COVID-19 Prevention and Control Measures in Workplace Settings: A Rapid Review and Meta-Analysis

Workplaces can be high-risk environments for SARS-CoV-2 outbreaks and subsequent community transmission. Identifying, understanding, and implementing effective workplace SARS-CoV-2 infection prevention and control (IPC) measures is critical to protect workers, their families, and communities. A rapid review and meta-analysis were conducted to synthesize evidence assessing the effectiveness of COVID-19 IPC measures implemented in global workplace settings through April 2021. Medline, Embase, PubMed, and Cochrane Library were searched for studies that quantitatively assessed the effectiveness of workplace COVID-19 IPC measures. The included studies comprised varying empirical designs and occupational settings. Measures of interest included surveillance measures, outbreak investigations, environmental adjustments, personal protective equipment (PPE), changes in work arrangements, and worker education. Sixty-one studies from healthcare, nursing home, meatpacking, manufacturing, and office settings were included, accounting for ~280,000 employees based in Europe, Asia, and North America. Meta-analyses showed that combined IPC measures resulted in lower employee COVID-19 positivity rates (0.2% positivity; 95% CI 0–0.4%) than single measures such as asymptomatic PCR testing (1.7%; 95% CI 0.9–2.9%) and universal masking (24%; 95% CI 3.4–55.5%). Modelling studies showed that combinations of (i) timely and widespread contact tracing and case isolation, (ii) facilitating smaller worker cohorts, and (iii) effective use of PPE can reduce workplace transmission. Comprehensive COVID-19 IPC measures incorporating swift contact tracing and case isolation, PPE, and facility zoning can effectively prevent workplace outbreaks. Masking alone should not be considered sufficient protection from SARS-CoV-2 outbreaks in the workplace.


Introduction
The novel SARS-CoV-2 is a respiratory pathogen causing COVID-19 [1]. Transmission is associated with exposure to droplets, fomites, and aerosols, particularly in crowded or confined spaces [2,3]. Asymptomatic infections occur in 17-21% of cases [4]. These dynamics can result in COVID-19 outbreaks and superspreading events (SSEs) that lead to changes in the community reproductive number [5].
Workplaces are common settings for explosive infectious disease outbreaks due to transmission between employees and their close contacts in respective households and communities [6]. As employees have a legal right to a safe workplace, employers must create safe working conditions [7]. However, in the absence of evidence on COVID-19 infection prevention and control (IPC) measures, employers have relied on two methods. First, many employers have applied the precautionary principle of a "better safe than sorry" approach comprising a bundle of measures for all employees [8]. Second, employers have sought direction from guidelines on preventing transmission of other respiratory pathogens such as influenza and severe acute respiratory syndrome (SARS) that have been modified for COVID-19 [9]. These approaches have had consequences, namely global shortages of personal protective equipment (PPE) and increased unemployment due to workplace closures [10]. Now, more than 15 months since SARS-CoV-2 emerged, it is crucial to identify measures that prevent and control pathogen transmission in workplace settings. Effective IPC measures can guide future efforts to respond to global health emergencies by protecting susceptible employees and citizens [11]. Therefore, this rapid review aims to investigate the effectiveness of measures to prevent and control COVID-19 outbreaks in workplace settings. Specifically, this study seeks to (1) map the IPC measures investigated in included studies and (2) assess COVID-19 positivity estimates associated with IPC measures implemented in workplace settings.

Methods
Given the urgency to identify evidence-based IPC measures for preventing COVID-19 outbreaks in workplace settings, a rapid review was completed.

Research Questions
The Problem/Population, Intervention, Comparison, and Outcome (PICO) framework was used to formulate research questions facilitating a precise search for occupational IPC measures (I/C) that reduce COVID-19 transmission (O) in the workplace environment (P) [12]. The following research questions were defined:
What IPC measures reduce COVID-19 infections in the workplace?

