The profound impact of the novel coronavirus (SARS-Cov-2) has been driven by the ease with which human-to-human transmission occurs, contributing to the rapid propagation of coronavirus disease 2019 (COVID-19). SARS-Cov-2 can be transmitted through cough or respiratory droplets, contact with infected bodily fluids, or less commonly, from contaminated surfaces. Reference Chan, Yuan and Kok1,Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2
Healthcare workers (HCWs) are particularly vulnerable to SARS-CoV-2 infection and other emerging, highly infectious diseases due to close contact with infected patients and contaminated materials. Reference Sepkowitz and Eisenberg3 Previous coronaviruses, such as severe acute respiratory syndrome coronavirus (SARS) and Middle East respiratory syndrome coronavirus (MERS), have demonstrated extensive transmission in healthcare settings even though they are relatively inefficient in transmission within the general community. Reference Booth, Matukas and Tomlinson4,Reference Lee, Hui and Wu5 As of July 14, 2020, data from Italy estimated that healthcare providers managing patients with COVID-19 account for 12% of cases. Reference Guzzetta, Marziano and Poletti6 Factors believed to contribute to the rapid spread among healthcare workers include suboptimal infection control practices, performance of aerosol-generating medical procedures, and failure to continue adequate mask use in break rooms. Reference Suwantarat and Apisarnthanarak7–Reference Çelebi, Pişkin and Çelik Bekleviç9 The prevalence of infected HCWs also differs by hospital units, being highest in medical intensive care units and emergency departments. Reference Wang, Brull, Patel, Massouh and Abdallah10
The preservation of health and wellness in HCWs is paramount because of their role in caring for critically ill patients as well as the need to prevent outbreaks in healthcare facilities. Reference Chirico, Nucera and Magnavita11 Currently, understanding of COVID-19 infection rates in HCWs and the risk factors predisposing to infection in pandemic settings is limited, and infection control guidelines across international organizations are inconsistent. Reference Islam, Rahman and Sun12 Prior systematic reviews have focused on subsets of viral respiratory infections, but none have focused on risk factors for HCW infection in pandemic settings. A recent meta-analysis found protective effects of face masks, eye protection, and physical distancing in preventing virus transmission in both public and healthcare settings. Reference Chu, Akl and Duda13 Healthcare settings are unique in their challenges to financial and PPE resources, workforce availability, inherent fear, and anxiety among frontline staff, which are exacerbated during novel viral outbreaks. Reference Kisely, Warren, McMahon, Dalais, Henry and Siskind14 The current study provides a thorough review of occupational risk factors for infection in HCWs and protective measures necessary to mitigate risk in such rare and challenging times. Therefore, in this systematic review and meta-analysis, we aimed to identify risk factors for HCW infection during a WHO-classified epidemic of a highly infectious viral respiratory infection, comparable to COVID-19.
Methods
Search strategy and selection criteria
The study was prepared according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines Reference Liberati, Altman and Tetzlaff15 and was guided by specifications outlined in the Meta-analysis of Observational Studies (MOOSE) recommendations. Reference Stroup, Berlin and Morton16 The study was registered on PROSPERO (CRD42020176232) on April 6, 2020.
The search strategy was developed in consultation with a medical librarian and was conducted according to recommendations in the Cochrane Rapid Review guide. Reference Garritty, Gartlehner and Kamel17 The searches were conducted in electronic databases MEDLINE, EMBASE, CINAHL, and Cochrane CENTRAL from database conception until July 6, 2020 (Appendix 1 online). We excluded case reports, case series, editorials, narrative reviews, consensus opinions, news articles, and letters to the editor. Searches were restricted to articles written in English and studies involving human subjects only.
Titles and abstracts were screened to identify potentially eligible studies, which subsequently underwent full-text review for study inclusion using predetermined inclusion and exclusion criteria. Literature screening and eligibility assessment was performed independently by 2 reviewers (C.T., O.L.) at all stages. Reasons for exclusion were documented at each stage. Data extraction was conducted independently by 2 authors (C.T., O.L.) using a standardized data extraction form. Opinions from senior authors were solicited to resolve any conflicts.
