This study was conducted between 8th and 19th June 2020 in Addis Ababa city administration, approximately three months after the first confirmed COVID-19 case was reported in Ethiopia on 13th March 2020. The city is both the capital and commercial hub of Ethiopia, including private and public commercial businesses. It is also a seat for the African Union and other international agencies. Administratively, the city is divided into 10 sub-cities and 116 districts. The projected population of the city was about 3.7 million in 2020, which accounted for 3.7% of the population of the country . COVID-19 has been the major problem in Addis Ababa, and about 71% of the confirmed cases in the country were identified from the city as of 21st June 2020 . During the period of the current study, the spread of COVID-19 in the city cumulatively increased from 1,625 on 8th June to 2,988 on 19th June, with an average of 114 cases per day. Overall, Addis Ababa reported about 66% (171,947) of the confirmed COVID-19 cases in the country as of 9th May 2021 .
The Regional State of Oromia is one of the 10 states in Ethiopia, with a projected population of about 38 million in 2020 . Oromia is the largest region in Ethiopia, and accounts for about 38% of the population. The landmass of Oromia stretches over the largest parts of eastern, southern, central and western parts of the country, and shares a boundary with almost every region except the Regional State of Tigray. The capital city of Oromia is the national capital, Addis Ababa (a.k.a. Finfinne). All national government offices including all ministry offices, the city and sub-city administration offices, the city’s sector offices and the Regional State of Oromia and its sector offices are located in Addis Ababa. Many people including government employees and business persons reside in the Oromia Special Zone Surrounding Finfinne and commute daily to the city. The State of Oromia is one of the regions highly affected by COVID-19, next to Addis Ababa city. As of 9th May 2021, a total of 36,817 confirmed COVID-19 cases were reported from Oromia, accounting for 14% of the national cases .
Study design and sample size
This study was an institution-based cross-sectional survey of government employees selected from 46 diverse public institutions located in Addis Ababa city administration. As it was not feasible to conduct a representative nationwide household or online survey during the study period, the researchers opted to use a self-administered survey. The expected prevalence of COVID-19 perceived risk was unknown at the time of the study and an estimate of 50% was used for the sample size calculation with a desired precision of 4%, 95% confidence level, a design effect of two and 30% non-response rate. Finally, the minimum total sample size required for this survey was 1,560 respondents. However, due to the growing concern of high non-response rate due to the fear of COVID-19 crisis, the sample size was increased to 1,730 after inflating it by about 10%.
Study population and sampling
The target population for this study comprised all employees of 46 government institutions or organizations located in Addis Ababa city, and the study population constituted all employee’s working in the selected institution at the time of the survey and willing to participate in the study. These included professionals, experts, technicians and support staff working at different hierarchies and divisions/directorates in the selected institution including higher and midlevel officials. The study institutions were purposively selected from government institutions located in Addis Ababa city administration, particularly from three government levels: (1) the Federal or National Government Ministries, (2) Addis Ababa city administration bureaus and sub-cities, and (3) Oromia Regional State bureaus located in Addis Ababa. These included 16 national offices (14 Ministry offices, a National Bank of Ethiopia, and Addis Ababa branch Commercial Bank of Ethiopia), 12 sector bureaus and six sub-city administration offices in Addis Ababa city administration, and 12 sector bureaus from Oromia Regional State [Additional file 1].
A systematic random sampling technique was used to finally select the study participants in each institution. At the first stage, the data collectors contacted the Human Resource Directorate of each institution to explain the purpose of the survey. They obtained information on the total number of employees, number of directorates and departments in the institution with their respective number of personnel. The total sample size was then proportionally distributed to the 46 selected institutions based on the relative size of their employees. The allocated sample size to the institution was further distributed to the directorates or departments proportional to the relative size of the employees. At the second stage, the survey team selected participants from the list of employees in the selected directorate or department using systematic random sampling or consecutive sampling procedures. Participants had to be the employees of that institution in order to participate in this study.
Survey instrument and data collection
The survey instruments were developed by the research team for the purpose of this survey. The questions were prepared in English after reviewing relevant literatures [30, 31], and the worry questions towards COVID-19 crisis were adapted from the World Health Organization (WHO) Cosmo protocol questionnaire . The final survey instrument was organized into three sections: (i) socio-demographics characteristics; (ii) worries about the COVID-19 crisis, and (iii) perceived risk of COVID-19 [Additional file 2]. The English survey questionnaire was translated into Amharic and Afan Oromo languages and back-translated into English by two independent experienced personnel in order to ensure consistency. Few minor revisions of the instruments were made. Various scales measuring the worry and perceived risk of COVID-19 have been recently developed [33–36]. Unfortunately, none of these scales were used because they were not yet available at the time of our study.