Workplace COVID-19 IPC Measures
Prior to study selection and analysis, COVID-19 IPC measures were broadly categorized according to World Health Organization (WHO) guidelines for health and safety in the workplace [13]. Categories were defined as:

1.
Surveillance measures-COVID-19 symptom monitoring, strategies to screen or test individuals, such as symptomatic or asymptomatic testing; 2.
Outbreak investigations and response-contact tracing and testing of close contacts, quarantine of potentially infected individuals or groups, self-isolation of confirmed cases; 3.
Environmental adjustments-improving airflow and ventilation, adding physical barriers to help employees avoid physical contact, environmental cleaning; 5.
Education initiatives-training on IPC measures, communication and signage, ongoing education and support; 6.
Changes in work arrangements-social distancing, facility zoning, entrance restrictions, changes in assignments for high-risk workers (i.e., individuals with medical conditions, pregnant women, over-60 population), facility shutdown, paid sick leave; 7.
Combined measures-approaches that combine measures from two or more categories.

Defining Effectiveness
Articles that formally assessed or modelled whether the implemented workplace measures prevented and/or controlled workplace transmission of SARS-CoV-2 were considered in this review. IPC measures that result in (i) lower numbers of confirmed cases among all employees tested following the intervention (i.e., COVID-19 positivity), (ii) lower secondary attack rates among (non-)household contacts, (iii) higher percentage reductions in effective reproductive number (R eff ), or (iv) percentage of COVID-19 cases prevented, may generally be considered more effective. However, in the case of surveillance and contact tracing, a high positivity percentage may be considered more effective, given that more cases are being successfully identified. We take care over the course of this review to consider contextual variability and endeavour not to overstate effective and ineffective interventions.

Search Strategy
Publications in any language were identified on four electronic databases: Medline, Embase, PubMed, and Cochrane Library. The following search terms were used: SARS-CoV-2 OR "severe acute respiratory syndrome coronavirus" OR "SARS" AND (workplace OR employers OR "healthcare workers" OR "nursing homes" OR "residential care" OR "meat factories" OR "factories" OR industry OR occupation OR "occupational health" OR "creche" OR "childcare facilities" OR work*).

Study Selection (Identification, Screening, and Inclusion)
Studies that were conducted in any workplace and geographical setting were included if they met at least one of the following criteria, by undertaking: (1) surveillance of a population over a specified time period, (2) an outbreak investigation, (3) an assessment of a COVID-19 IPC measure or policy, or (4) mathematical modelling to estimate effectiveness.
Studies that met the following criteria were excluded: (1) mental health outcomes, (2) modelling studies with assumptions of effectiveness (e.g., effectiveness of IPC measures as modelling assumption(s) rather than outcome(s) to be predicted), (3) systematic reviews, and (4) qualitative investigations, including commentaries and editorials. Though included in our initial search strategy, studies related to SARS-COV or MERS-CoV were eventually excluded as they were less relevant to the novel SARS-CoV-2/COVID-19.
Identified studies were imported into Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Covidence automatically removed duplicate studies. To rapidly screen studies, 20 postgraduate researchers trained in public health and medicine were enlisted. In pairs, researchers screened the title and abstract of imported studies. The same process was repeated for full text review, with eligible articles subsequently included for data extraction.

Data Extraction
For each included article, two researchers independently extracted data into structured Excel sheets. Any discrepancies in extracted data were identified by the principal investigator (PI) and resolved by discussion between extractors. Data on the country of study conduct, study design, workplace setting, population investigated, IPC measure(s) implemented, outcome(s) measured, study start and end dates, and study duration were extracted from the included studies. Studies were selected for quantitative synthesis if they (1) had complete denominators, (2) assessed a similar intervention to at least one other study, (3) included a comparable outcome measure, and (4) were deemed sufficiently homogeneous in clinical and methodological characteristics to permit meta-analysis. For quantitatively aggregable studies, additional information was compiled on the national COVID-19 positivity estimate(s) during the study period, and the trajectory of the national pandemic wave during that time (e.g., acceleration, deceleration, peak) according to the Johns Hopkins Coronavirus Resource Center [14]. For included studies that implemented combined measures, the total number of interventions implemented was also recorded.

Quality Assessment
Given the diversity of included study designs and interventions, risk of bias could not be assessed using a single, existing tool. An adapted checklist modelled after Gulumian et al. [15] was used to score longitudinal studies based on experimental study (yes/no), total study population reported (yes/no), gold-standard PCR testing used (yes/no), and intervention follow-up time reported (yes/no). Scores ranged from 1 (lowest) to 4 (highest). Cross-sectional studies automatically received a quality score of 1 due to their low quality of evidence [16].
The Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) was used to assess the quality of modelling studies. Modelling studies were assessed for risk of bias according to source of data, outcome to be predicted, candidate predictors, sample size, missing data, model development, model performance, model evaluation, results, interpretation, and discussion. Quality assessment was completed by six reviewers and discrepancies were solved mutually between reviewers.