Studies were included if the study population was comprised of HCWs in a healthcare setting with pandemic respiratory disease with a similar outbreak and transmission dynamics (droplet size) to COVID-19, including MERS, SARS, H1N1, and H5N1. Studies describing nonrespiratory infectious diseases, infectious diseases not defined by the World Health Organization (WHO) as epidemic or pandemic (eg, seasonal influenza), and diseases occurring in nonhealthcare settings were excluded. HCWs were defined as all staff in a healthcare facility involved in the provision of care to patients, not only those directly providing patient care. 18 Only comparative studies with a valid infected HCW (cases) group and a noninfected HCW (control) group were included. Therefore, studies that reported the prevalence of risk factors (described below) in both case and control groups were eligible for inclusion. We included observational studies (eg, cross-sectional, cohort, or case-control studies) and experimental studies (eg, randomized control trials [RCTs]).
Outcomes of interest
We sought to answer 3 knowledge questions: (1) Which types of HCWs and which medical departments are at increased risk of infection? (2) Which infection prevention and control practices are associated with protective effects for infection in HCWs? (3) Which exposures or procedures are associated with infection in HCWs? We collected data related to occupational risk factors that addressed these questions using 4 outcomes (categorical variables) in the case (infected HCWs) and control (non-infected HCWs) groups. (1) We collected data related to HCW occupation type as described previously. 18 (2) We collected data related to work department (eg, ward, emergency [ER], intensive care unit [ICU]). Frontline HCW were defined as those with high occurrence of patient face-to-face contact, including ER staff, ICU staff, and HCW who responded affirmatively to having exposure with patients. We sought to determine whether the health facility was a designated treatment center or was unidentified as a designated center. (3) We collected data related to the following infection prevention and control practices (IPAC): personal protective equipment (PPE) use (eg, surgical mask, N95 respirator or equivalent, gowns, full-body protection, eye and face protection, gloves, proper donning and doffing techniques), hand hygiene, IPAC training, vaccination status, pharmaco-prophylaxis. (4) We collected data related to exposure and procedural risks, that is, exposures to infected patients and colleagues, contaminated materials, participation in intubation or other aerosol-generating medical procedures (AGMPs). Reference Verbeek, Rajamaki and Ijaz19
Data analysis
All statistical analyses and the meta-analysis were performed on STATA version 15.1 software (StataCorp, College, TX) 20 and Comprehensive Meta-Analysis version 3 software (Englewood, NJ). Reference Borenstein, Hedges, Higgins and Rothstein21 We performed meta-analyses using a DerSimonian and Laird random-effects model for continuous and dichotomous outcomes, wherever applicable. Pooled effect estimates were obtained by calculating the odds ratios (ORs) for dichotomous outcomes along with their respective 95% and 99% confidence intervals (CIs). A subgroup analysis was conducted for each infectious agent.
Inverse variance weighted meta-regression analysis was performed to investigate the association between study characteristics and relevant outcomes. We included categorical variables (eg, virus type, designated centre, IPAC training, and ICU status) in the meta-regression models, wherever applicable. The R Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 statistic was used to measure the proportion of the variability in the outcome measure explained by the statistical model.
The quality of nonrandomized studies was assessed using the Newcastle-Ottawa scale (NOS) adapted to each study’s design. Reference Stang22–Reference Moskalewicz and Oremus24 Sensitivity analyses were conducted excluding studies with higher risk of bias. Heterogeneity between studies was assessed qualitatively and quantitatively using the Higgins I Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 statistic. Publication bias was assessed using Egger regression and visual inspection of funnel plots. Evidence was evaluated according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework. Reference Guyatt, Oxman and Akl25
Results
After the removal of duplicated search results, 6,936 articles underwent title and abstract screening. Of these, 204 full-text articles were assessed for eligibility for inclusion. Overall, 54 studies were included for analysis (Fig. 1).