Trained data collectors with previous experience in field data collection distributed the standardized paper-based self-administered questionnaires in two local languages (Afan Oromo in the offices of the Oromia Regional State and Amharic otherwise). The data collectors visited the selected institution, approached each potential participant who met the enrollment criteria and invited them to join the study. After obtaining consent, they handed over the questionnaires to the study participants with a cover letter (consent form), introducing the study and explaining the purpose of the survey, instructions on how to complete the questionnaire, the confidentiality of individual responses and ethical safeguards, and researchers contact information for any questions the respondent might have. The respondents were free not to answer questions that they felt uncomfortable in answering. Participants completed the questionnaires themselves in the local languages. Individuals who declined to participate were excluded from the study.
The socio-demographic variables included gender, age, education level, category of institution, years of service in the institution, location of residence, the COVID-19 worry scale and perceived risk scale.
The worry questions towards the COVID-19 crisis were measured using a 12-item worry scale, each rated on a three-point response options 1=’don’t worry at all’, 2=’worry somehow’ and 3=’worry a lot’. Respondents were asked to indicate how worried they were for each of the following 12 questions: (1) losing someone I love, (2) health system being overwhelmed, (3) own mental health, (4) own physical health, (5) own loved ones health, (6) restricted liberty of movement as a result of COVID-19, (7) small companies running out of business, (8) economic recession in the country, (9) restricted access to food supplies, (10) becoming unemployed because of COVID-19, (11) not being able to pay my bills, and (12) not being able to visit people who depend on me because of COVID-19. Responses to the questions were summed to obtain the total worry score of the 12-items, ranging from 12 to 36, with a higher score indicating a higher degree of worry regarding the COVID-19 crisis. The Cronbach’s alpha coefficient of the reliability of scale used in this study had a higher internal consistency of 0.86 .
COVID-19 perceived risk
Perceived risk toward COVID-19 was measured with three-items examining the extent to which the participant thinks about COVID-19: (1) “How likely is your chance of acquiring coronavirus?”, (2) “How likely susceptible do you consider yourself to an infection with coronavirus?”, and (3) “How likely do you feel that your own personal health is at risk of COVID-19 due to your work or occupational characteristics?”. Responses for each question were rated on a 4-point Likert scale with higher scores indicating higher perceived risk: 1=’very unlikely, 2=’unlikely’, 3=’likely’, and 4 = very likely. The total perceived risk score was the sum of the three items, ranging from 3 to 12. A higher total score indicates a greater perceived risk of COVID-19. The perceived risk index had the Cronbach’s alpha coefficient of the internal consistency reliability of scale of 0.71. Cronbach’s alpha values > 0.7 were considered acceptable .
Data were entered into the Census and Survey Processing System (CSPro) version 7.2 statistical software database (U.S. Census Bureau and ICF Macro). All descriptive analyses were performed using the Statistical Package for Social Sciences (SPSS) version 23.0 for Windows (IBM Corp., Armonk, NY, USA), and two-level linear mixed-effects regression models were applied in STATA/SE version 14 for Windows (StataCorp., College Station, TX, USA). The main outcome variables were worries and perceived risk scores about COVID-19. The level-1 independent variables consisted of gender, age, years of service, educational status (diploma or lower, bachelor’s degree, and master’s degree or above), and reported chronic illness. The level-2 independent variables considered in the study included place of residence (Addis Ababa or other) and the group of institutions (national, Addis Ababa, and Oromia). The worry scores about the COVID-19 crisis were also used as the main independent variable for predicting the perceived risk score. The two continuous level-1 independent variables, age and service year of the respondents, were centered around their means in order to facilitate interpretation.
Descriptive statistics such as means, standard deviations, frequencies, percentages and cross-tabulations were used to summarize the data as appropriate. Exploratory analyses were conducted to evaluate the assumptions of normality, linearity, multicollinearity and interaction. Normality and linearity tests for the data of worry scores and perceived risk scores were checked through graphically. Both the data of the outcome variables were slightly skewed, but no transformation was undertaken since the requirement of the large sample size was met and modelling with and without transformation of the outcome variables produced results that did not differ substantially . Therefore, untransformed scores of the outcome variables were used in the regression analyses. The variance inflation factor (VIF) test was performed for the predictor variables, and no evidence of multicollinearity was detected in the regression model (all VIF values < 1.6) . In addition, interactions between level-1 independent variables (age, gender, service year, and education) were also evaluated, but no significant interactions were observed (all p-values > 0.05).
All variables were analyzed individually for an association with the outcome variables using bivariate analyses. In addition, bivariate correlations between COVID-19 worry and perceived risk scores were investigated using Pearson product-moment correlation coefficient. Finally, mixed-effects linear regression models were performed to estimate adjusted beta-coefficients (β) and associated 95% confidence intervals (CIs) as measures of the associations between the outcome variables (worry scores and perceived risk scores) and the independent variables, adjusted for clustering and potential confounders. The intercept-only model (a null model) was modelled at the first step, followed by a model with all level-1 predictors. At the last step, the full model with all level-1 and level-2 predictors were modelled. The full model was better than the one with the level-1 variables, and the results presented in this study were obtained from the full model. Maximum likelihood was used to estimate the different parameters. The level of significance was set at p < 0.05 for all tests.