Statistical Analysis
Random-effects meta-analyses were performed using the DerSimonian-Laird method to estimate the pooled effect of IPC measures implemented in two or more studies. Because we assumed, prior to model selection, that aggregated studies would not share a common effect size (due to between-study variation in study design, setting, and intervention), random-effects models were considered more appropriate than fixed-effects models [17]. Study weights were assigned using the inverse of each study's total variance (i.e., by combining within-and between-study variation). An arcsine-based transformation was used to stabilize the variance of each study's proportion estimate through a conventional, two-step meta-analytic approach [18]. First, each study's COVID-19 positivity estimate was transformed using the Freeman-Tukey double arcsine transformation to approximate a normal distribution required for meta-analysis. Then, meta-analytic results on the transformed scale were back-transformed to interpret pooled estimates and 95% confidence intervals (95% CI) [19]. The Freeman-Tukey double arcsine transformation method was chosen as it can account for studies with zero or one effect size [19]. Cochran's Q test was performed to estimate whether total variation was statistically different compared to expected variation when assuming that all aggregated studies share a common, underlying positivity estimate. We also calculated a Tau-squared (T 2 ) parameter and I 2 statistic to estimate between-study variance and the proportion of total variation that is due to real differences between studies' positivity estimates, respectively. We selected I 2 > 75% to define "high" heterogeneity. Outlying effect sizes were identified by screening for externally studentized residuals (Z > 2) and excluded if they exerted considerable influence on summary effect size [20].
When 10 or more studies assessed similar interventions using the same outcome variable, univariate meta-regression analysis was performed to examine the influence of other factors on intervention effectiveness. The following predetermined factors were studied: study region, duration of intervention, community transmission rate during intervention period, community pandemic wave interval, implementation of specific interventions (asymptomatic PCR testing, facility zoning, education, environmental cleaning, PPE, syndromic surveillance, contact tracing), and the total number of interventions implemented. To account for multiple comparisons, a Bonferroni correction was applied, and the significance threshold was set at p < 0.01. The significance threshold for all other tests was set at p < 0.05.
When study numbers allowed (n ≥ 10), publication bias was tested for using Egger's test. Sensitivity analysis was performed by repeating meta-analyses while excluding lowquality studies (adapted checklist quality score < 2). All statistical analyses were completed using R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Metaanalyses and meta-regressions were performed using the R packages meta 4.18-0 and metafor 2.4-0.
Hospitals and healthcare centres (n = 45) were the most common workplace settings in the included studies, followed by nursing homes (n = 11), offices (n = 2), manufacturing facilities (n = 1), meat factories (n = 1), and modelling of general workplaces (n = 1). Tables 1-3 present an overview of the study characteristics and IPC measures from healthcare studies (Table 1), nursing home studies (Table 2), and other workplaces (Table 3).  Anti-SARS-CoV-2 antibody testing in addition to RT-PCR testing

Study Quality
Of the 61 included studies, 48 were of longitudinal design (i.e., researchers examined the same group of workers over a period of time). The amended checklist assessment for longitudinal studies showed that only one of these was experimental [70]. Two of the 48 longitudinal studies did not report a complete denominator [41,67], whereas 19 out of 48 did not specify intervention follow-up time. No study attained the highest overall quality score of 4. Twenty-three studies had a quality score of 3, and 19 had a quality score of 2.
Studies with a quality score of 1 (k = 12), including cross-sectional studies, were considered low-quality evidence (Tables 1 and 2). Evaluations of mathematical modelling studies using the CHARMS checklist identified low risk of bias in the seven modelling studies included in this review. Detailed quality assessment findings for all studies are provided in Supplementary Materials.

Effectiveness of Workplace COVID-19 IPC Measures
COVID-19 IPC measures implemented by studies included in the rapid review comprised six categories: (i) surveillance, (ii) outbreak investigation and response, (iii) PPE, (iv) changes in work arrangements, (v) worker education, and (vi) combined measures. No studies assessing the effectiveness of environmental adjustments were identified. Measures tested as part of mathematical modelling studies were considered separately due to their hypothetical nature. Tables 4-6 map the array of single, combined, and modelled measures implemented by category and the studies that assessed their effectiveness.