Fig. 1. Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) reporting of systematic reviews and meta-analysis flow diagram outlining the search strategy results from initial search to included studies. PRISMA indicates preferred reporting items for systematic reviews and meta-analyses.
The included studies represented a total population of 191,004 healthcare workers and 7,375 cases of confirmed infection by the pathogen under study. All included studies were comparative and observational in nature, including 28 retrospective cohort studies, 10 case-control studies, 11 prospective cohort studies, and 5 cross-sectional studies, and the studies were conducted across 5 continents among 20 countries (Table 1). The infectious agents evaluated included COVID-19 (17 studies, n = 152,019), Reference Zheng, Wang and Zhou26–Reference Mani, Budak and Lan42 H1N1 (18 studies, n = 26,349), Reference Balkhy, El-Saed and Sallah43–Reference Kuster, Coleman and Raboud60 SARS (15 studies, n = 6,360), Reference Caputo, Byrick, Chapman, Orser and Orser61–Reference Reynolds, Anh and Thu75 MERS (3 studies, n = 5,750), Reference Hastings, Tokars and Abdel Aziz76–Reference Kim, Choi and Jung78 and H5N1 (1 study, n = 526). Reference Bridges, Katz and Seto79 No eligible RCTs were identified. The vast majority of studies (49 of 54; 90%) used WHO-defined criteria for confirmation of cases (Table 1).
Table 1. Study Characteristics
Note. SARS, severe acute respiratory syndrome; MERS, Middle East respiratory syndrome coronavirus; WHO, World Health Organization. Higher number of stars indicates lower risk of bias. WHO case definition in Appendix 6 (online).
Study quality ranged from poor (n = 27), to fair (n = 2), to good (n = 25) (Appendix 2 online). Reference Viswanathan, Ansari and Berkman80 To adjust for study quality, sensitivity analyses including only studies with low risk of bias (NOS ≥ 7) did not yield any significant change in effect estimates for outcomes. Evidence of publication bias from visual inspection of funnel plots and the Egger test was not strongly indicative (Table 2; Appendix 3 online).
Table 2. Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) of Meta-Analyzed Outcomes by 3 Knowledge Questions
a All studies were nonrandomized and evaluated using the Newcastle-Ottawa Scale (NOS). Most studies were at a lower risk of bias (NOS ≥7 stars). Furthermore, sensitivity analysis excluding studies with higher risk of bias did not yield any important difference in effect. Therefore, risk of bias was not downgraded.
b While there was a high I Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 value, there was a large amount of overlapping of confidence intervals and low variation of effect estimates across studies. Thus, inconsistency was not downgraded.
c Low heterogeneity was detected with overall I Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 <50% or some heterogeneity was explained through subgroup analysis demonstrating lower I2 value(s) <50%.
d Clinical heterogeneity associated with variable definitions of hand hygiene was probably introduced and inconsistency was downgraded.
e All studies included reported risk factors for health care workers infection of a highly infectious respiratory virus (SARS, H1N1, MERS, or H5N1) with a valid noninfected comparator group. Each disease-causing pathogen have caused epidemics with sufficient similarity in severity and transmission patterns. All outcomes (ie, infected cases) were ‘confirmed’ or ‘probable’ based on World Health Organization case definition criteria. Therefore, we did not rate down for indirectness of population, exposure, comparator, or outcomes.
g Magnitude of effect is large considering the thresholds set by GRADE (RR >2 or <0.5) with consistent evidence from at least 2 studies. Effect size assumes that the odds ratios translate into similar magnitudes of relative risk estimates.
h Although publication bias was suggested through the Egger test, visual inspection of funnel plots was largely symmetrical and thus, we did not downgrade for strongly suspected publication bias.
i No other virus-specific immunizations were identified in the literature.
fDowngraded 1 point because of large confidence intervals that overlaps both little to no effect, as well as appreciable benefit or appreciable harm of the intervention/exposure. This suggests that more studies with larger sample sizes are needed to calculate precise effect estimate.
