workplace virus exposure [91, 92]. Systematic reviews that assessed the effect of physical distancing and screening in non-workplace settings or on other respiratory viruses provide indirect evidence for the effect of these measures on COVID-19 illness in non-healthcare workplaces. A systematic review assessed the effectiveness of physical distancing measures in non-healthcare workplaces on influenza attack rates [93]. One review included studies of physical distancing on COVID-19 illness in settings other than workplaces (e.g., ≥ 3 vs. ≥ 6 feet distancing policies in schools; frequency of close contact with a primary case in a household) [94]. A Cochrane rapid review assessed the effect of symptom/exposure-based or test-based screening of international travelers for SARS-CoV-2 at borders before or after travel [95]. Systematic reviews of modeling studies on the effect of NPIs within non-healthcare workplaces on COVID-19 illness are needed because modeling studies fill in gaps of information when decisions must be made and there is limited information [96, 97].

Recent systematic reviews of empiric studies have assessed the effect of workplace closures and lockdowns [18, 94, 98]. Two of these reviews included cross-sectional studies [94, 98]. We excluded cross-sectional studies because it is difficult to assess cause-and-effect relationships from such studies [99]. The previous reviews reported that workplace closures and lockdowns reduced COVID-19 incidence, case growth rate, reproduction number, COVID-19 mortality, and death growth rate in the general population [18, 94, 98]. Lockdowns have been shown to reduce population mobility, with increased time at home, reductions in visits to shops and workplaces, and decline in use of public transport [17].

Our systematic review did not find conclusive evidence that lockdown increased anxiety and depression symptoms. A previous rapid review of studies published from January 2020 to June 2020 reported small effects of lockdown on anxiety and depression symptoms [100]. Among the 11 empiric studies on anxiety and depression symptoms included in the review, four were conducted in college or university students and thus not directly relevant to our systematic review. Another review estimated that the global prevalence of anxiety and depression symptoms increased during the COVID-19 pandemic compared to the pre-pandemic period [101]. The authors attributed the increase in anxiety and depression symptoms to the combined effects of the spread of SARS-CoV-2 and the interventions, including lockdown, school and workplace closures, decreased public transport, and reduction of social interactions. Several risk factors for anxiety and depression during lockdown have been reported. Risk factors for anxiety include loneliness and history of mental health issues, while higher level of resilience and spiritual well-being are associated with lower anxiety [77, 78]. Risk factors for depression include loneliness, detachment, negative affect, history of mental health issues, concerns about changes at work and running out of money, and unemployment [71, 77, 84]. On the other hand, protective factors associated with depression include more resilient coping style, higher level of resilience, spiritual well-being, and moderate-to-vigorous physical activity [75, 77, 78, 84].

Our systematic review showed lockdown increased unemployment and decreased labor force participation. Lockdown can directly lead to layoffs because of business closures, cancelation of events, and reduced economic activities. However, in the absence of lockdown, employment can be affected by individuals’ refraining from activities outside their household to reduce their risk of infection, which can lead to decreased consumer spending and business revenues [5, 88]. We did not identify any previous systematic reviews of the effect of lockdown on unemployment and labor force participation.

Findings of our systematic review should be considered in context of at least seven limitations. First, some studies on the effects of workplace closures and lockdowns on COVID-19 outcomes used quasi-experimental designs (controlled before after, interrupted time series) that can allow for causal inferences without randomized trials [102, 103], but it is unclear if the assumptions required to ensure valid causal inference were met. The findings therefore need to be interpreted as showing an association. Second, the included studies often did not describe in detail the interventions that were assessed, which may make it difficult to compare findings across studies. Third, many NPIs were implemented together or within a short time, and so the independent effects of interventions may be difficult to determine [104], particularly for studies that did not have a concurrent control group. Fourth, the number of COVID-19 cases could have been underestimated to a greater degree during the early phase of the pandemic because of limited availability of COVID-19 tests. However, the underestimation would likely bias the effect of an intervention toward the null [105]. Fifth, several studies on the effect of lockdown on anxiety and depression symptoms collected baseline data after the start of lockdown, and so the magnitude of the effect may be under-estimated. In addition, anxiety and depression were assessed using screening questionnaires that identified probable cases, and the findings may not be extrapolated to diagnosed cases of anxiety and depression [101]. Sixth, although our electronic search identified grey literature (e.g., working papers, medRxiv preprints) [36, 40, 51, 86, 88], we did not specifically search preprint databases or dissertations and theses databases. Finally, we limited studies to English when we performed the electronic searches and screened articles, and thus the findings may not be generalizable to studies published in other languages.

However, this systematic review also has several strengths. We assessed both desired effects (i.e., public health benefits) and secondary (unintended / unwanted) effects of NPIs during the initial year of COVID-19 pandemic. Additionally, we used several electronic databases to search for studies and examined the references of previous systematic reviews, which increased the comprehensiveness of the literature search. Next, our review was based on empiric studies that provide direct evidence of effectiveness in real-world settings.

The COVID-19 pandemic had unequal effects on the population, with people who could work remotely faring better in terms of health and socioeconomic wellbeing than persons who were required to work in-person, such as those in goods production or essential industries [1]. Minority and low-income vulnerable persons are over-represented in high-risk essential industries [1, 45, 106]. COVID-19 death rates in the U.S. have been estimated to be substantially higher in Hispanics and non-Hispanic Blacks compared to non-Hispanic Whites [107, 108]. Compared to people working in non-essential sectors, those working in essential sectors (particularly in agriculture, emergency, manufacturing, facilities, and transportation or logistics) were found to have higher COVID-19 deaths [109, 110]. It is important to deploy effective science based NPIs to reduce health inequities and decrease overall disease transmission, especially in industries where work cannot be performed remotely.

Our systematic review showed that several empiric studies assessed the effect of lockdowns, but there is a paucity of studies on the effects of other interventions undertaken in many workplace settings, including temperature/symptom screening, use of different barrier protections including some which were not previously proposed as an NPI or tested (e.g., plexiglass or curtain partitions), and physical distancing measures within the workplace. With the availability of COVID-19 vaccines and effective therapeutics that reduce hospitalizations and deaths [1], as well as the desire to avoid detrimental effects on daily life and the economy, the use of workplace closures and lockdowns abated after the initial year of the pandemic in most countries. However, because SARS-CoV-2 remains endemic and because it evolved into variants which can evade immunity acquired through prior infection or vaccination and transmit more efficiently [111], use of less disruptive NPIs including better ventilation, face masks, and some variations of physical distancing within the workplace may still have relevance. Addressing the gaps in the evidence base on the effects of NPIs pertaining to workplaces is therefore important for informing ongoing prevention strategies as well as future pandemic preparedness.

There was scarce direct evidence on the benefits of symptom and/or temperature screening, physical barriers, and physical distancing measures to reduce COVID-19 illness within workplaces that are open. While the use of these interventions is less likely to be perceived as disruptive for work process than lockdowns, they are not likely to be effective in reducing the transmission of an airborne virus like SARS-CoV-2 that can be readily spread in indoor settings by asymptomatic or pre-symptomatic individuals. There was evidence to indicate that lockdown helped reduce COVID-19 morbidity and mortality in the general population, but it increased unemployment and reduced labor force participation. It is important for countries that implement lockdown in future outbreaks of emerging infectious diseases or pandemics to consider strategies to mitigate these unintended consequences.