Combined Preventive Measures. Study Reference
Hospital layout adjustments, training, pre-testing and triage, environmental cleansing, PPE [48] Standard operating procedure, staff training, symptom reporting, enhanced cleaning, inventory monitoring protocols [34] Social distancing, universal masking, testing of all symptomatic patients [61] Home-based 7-day infection control strategy for exposed HCWs-symptomatic, asymptomatic RT-PCR testing [29] General screening and cohort isolation [42] PPE and sanitation guidelines implemented, epidemiological investigation and contact tracing of high-risk HCWs, symptomatic swab testing [46] Integrated infection control strategy: zoning, PPE, mass surveillance [62] PPE, visitor restrictions, universal face masks, universal RT-PCR patient admission screening [71] Hospital shut down, universal testing of all inpatients, medical staff, and employees [56] Systematic testing, social distancing, monitoring of sick leaves, on-site visits by hygiene experts, staff training, direct communication of all measures to personnel and patients [67] Social distancing, surgical masks, systematic testing [68] Regulation of access to the company, social distancing, hygiene and PPE, cleaning and sanitizing of environments, worker education [22] Contact tracing, reinforced hygiene practices, PPE, education, and signage [44]

Modelled Preventive Measures Study Reference
Variations in employee testing frequency (daily, weekly, bi-weekly, monthly) [80] Variations in testing frequency; outbreak vs. non-outbreak testing [81] Testing and symptomatic isolation; regular screening of high-risk groups; close contact quarantine [75] Non-adaptive combinatorial matrices used for group testing [79] Self-isolation and variations in contact tracing methods; mass testing [76] Variations in PPE use; worker age restrictions [78] Variations in PPE efficacy and testing frequency [77] All 61 studies contained a statistical measure of effectiveness for one or more COVID-19 IPC measures. More than half of the studies (34/61) used COVID-19 positivity rates to assess effectiveness between 4 and 270 days following the implementation of IPC measures. Except in the case of single testing and contact tracing measures, lower positivity was considered more effective. For test and trace measures, higher positivity was considered more effective as it meant COVID-19 cases were successfully being captured. Other outcome types included attack rates, mean reduction in R eff , odds ratios (OR), relative risk (RR), and hazards ratios of COVID-19 infection. The median intervention duration was 41 days (min 1-max 300). Thirty-three studies contained results that were amenable to meta-analyses of proportion estimates, stratified by intervention type ( Table 7). Note that no studies falling into the "changes in work arrangements" and "worker education" intervention categories met inclusion criteria for formal quantitative synthesis.
Pooled COVID-19 positivity estimates amongst employees who underwent universal, asymptomatic RT-PCR testing (25,023 HCWs from nine studies; 54 office employees from one study) was 1.7% (95% CI = 0.9-2.9) ( Table 7). Egger's regression asymmetry test was non-significant (p = 0.93), indicating the potential absence of publication bias. To explore heterogeneity between asymptomatic testing interventions (I 2 = 96%), univariate meta-regression was conducted; however, none of the pre-identified potential moderating factors-that is, study region and duration, rates of community transmission, and community pandemic wave interval-were significantly associated with greater COVID-19 positivity or reduction in heterogeneity.