Infection rates in frontline HCWs were analyzed from 32 studies Reference Lahner, Dilaghi and Prestigiacomo28,Reference Korth, Wilde and Dolff29,Reference Barrett, Horton and Roy31,Reference Eyre, Lumley and Donnell33,Reference Houlihan, Vora and Byrne34,Reference Ran, Chen, Wang, Wu, Zhang and Tan36,Reference Heinzerling, Stuckey and Scheuer38,Reference El-Boghdadly, Wong and Owen39,Reference Bai, Wang and Huang41,Reference Mani, Budak and Lan42,Reference Lobo, Oliveira, Garcia, Caiaffa Filho and Levin45–Reference Zhang, Seale and Yang51,Reference Bhadelia, Sonti and McCarthy53,Reference Chen, Lee and Barr54,Reference Chu, Li and Wang56,Reference Jefferies, Earl and Berry59,Reference Kuster, Coleman and Raboud60,Reference Teleman, Boudville, Heng, Zhu and Leo63–Reference Wilder-Smith, Teleman, Heng, Earnest, Ling and Leo65,Reference Lau, Fung and Wong69–Reference Loeb, McGeer and Henry71,Reference Bridges, Katz and Seto79,Reference Alraddadi, Al-Salmi and Jacobs-Slifka81–Reference Ho, Singh and Habib83 and were significantly higher in this group of HCWs compared to nonfrontline HCWs (OR, 1.66; 95% CI, 1.24–2.22, P = .001; 12.0% in frontline vs 4.4% non-frontline; low certainty) (Fig. 2; Table 2). Meta-regression analysis using random effects was performed by including covariates, wherever applicable. The overall risk of infection was higher among frontline workers (2-sided P = .039; τ2 = .3435; R2 = 72%). Furthermore, working within a designated center versus an unidentified center was protective (OR, − 1.04, 95%CI, −1.53 to −0.33; P = .004) (Table 1; Appendix 5, Fig. 1 online). Our model was unable to detect statistical difference in infection risk between the various virus types (P = .566). Similarly, there was low certainty that the difference in overall infection rates between physicians and nurses was not statistically significant (Table 2; Appendix 4, Fig. 1 online). Reference Zheng, Wang and Zhou26,Reference Lahner, Dilaghi and Prestigiacomo28,Reference Chen, Tong and Wang30–Reference Chatterjee, Anand and Singh32,Reference Houlihan, Vora and Byrne34–Reference Lai, Wang and Qin37,Reference Bai, Wang and Huang41,Reference Balkhy, El-Saed and Sallah43–Reference Lobo, Oliveira, Garcia, Caiaffa Filho and Levin45,Reference Sandoval, Barrera and Ferres49–Reference Zhang, Seale and Yang51,Reference Chen, Lee and Barr54–Reference Costa, Silva, Tavares and Nienhaus57,Reference Pei, Gao and Yang67,Reference Liu, Tang and Fang70,Reference Nishiyama, Wakasugi and Kirikae73–Reference Hastings, Tokars and Abdel Aziz76,Reference Alraddadi, Al-Salmi and Jacobs-Slifka81,Reference Chen, Leo, Ang, Heng and Choo82,Reference Hudson, Toop, Mangin, Brunton, Jennings and Fletcher84
Fig. 2. Forest plot of random effect meta-analysis of the risk of infection in frontline healthcare workers (HCWs) by virus type. Frontline HCWs were defined as those with high occurrence of patient face-to-face contact, including emergency department staff, intensive care unit staff, and HCWs who responded affirmatively to having direct exposure with patients.
