The protocol for this evidence synthesis has been registered with the International Prospective Register of Systematic Reviews (PROSPERO; Registration number: CRD42020204174). The protocol was developed following PRISMA-P+ (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols) checklist (4) – additional file 2.
The question under review and PECOS framework
The study objective is based on the elements of PECOS (P: Population, E: Exposure, C: Comparator, O: Outcome and S: Setting). The Population is healthcare workers; Exposure is SARS-CoV-2/COVID-19; Comparator is the general population; the primary Outcome is COVID-19 infection, illness or deaths. The secondary outcomes shall be the access to COVID19 testing, access to COVID19 treatment, mental health co-morbidities related to COVID19, and risk factors for COVID19 death like pre-existing NCDs.
Study designs that will be considered are predominantly observational including cross-sectional, case-control, cohort studies, as well as systematic reviews and meta-analyses. Cross sectional studies and systematic reviews of cross sectional–studies will help in establishing the prevalence of COVID-19 infection, illness and the case fatality rate among healthcare workers. Also, case-control and cohort studies will help in establishing the case-fatality rate of COVID-19 among healthcare workers in LMICs. In addition, the descriptive and analytic cross sectional, case-control, and cohort studies plus systematic reviews of the same designs will be used for descriptive analysis of the nature of illnesses and establishing risk factors for deaths from COVID-19 among healthcare workers. Any of these observational study designs will be used to provide information on access to COVID-19 testing, access to treatment, and mental health outcomes.
Eligibility Criteria
Inclusion criteria
The period covered will be from December 2019 to date. Eligible studies should report at least one outcome of infection, illness and/or death due to COVID-19. Studies will be limited to those conducted among adult healthcare workers, above the age of 18 years, working in healthcare settings in LMICs during the COVID-19 pandemic. The review will also include studies conducted among the general population with healthcare workers as a sub group if results are stratified. Primary observational studies with cross sectional or prospective research designs, case control studies, systematic reviews, and meta-analyses shall be included. Only studies published after December 2019 will be included in the review.
Exclusion criteria
Case reports and series of single case studies, programme reports, conference abstracts, author’s opinions and editorials shall not be included in this review. Studies not published in the English language shall be excluded as those full texts that are inaccessible.
Information sources
Data sources for the review will include PubMed (preliminary search done on June 30th, 2020), MEDLINE Ovid and Google Scholar (July 9th, 2020), COVID-END, and the Cochrane library (September 15th, 2020). Websites of health and medical related organizations (World Health Organization (WHO), Ministry of Health, Uganda (MoH), and Medscape.
Search terms
The search terms used for this review included those for COVID-19 like Coronavirus, novel Coronavirus, Coronavirus Infection, Coronavirus disease 2019, Corona virus disease 2019, Coronavirus disease-19, 2019 novel coronavirus, 2019 novel coronavirus disease, 2019 novel coronavirus infection, novel Respiratory 2019 Coronavirus, 2019-nCoV infection, 2019-nCoV disease, COVID-19, COVID19, COVID-19 virus infection, COVID-19 pandemic, SARS-Cov-2, SARS-CoV-2 infection, Wuhan pneumonia. Terms for healthcare workers were tested in the pilot Health worker, healthcare worker, health professional, healthcare professional, nurses, doctors, ambulance driver, physician, laboratory technician, and the LMIC filter by Cochrane was used to filter articles in all Low- and middle-income countries.
Search strategy
An experienced information scientist (AAK) will conduct article search. Medical Subject Headings (MeSH) will be used in article search. The unique terms will be combined using Boolean operators “OR”, “AND” or “NOT”. Targeted search for additional relevant articles for this review will be done manually from references of included articles, and grey literature will be considered. The pilot PubMed search string is attached as an additional file 3.
Article screening
All the articles identified from the search will be exported into Endnote and duplicates removed. Article screening will be done using a priori criteria. BAK, RNN and RNW will independently screen articles following the PRISMA flow. Any disagreements between the reviewers will be resolved through consensus. A tiebreaker will resolve further disagreements (OM, EAO).
Data abstraction and management
A data abstraction tool will be developed in Excel spreadsheet 2007 (Microsoft Corporation, WA, USA, www.microsoft.com). This tool will capture administrative information as author, title, citation, country, ethical approval, funding, year of publication; methodological data such as study design, sample size, COVID-19 test confirmation method, profession, type of healthcare setting; and outcomes such as number infected, mortality, duration of illness, nature of care for example, need for Intensive Care Unit (ICU) or not, prevalence and case fatality. Ascertainment of infection as an outcome of interest will be through WHO approved laboratory tests for COVID-19, whereas ascertainment of illness or death will be through verbal autopsy or clinical assessment reports. Three reviewers BAK, RNW and RNN will do abstraction.
Risk of bias assessment
This evidence synthesis shall undertake critical quality control assessment at every stage of the review. Biases such as selection bias, outcome assessment bias, attrition bias and analysis bias are expected within the included primary studies. Quality assessment will be done using the New Castle Ottawa tools for Case-control as well as for Observational Cohort and Cross-Sectional Studies to critically appraise the selected studies (5). The AMSTAR-2 (A Measurement Tool to Assess Systematic Reviews 2) Systematic Review Critical Appraisal Tool (6) shall be used to assess quality of the included systematic reviews. The overall quality of the body of evidence will be graded using the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) framework (8). The immediate output will be a rapid evidence brief for policy in line with the SURE Guidelines (Supporting the Use of Research Evidence), which will inform a results deliberation/policy dialogue (9).
Publication bias
Publication bias will be examined using a funnel plot while using the asymmetry of the plots to detect likelihood of publication bias among included articles (19).
Data synthesis and analysis
Synthesis will be by study type. Abstracted data from the included articles will be exported to an open access review management software or EPPI-Reviewer (Evidence for Policy and Practice Information) for analysis. Results from the analysis will be presented as structured narratives or summaries in data tables and graphs where applicable. Measures of central tendencies (means, medians) and frequencies will be used for descriptive statistics. Random effects model, which assumes a hierarchical linear model will be used in meta-analysis, if appropriate.
Heterogeneity and sensitivity analysis
The level of heterogeneity will be assessed using the Higgins I2 statistic. It will be used to indicate percentage (%) heterogeneity that can be attributed to between-study variance. Interpretation: I2 = 25% (low heterogeneity), I2 = 50% (moderate heterogeneity), I2 = 75% (high heterogeneity). The different study designs will be analyzed using sub-group analysis. For the studies that will show dichotomous outcomes, the DerSimonian and Laird random effects method shall be used in meta-analysis.
Meta-analysis will be conducted to establish aggregate or pooled estimates of the review outcomes from the single studies. In accordance with the PRISMA statement, studies will be combined for meta-analysis following heterogeneity analysis. In our study, based on Higgins I2 statistic, if the heterogeneity between studies is above 90%, it shall be considered substantial heterogeneity and thus a meta-analysis may not be opted for. Further, both random effects meta-analysis and fixed effects meta-analysis will be conducted and the two results will be compared.
Sensitivity analysis will be done taking into regard risk of bias assessment, heterogeneity and differences in health care workers cadres (nurses, doctors, non-clinical staff).