PPE
The effectiveness of PPE in reducing workplace COVID-19 transmission was assessed in 12 studies [26,27,32,33,41,43,45,52,58,66,70]. In the healthcare setting, three studies assessed the effectiveness of general PPE [26,33,43], and three assessed universal masking policies [41,45,55]. One study looked at the effectiveness of using N95 respirators in the hospital setting compared to surgical masks [27]. Studies also assessed the effectiveness of cloth masks compared to medical masks [70], of having adequate PPE supply [66], of homemade PPE for laboratory workers in India [32], and of how eye protection compares to face protection [58].
Studies that used combined measures had a lower pooled COVID-19 positivity estimate at the end of their intervention period than studies that implemented single measures (Table 7, Figure 2).
Across the 15 studies that implemented combined IPC measures and could be included in meta-analysis, 0.2% (95% CI 0.0-0.4) of 31 196 healthcare and nursing home workers tested positive for COVID-19. Univariable meta-regression analysis was conducted to assess for pre-determined moderators that might explain between-study heterogeneity (I 2 = 80%) ( Table 8). Studies conducted in Europe had higher positivity than studies conducted in Asian countries (QM (df = 2) = 7.14, p = 0.03), with study region explaining~43% of effect size heterogeneity. Studies conducted in communities with over 5% positivity during the intervention period were not significantly more likely to experience higher COVID-19 positivity estimates (QM (df = 1) = 3.39, p = 0.07), whereas implementing a higher number of interventions resulted in significantly lower COVID-19 positivity estimates (QM (df = 1) = 6.10, p = 0.01). Other than asymptomatic testing (QM (df = 1) = 4.96, p = 0.02), no single intervention type resulted in significantly lower COVID-19 transmission in studies that implemented it compared to those that did not. Study duration was also not significantly associated with COVID-19 positivity (QM (df = 1) = 43.96, p = 0.21).  Table 5. Note that two studies ( [64,73]) were classified under the "Outbreak Investigations and Response" category in Table 4. We have chosen to include them in the Combined Measures meta-analysis due to their comprehensive nature. Egger's regression asymmetry test indicated potential for publication bias amongst the 15 studies that implemented combined measures (p = 0.04) (Table 4); however, sensitivity analyses excluding lower-quality evidence (Quality Score < 2) did not affect pooled effect sizes or meta-regression results.

Key Findings from Modelling Studies
Modelling studies provided additional support for the effectiveness of combined IPC measures versus single interventions, particularly when implemented early (Table 9). Models showed that random, asymptomatic testing in settings with low to moderate SARS-CoV-2 transmission rates did little to stop outbreaks [77,80], whereas asymptomatic outbreak testing effectively prevented workplace infections [81]. Testing alone was not enough to reduce R 0 below 1 [76] but effectively reduced transmission when paired with high-quality infection control practices [81]. Timely and widespread contact tracing (<24 h contact after the first case identified,~80% coverage) [75,76], swift self-isolation of cases [76], and effective use of PPE [77] were reported to slow and/or prevent SARS-CoV-2 transmission. Other effective IPC measures included creating smaller worker/patient cohorts in hospitals [77], and restricting patient contact for high-risk HCWs over the age of 60 [78].  Due to the high level of heterogeneity in terms of study type, interventions assessed, and outcome measured, results from 28 studies could not be included in formal quantitative synthesis. From these, several studies exemplary (quality score > 1) that demonstrate the effectiveness of diverse interventions have been selected, key findings from which are summarized in Table 10. The risk of COVID-19 infection was more than double among HCWs self-washing their masks compared with the hospital laundry (HR 2.04 (95% CI 1.03 to 4.00); p = 0.04). There was no significant difference in infection between HCWs who wore cloth masks washed in the hospital laundry compared with medical masks (p = 0.5).
• Self-washing cloth masks by hand more than doubles the risk of seasonal respiratory illnesses. • Double-layered cloth masks washed in the hospital laundry were as protective as medical masks.
[52] PPE/Meatpacking After initiating both universal masking and physical barrier interventions, 8/11 facilities showed a statistically significant reduction in COVID-19 incidence in <10 days. Facilities that only initiated a universal mask policy showed no significant difference before and after the intervention.
• Together, universal masking and physical barriers can prevent COVID-19 transmission in meatpacking plants.

•
These interventions should be accompanied by ventilation enhancements and worker education on mask use and adherence. • Monitored adherence to infection control processes, especially proper wearing of PPE and cohorting, can reduce weekly infections and mortality. [72] Changes in work arrangements/Nursing home Long-term care facilities (LTCF) that organized staff compartmentalization within zones were significantly more likely to avoid a COVID-19 outbreak (OR = 0.19 (0.07-0.48)) as were LTCFs whose staff perceived high-quality implementation of preventive measures (OR = 0.65 (0.43-0.98)).
• Staff should be encouraged and incentivized to work in single care homes and movement of staff across multiple care homes should be limited. • Infection control should be extended for all contacts, including those between staff, whilst on the care home premises.