Compared to control (ie, no use of corresponding PPE item), use of gloves (16 studies) Reference Barrett, Horton and Roy31,Reference Chatterjee, Anand and Singh32,Reference Heinzerling, Stuckey and Scheuer38,Reference Marshall, Kelso and McBryde46,Reference Toyokawa, Sunagawa and Yahata50,Reference Zhang, Seale and Yang51,Reference Chokephaibulkit, Assanasen and Apisarnthanarak55,Reference Jaeger, Patel and Dharan58,Reference Caputo, Byrick, Chapman, Orser and Orser61,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Wilder-Smith, Teleman, Heng, Earnest, Ling and Leo65,Reference Pei, Gao and Yang67,Reference Lau, Fung and Wong69,Reference Liu, Tang and Fang70,Reference Nishiura, Kuratsuji and Quy72,Reference Raboud, Shigayeva and McGeer74 , gowns (8 studies) Reference Barrett, Horton and Roy31,Reference Chatterjee, Anand and Singh32,Reference Toyokawa, Sunagawa and Yahata50,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Pei, Gao and Yang67,Reference Lau, Fung and Wong69,Reference Nishiura, Kuratsuji and Quy72,Reference Raboud, Shigayeva and McGeer74 , surgical masks (12 studies) Reference Heinzerling, Stuckey and Scheuer38,Reference Sandoval, Barrera and Ferres49–Reference Zhang, Seale and Yang51,Reference Chokephaibulkit, Assanasen and Apisarnthanarak55,Reference Jaeger, Patel and Dharan58,Reference Pei, Gao and Yang67,Reference Liu, Tang and Fang70–Reference Nishiyama, Wakasugi and Kirikae73,Reference Reynolds, Anh and Thu75 , N95 respirators (15 studies) Reference Wang, Pan and Cheng27,Reference Wang, Huang and Bai35,Reference Guo, Wang and Hu40,Reference Toyokawa, Sunagawa and Yahata50,Reference Zhang, Seale and Yang51,Reference Chokephaibulkit, Assanasen and Apisarnthanarak55,Reference Caputo, Byrick, Chapman, Orser and Orser61,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Wilder-Smith, Teleman, Heng, Earnest, Ling and Leo65,Reference Lau, Fung and Wong69–Reference Loeb, McGeer and Henry71,Reference Raboud, Shigayeva and McGeer74,Reference Kim, Choi and Jung78,Reference Alraddadi, Al-Salmi and Jacobs-Slifka81 , and face protection (11 studies) Reference Chatterjee, Anand and Singh32,Reference Toyokawa, Sunagawa and Yahata50,Reference Kuster, Coleman and Raboud60,Reference Caputo, Byrick, Chapman, Orser and Orser61,Reference Pei, Gao and Yang67–Reference Liu, Tang and Fang70,Reference Raboud, Shigayeva and McGeer74,Reference Alraddadi, Al-Salmi and Jacobs-Slifka77,Reference Kim, Choi and Jung78 were associated with large reductions in the risk of infection (moderate certainty; Table 2; Appendix 4, Figs. 2–6 online). The definition of N95 respirator use varied greatly across studies. The 2 studies with the strongest effects for use of N95 respirators both investigated COVID-19, but they did not clearly define the setting in which this occurred. Reference Wang, Pan and Cheng27,Reference Wang, Huang and Bai35 Furthermore, most studies of N95 respirators did not provide detail on the comparison group (eg, surgical mask, no mask) and had varying definitions for the use of N95 respirators, such as use all of the time, Reference Guo, Wang and Hu40 always while in an infected patient’s room, Reference Raboud, Shigayeva and McGeer74,Reference Elkholy, Grant, Assiri, Elhakim, Malik and Van Kerkhove85 or during intubation. Reference Caputo, Byrick, Chapman, Orser and Orser61