Summary of Findings
The review identified 61 articles that implemented and assessed COVID-19 IPC measures in the workplace. The studies showed that universal asymptomatic RT-PCR testing yielded low employee positivity rates, indicating few cases identified at potentially high cost in moments of reduced community transmission. Asymptomatic testing more effectively captured cases when implemented following facility outbreaks or environmental monitoring; however, studies generally concluded that all testing should be combined with high-quality workplace infection control practices. Staff compartmentalization within zones and/or cohorts (worker bubbles), for example, was identified as an effective way to prevent workplace transmission. Universal masking, though a critical component of most initiatives to protect workers, proved inadequate in reducing workplace transmission when implemented alone. Masking was more effective when combined with physical barriers. While studies mentioned the added value of environmental adjustments and worker education for maximizing masking efficiency, there was a gap in quantitative evidence supporting this. Results from contact tracing interventions varied widely, indicating the potential moderating role of community transmission rates and/or other contextual factors in contact tracing effectiveness. Nevertheless, ongoing syndromic surveillance and outbreak investigations tended towards lower post-intervention COVID-19 positivity estimates than once-off contact tracing and testing initiatives. Mathematical modelling demonstrated the role that contact tracing coverage and timeliness, and added physical distancing measures, could play in maximizing the effectiveness of test-and-trace initiatives.
Meta-analyses using random-effects models supported these findings, highlighting lower COVID-19 positivity estimates in workplace settings that implemented combined measures compared with settings that applied single measures. Though different regional and workplace contexts prevent the identification of a cure-all combination of measures, PPE, timely and thorough outbreak investigations, syndromic surveillance and testing, and staff compartmentalization within zones emerge as important considerations.

Discussion
Over one year into the COVID-19 pandemic, workplace settings remain a high-risk environment for SARS-CoV-2 outbreaks, presenting great risk to the health and wellbeing of employees, their families, and surrounding communities. Due to the rapidly changing nature of COVID-19 and related management, control, and prevention guidelines, employers have struggled to implement timely and effective COVID-19 protective measures in the workplace [82]. To develop a greater understanding, this rapid review of the literature was designed to compile evidence on COVID-19 IPC measures implemented in global workplace settings through April 2021. Specifically, this review (1) mapped existing measures and (2) assessed their effectiveness.
Despite high levels of heterogeneity in study type, region, setting, and outcomes measured, there was a consensus in the literature on the increased effectiveness of combined versus single measures, thereby providing evidence in support of layered mitigation strategies recommended by national and international health authorities [13,83]. Unsurprisingly, meta-regression revealed a positive association between the total number of interventions implemented and reduced employee COVID-19 positivity estimates. However, this begs the question of cost-effectiveness since workplaces, particularly in low-income developing countries, may lack resources to implement exhaustive measures. Fortunately, Juneau and colleagues considered the question of cost-effectiveness in their own systematic review and identified swift contact tracing and case isolation, surveillance networks, PPE, and early vaccination when possible as the most cost-effective combination of interventions, particularly when adopted early [84]. Workplace closures have been reported as effective but costly [84].
Appropriate combinations of measures may also vary depending on workplace context. Healthcare settings, for example, must consider which type of PPE is most effective (low-quality evidence suggests that N95 respirators should be used during aerosol generating procedures (AGP) instead of surgical masks) [27,85], or the possibility of restricting the age of workers with direct patient contact [77]. Food and manufacturing facilities have other, unique considerations such as how environmental conditions, sick leave policies, and access to health insurance for employees affect worker vulnerability to COVID-19. Unfortunately, no evidence was identified on how these factors related to COVID-19 transmission in the meatpacking and manufacturing facilities included in this review.
Other significant research gaps emerged. Most studies assessed the effectiveness of IPC measures in hospital and nursing home settings, demonstrating the extent to which healthcare facilities and staff have been disproportionately impacted by the COVID-19 pandemic. However, few if any studies assessed IPC measures in manufacturing, industrial, essential retail, and public service settings, which have continued to provide essential goods and services to the public throughout the global pandemic. The geographical spread of included studies was concentrated in Europe, Central Asia, and North America, revealing a gap in emerging literature from South America, Southern Asia, Oceania, and Africa. Research on COVID-19 occupational safety measures is critical to protect essential workers in epicentres such as Brazil, Argentina, Colombia, India, and Peru, where case numbers rank amongst the highest in the world [14]. Conversely, Australia and New Zealand have been remarkably adept at preventing local COVID-19 outbreaks [14]; however, no literature was identified detailing how their workers were effectively protected.
Though several studies implemented education initiatives as part of combined interventions, few assessed the effectiveness of worker education specifically. Research has shown that HCWs can feel overwhelmed by constantly changing IPC guidelines, that adherence to guidelines is influenced by levels of support and communication from management, that training is most effective when mandatory, and that there is a need for training on the infection itself and proper PPE use [9]. In meatpacking plants, home to internationally diverse worker cohorts, language barriers have been identified as a primary challenge in deploying COVID-19 IPC strategies but the effectiveness of implemented strategies such as multilingual signage and mass-communication apps remains unassessed [86]. Moreover, this review found little evidence on how cultural factors may be influencing workplace viral transmission and/or interacting with COVID-19 IPC measures. Such findings underline the pressing need for research on how tailored education initiatives can help protect workers.
Our rapid review identified another gap in evidence surrounding effective contact tracing methods. While included studies mentioned contact tracing frequency and coverage, none discussed methods for ensuring adequate public engagement despite research identifying privacy concerns, mistrust, unmet needs for information, and digital challenges as potential barriers to engagement and subsequent contact tracing effectiveness [87]. Finally, given strong evidence that SARS-CoV-2 spreads by airborne transmission [3], this review revealed a need for experimental research on how indoor environmental adjustments relate to COVID-19 outbreaks and superspreading events. While the WHO has released guidelines for improved workplace ventilation, our lack of findings on indoor air adjustments supports claims that precise ventilation or air-purification regimes to improve workplace safety remain unknown [88]. Countries have begun implementing promising, low-cost strategies for reducing workplace risk through improved indoor air quality. In Belgium, for example, COVID-19 ventilation rules requiring carbon dioxide monitors to be on public display have been imposed, allowing workers and members of the public to determine when air quality reaches unsafe levels [89]. Yet, experts note that a lack of studies addressing CO 2 monitoring as a public health tool during the pandemic leads to uncertainty surrounding optimal CO 2 levels for COVID-19 conditions [90]. This type of evidence will be of great value to the indoor workforce moving forward. Scientists and public health agencies recommend free and low-cost options for preventing viral transmission inside such as opening windows, using window fans, turning off demand-controlled ventilation controls, and repositioning supply/exhaust diffusers [83]. Workplaces should strongly consider adding these adjustments to the layered IPC approach identified as effective by this review.