Across 13 studies, Reference Ran, Chen, Wang, Wu, Zhang and Tan36,Reference Guo, Wang and Hu40,Reference Sandoval, Barrera and Ferres49–Reference Zhang, Seale and Yang51,Reference Chokephaibulkit, Assanasen and Apisarnthanarak55,Reference Kuster, Coleman and Raboud60,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Wilder-Smith, Teleman, Heng, Earnest, Ling and Leo65,Reference Chen, Ling and Lu68,Reference Lau, Fung and Wong69,Reference Nishiura, Kuratsuji and Quy72,Reference Raboud, Shigayeva and McGeer74 compared to controls (no hand hygiene), hand hygiene following exposure to patients showed an overall significant protective effect (OR, 0.54; 95% CI, 0.34–0.87; P = .012) (low certainty; Table 2; Appendix 4, Fig. 7 online). IPAC training (6 studies Reference Guo, Wang and Hu40,Reference Pei, Gao and Yang67–Reference Liu, Tang and Fang70,Reference Raboud, Shigayeva and McGeer74 ) was associated with a large reduction in infection risk (OR, 0.24; 95 CI%, 0.14–0.42; P < .001) with an overall risk reduction of 17.1% (95% CI, 12.4%–20.1%; moderate certainty; Table 2; Appendix 4, Fig. 8 online). Compared to no H1N1 vaccine, H1N1 vaccine was strongly protective during the H1N1 pandemic (OR, 0.10; 95% CI, 0.04–0.22; P < .001) (moderate certainty; Appendix 4, Fig. 9 online). Reference Zhang, Seale and Yang51,Reference Costa, Silva, Tavares and Nienhaus57,Reference Kuster, Coleman and Raboud60
Compared to control (no involvement in intubation procedures), involvement in intubation (8 studies Reference Chatterjee, Anand and Singh32,Reference Heinzerling, Stuckey and Scheuer38,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Pei, Gao and Yang67,Reference Chen, Ling and Lu68,Reference Liu, Tang and Fang70,Reference Loeb, McGeer and Henry71,Reference Raboud, Shigayeva and McGeer74 ) was associated with a significant increase in infection risk (OR, 4.72; 95% CI, 2.71–8.24; P < .001) (57.3% in intubation vs 22.1% in no intubation; moderate certainty) (Table 2; Appendix 4, Fig. 10 online). Across 19 studies, Reference Chatterjee, Anand and Singh32,Reference Ran, Chen, Wang, Wu, Zhang and Tan36,Reference Heinzerling, Stuckey and Scheuer38,Reference Zhang, Seale and Yang51,Reference Kuster, Coleman and Raboud60,Reference Teleman, Boudville, Heng, Zhu and Leo63,Reference Pei, Gao and Yang67–Reference Loeb, McGeer and Henry71,Reference Raboud, Shigayeva and McGeer74 a composite measure of AGMPs was associated with significant increased risk of infection (OR, 2.42; 95% CI, 1.53–3.82; P < .001) (41.5% in AGMPs vs 22.7% in no AGMPs; moderate certainty; Fig. 3; Table 2). On subgroup analysis, significantly increased odds of infection were only seen with SARS (OR, 2.95; 95% CI, 1.68–5.18; P < .001) and not for COVID-19 or H1N1 (Fig. 3). Meta-regression analysis, including covariates of designated status (designated center vs unidentified center), IPAC measures (implemented vs unimplemented or undefined), AGMP type (intubation vs other AGMPs), ICU versus non-ICU, and virus type was performed (τ2 = 0.2428; I Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 = 73%; R Reference Mondelli, Colaneri, Seminari, Baldanti and Bruno2 = 0.61) (Appendix 5, Table 2 online). The rate of infection associated with performing AGMPs was a significantly lower in designated facilities compared than in those not identified as such (OR, −1.30; 95% CI, −2.52 to −0.08; P = .037). A higher rate of infection was associated with intubation compared to other AGMPs (OR, 1.04; 95% CI, 0.30–1.77; P = .006) (Appendix 5, Fig. 3 online).
Fig. 3. Forest plot of random effect meta-analysis of the association of aerosol-generating medical procedures (AGMPs) on infection in HCWs by virus type. AGMPs include endotracheal intubation, chest compressions, and other airway manipulations.
Summary odds ratios for meta-analyzed risk factors are reported in Figure 4. To emphasize the meta-analysed effect estimates of risk factors with greater robustness, additional meta-analysis with 99% confidence intervals were conducted. In this analysis, risk factors with effect estimates that persisted toward significant effect included frontline HCW, gloves, surgical masks, N95 masks, face protection, IPAC training, H1N vaccination, intubation, and participation in AGMPs (Appendix 6 online).