Completeness and Applicability of Evidence
Concerning the compiled evidence base, most included studies were observational in nature, which may impact the strength of findings and recommendations. The necessity of responding to a global emergency meant that some researchers had to document workplace transmission as it unfolded, or to assess outbreaks retrospectively. These assessments provide important insight on how to protect workers from COVID-19; however, they do not allow for strong conclusions on the effectiveness of particular IPC measures. Experimental studies provide higher-quality evidence but, as they are difficult to design and carry out in emergency settings, only one was identified for inclusion in this rapid review. This review also included mathematical modelling studies that, while providing support for the effectiveness of timely and comprehensive workplace interventions, may vary in terms of real-world applicability. Results from this review are therefore not intended to guide clinical practice, but to enhance the understanding of what IPC measures have been implemented, and/or implemented effectively, in workplaces thus far.

Limitations and Strengths
Our rapid review and meta-analysis presented several limitations and challenges. First, due to the time frame of the grant through which this rapid review was funded (Science Foundation Ireland Grant 20/COV/8539) and limited availability of graduate student volunteers, we were required to begin data extraction as soon as our protocol was complete. This disqualified the review from subsequent submission to the online PROSPERO database.
Additional challenges were encountered during formal analysis. Nearly half of included studies could not be included in meta-analyses due to variability of outcome variables and/or incomplete denominators. This resulted in several small pools of comparable studies (k = 2-3), potentially impacting the robustness of meta-analytic results. Though we sought to test community COVID-19 transmission rates as a potential moderating factor using meta-regression, many studies reported long intervention periods and/or imprecise study locations, making it difficult to identify precise community transmission rates. This may have contributed to why community transmission was only marginally associated with COVID-19 positivity in the workplace. Many potential moderators of the relationship between IPC measures and pooled positivity estimates were deemed to be non-significant using meta-regression. However, our inability to detect statistically significant associations for many of these pre-determined moderators may not be due to a non-existent or meaningless effect but rather a lack of statistical power (due to too few studies) to detect a small but meaningful difference between subgroups. Additionally, few studies reported changes in COVID-19 outcomes over time, making it difficult to account for temporal dynamics of workplace transmission when assessing intervention effectiveness. Finally, this rapid review contains a potentially unrepresentative sample of global workplaces as most included studies yielded from healthcare, nursing home, European, Asian, and North American settings. Results should therefore be interpreted with caution when applied to other professional, geographical, and/or cultural contexts.
Despite these limitations, this rapid review makes several important contributions to global understanding of COVID-19 in workplace settings. To our knowledge, this is the first study to map all COVID-19 IPC measures implemented and assessed in workplace settings. Most included studies were performed in healthcare settings, which has important implications for other workplace settings. By demonstrating that high-risk healthcare settings were able to effectively contain and/or prevent workplace COVID-19 outbreaks, this rapid review demonstrates the feasibility of lower-risk workplace settings remaining open at minimized risk to employees and the wider community. As our study demonstrates, this will require combinations of surveillance, swift contact tracing and case isolation, facility zoning, and universal masking. This study shows that masking alone should not be considered sufficient protection against workplace outbreaks. Despite widespread vaccination programs, COVID-19 case numbers remain high worldwide and workplace safety is of critical importance. Our findings can help to protect workers in countries where businesses are beginning to reopen and public interactions increasing; in developing countries where infections remain uncontrolled; and in future settings when other respiratory diseases with airborne transmission threaten workplace safety.

Implications for Research
This rapid review highlights the extent to which workplaces including hospitals, nursing homes and, to a lesser extent, manufacturing, meatpacking, and office settings are at increased risk for SARS-CoV-2 outbreaks and in need of timely and effective infection prevention and control measures. Moving forward, experimental studies that address identified gaps in the workplace COVID-19 IPC literature base are of pressing importance. This includes research on effective IPC education initiatives, on ways to maximize public engagement in contact tracing, and on how precise environmental conditions relate to COVID-19 transmission in the workplace. While this rapid review identified a large body of evidence on how to protect healthcare and social care workers, further research should consider ways to better support essential workers in retail, education, construction, transportation, and manufacturing settings. High-quality evidence on effective workplace COVID-19 IPC measures in under-studied and highly positive regions such as Central-South America and South Asia is also necessary to ensure the protection of all workers, not just those in wealthy countries.

Conclusions
Workplaces remain common settings for infectious disease outbreaks and superspreading events. Though many national and global health authorities provide recommendations for workplace health and safety, there has been a reported absence of evidence on precise COVID-19 IPC regimes to make workplaces safe. By conducting a synthesis of the evidence base on workplace COVID-19 IPC measures and their effectiveness, our results can inform guidelines on how to better protect workers from COVID-19 and future infectious disease outbreaks. Moving forward, timely and comprehensive IPC measures should be favoured in workplace settings. Though exact combinations of measures may vary depending on professional, geographical, or cultural context, our rapid review identified swift and thorough contact tracing and case isolation, effective PPE, syndromic surveillance and testing, and staff zoning and/or cohorts as important considerations. These measures should be paired with improved building ventilation and indoor air quality. Governments should prioritize funding for these initiatives, particularly for small businesses who may lack financial resources for adequate IPC programs. Masking alone should not be considered sufficient protection from SARS-CoV-2 outbreaks in the workplace. Our findings indicate that applying timely and comprehensive infection prevention and control measures can allow workers to safely remain in or return to the workplace in the context of COVID-19 and at the outset of future epidemics.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ijerph18157847/s1, Table S1: Quality assessment checklist for included longitudinal studies; Table S2: CHARMS Quality Assessment of Modelling Studies; Sheet S1: Meta-analysis data. Institutional Review Board Statement: Ethical review and approval were waived for this study by the University College Dublin Human Research Ethics Committee-Sciences based on its status as a low-risk project.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data used for meta-analysis is provided in the Supplemental Materials under Supplemental Sheet S1: Meta-analysis data.