Fig. 4. Forest plot of all the summary odds ratios for meta-analysed risk factors. *Represents the overall odds ratios for meta-analysed risk factors on healthcare worker infection during all included viral respiratory pandemics. Comparator groups: intubation versus no intubation; AGMP versus no AGMP; frontline HCW versus non-frontline HCW; physician versus nurse; surgical mask versus no surgical mask; N95 mask versus no N95 mask; IPAC training versus no IPAC training; hand hygiene versus no hand hygiene; gowns versus no gowns; gloves versus no gloves; face protection versus no face protection.
Discussion
This systematic review and meta-regression analysis provides a comprehensive summary of occupational risk factors for HCW infection during viral respiratory pandemics. Our findings suggest that compared to nonfrontline HCWs, frontline HCWs are at significantly increased risk of infection during an outbreak (low certainty). Use of gloves, gowns, surgical masks, N95 respirators, and face protection, as well as receiving IPAC training were each associated with large reductions in infection (moderate certainty). Compared to other AGMPs, endotracheal intubation of patients with coronaviruses SARS-CoV-1 and SARS-CoV-2 was associated with a very large increase in the HCW infection rate (moderate certainty). Meta-regression analysis revealed that the availability of isolation wards was protective from infection among frontline HCWs and those performing AGMPs.
The safety of HCWs is paramount for many reasons, including the facilitation of continuous patient care, prevention of virus infection for themselves and also spread to other patients, as well as an ethical duty to protect those who put themselves on the frontline to treat others. The results of our review demonstrate the efficacy of well-known measures, such as PPE adherence and IPAC training, against viral respiratory pathogens that have stood as pillars of infection prevention and control in healthcare settings. The delivery of adequate IPAC training also poses its own barriers, including constantly changing guidelines, poor communication and enforcement of guidelines, and increased workload and fatigue in HCWs, which may be heightened during a pandemic lasting many months. Reference Houghton, Meskell and Delaney86 Thus, despite the novelty of SARS-CoV-2, it is likely that interventions long-practiced in acute-care sites across the globe are adequate to protect frontline staff against the virus. Reference Ellingson, Haas and Aiello87
Our findings regarding the protective effects of PPE use and increased transmission risk associated with AGMPs are generally consistent with results from previous reviews in the HCW population. Reference Chu, Akl and Duda13,Reference Tran, Cimon, Severn, Pessoa-Silva and Conly88–Reference Chou, Dana, Buckley, Selph, Fu and Totten91 A recent rapid review reported that in healthcare settings, risk for infection with SARS-CoV-1 was likely decreased with mask use versus no mask use and possibly decreased with N95 versus surgical mask use, with uncertain applicability to SARS-CoV-2 due to lack of direct evidence, This finding is generally consistent with our report relating to mask effectiveness and SARS-CoV-2. Reference Chou, Dana, Jungbauer, Weeks and McDonagh92 Of the 3 studies reporting a significant increase in risk for involvement in AGMPs, these procedures included endotracheal intubation and nebulization therapy with inconsistent reports of PPE use during the procedure. Reference Kuster, Coleman and Raboud60,Reference Chen, Ling and Lu68,Reference Liu, Tang and Fang70 Critically, none of these 3 studies addressed whether proper PPE was worn by personnel during these procedures, including use of N95 respirators. Based on these and other findings, national guidelines therefore universally recommend N95 respirators during AGMPs performed on patients with COVID-19. 93–Reference Chughtai, Seale, Islam, Owais and Macintyre96
The strengths of this study are that it identified a multitude of different factors relating to infection risk during previous respiratory viral epidemics representative worldwide through stringent methodology of data synthesis. Nearly all included studies met the WHO criteria for confirmed positive cases for each respective disease, ensuring the accuracy of cases and controls (Appendix 7 online). Our review highlights respiratory viruses with transmission profiles and reproductive numbers comparable to SARS-CoV-2, thereby increasing the generalizability of our findings and their applicability to the ongoing pandemic, distinct from previous reviews. Reference Verbeek, Rajamaki and Ijaz90,Reference Bartoszko, Farooqi, Alhazzani and Loeb97,Reference Long, Hu and Liu98 Finally, we used the GRADE approach to facilitate transparent recommendations and interpretations of the data. Reference Guyatt, Oxman and Akl25
Although stringent methods were adhered to, limitations were inherent in the current review. First, randomized trials were lacking due to the inherent ethical risk of restricting protective measures during an emerging epidemic. Most studies were of retrospectively design, potentially leading to selection and measurement biases and failure to match for potential confounding variables such as age, sex, and baseline comorbidities. Reference Chou, Dana, Buckley, Selph, Fu and Totten91 We also observed also heterogeneity introduced in the meta-analysis of many unique viral pathogens, each with different epidemiological profiles. Furthermore, the differences in global impact of the various pathogens (8,098 worldwide SARS-CoV-1 cases versus 56 million worldwide SARS-CoV-2 cases and increasing) introduced heterogeneity in meta-analyzed risk factors, potentially reducing the certainty of evidence for certain findings. 99,100 We conducted a pathogen-specific stratified meta-analysis to address these differences. However, few individual patient factors were reported (eg, ethnicity, sociodemographic factors, and comorbidity status) that likely influence HCW susceptibility to infection. Reference Cook101 Emerging literature suggests that black, Asian, and minority ethnic individuals are at increased risk of SARS-CoV-2 infection, with worse clinical outcomes. Reference Pan, Sze and Minhas102 Heterogeneity was observed in classifying the various risk factors. Few studies have explored the role of HCW-to-HCW transmission of pathogens, which has been associated with an increased risk of SARS-CoV-2 transmission in HCW without adherence to medical mask use in break rooms and during meals. Reference Çelebi, Pişkin and Çelik Bekleviç9 Moreover, data on compliance with hand hygiene or proper donning and doffing technique and staff surveillance strategies were limited, and both of these factors have been shown to be critical in reducing the infection risk. Reference Verbeek, Rajamaki and Ijaz90,Reference Phan, Maita and Mortiz103,Reference Wee, Sim and Conceicao104 These limitations were addressed by conducting meta-regression, controlling for virus type, and various covariates, and thereby adjusted estimates provide a conservative assessment of the risk to HCWs. The protective effects of each individual PPE item may be confounded by the reality that PPE is usually worn in bundles (eg, mask with face shield, gloves, and gown) and therefore may not reflect the true effect estimates of each PPE item, and these protective effects may be additive in when adhering to PPE bundles. Lastly, restriction of articles to the English language, to produce a timely review, may have excluded potentially relevant studies.
Amid the evolving COVID-19 pandemic, rapidly released research has attempted to answer many questions regarding the safety of HCWs caring for patients with COVID-19. Our review has shown that some key questions remain to be answered, including efforts to report detailed data for ethnicity, sociodemographic factors and comorbidity status, and direct head-to-head comparison of N95 respirators and surgical masks in the routine care of patients with COVID-19, a topic which has yet to be directly addressed by current evidence. Reference Long, Hu and Liu98,Reference Cook101,Reference Bartoszko, Farooqi, Alhazzani and Loeb105
In conclusion, this systematic review and meta-analysis synthesizes the current evidence for the risk of infection among HCWs in a viral respiratory outbreak and draws attention to useful protective strategies while caring for patients, especially for frontline HCWs performing risk-prone exposures. IPAC measures should be instituted, preferably in dedicated settings, to protect frontline HCWs during current and future waves of respiratory virus pandemics.
Acknowledgments
Financial support
Support was provided solely from institutional and/or departmental sources. MS is supported by the Canadian Anesthesiologists Society Career Scientist Award, as well as the Merit Awards Program from the Department of Anesthesia at the University of Toronto.
Conflict of interests
M.S. serves on the medical advisory board of the Hypersomnia Foundation on a voluntary basis. The remaining authors declare no competing interests.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2021.18
Open access
