2. Methodology
2.1 Quantitative survey
Data for Round One BSS–Ethiopia were collected from all ten regions and Dire Dawa AC between December 2001 and May 2002; on average, data collection lasted three months for each region.
2.1.1 Target groups
The BSS considered six target groups (including 10 subgroups), namely, the youth (both in-school and out-of-school), female sex workers (FSWs), transport workers (long-distance truck drivers, intercity bus drivers and minibus taxi drivers and their assistants), rural residents (farmers and pastoralists), factory workers and the uniformed services (ground force and air force). The justifications for focusing on these target groups were as follows:
- Young people are particularly vulnerable and are the key to the future course of the HIV epidemic. They are the essential focus for prevention messages in every sexual health program. Since most new infections are in young people, even modest changes in behavior will have significant impacts on the epidemic.
- Sex workers are considered to be one of the core groups driving the epidemic because of their high risk for HIV infection and exposure to multiple partners.
- Because transport workers are highly mobile and have contact with many people they are considered a vulnerable group for sexually transmitted infections (STIs) including HIV/AIDS.
- Over 80% of Ethiopia's population is rural, consisting mainly of farm households and pastoralists. Although the importance of this sector to national development is recognized, the state of the HIV/AIDS epidemic in this group is not well documented.
- Factory workers are an important target group for behavioral surveillance survey (BSS), primarily because of their importance to the national economy.
- Because of the importance of the uniformed services to national security, the size of the group and its potentially high-risk behavior, the uniformed services have been given high priority in national HIV/AIDS prevention and control programs.
2.1.2 Sample design
The BSS–Ethiopia used sampling strategies that are crucial to the measurement of trends over time. Some target groups were not easily accessible through conventional household or institutional sampling techniques. For most target groups, two-stage sample designs were used i.e. primary units (clusters) chosen at the first stage and individual respondents chosen at the second stage. When members of the target group were associated with sites in a fixed manner, conventional clusters were used; however, when the target groups were ‘floating’, time-location clusters and targeted snowball sampling were used.
The probability-sampling method was used for the selection of respondents. The sampling frames for the household surveys and institution-based surveys were simple and readily available. However, sampling frame development required preliminary qualitative research and some level of mapping (social and geographic mapping). The objectives of the BSS mapping were to define target groups and identify their locations, to estimate sizes of the potential target groups and systematically list ‘clusters’ of groups to serve as a sampling frame for the main survey. The mapping also aimed to provide information on potential target groups for immediate commencement of intervention activities by governmental and non-governmental agencies, which were planning to operate in the area. The BSS team used various methods for mapping, including interview of key informants, observation by walking through the community, review of records and reports, and discussion with experts and program managers involved in STI/HIV/AIDS control.
The following formula was used to determine sample size for all target groups.
Where
D = design effect;
P1 = the estimated proportion at the time of the first survey;
P2 = the proportion at some future date such that the quantity (P2 – P1) is the size of the magnitude of change it is desired to be able to detect;
P = (P1 + P2)/2;
2 = (P2 — P1)2
Z1-a = the z-score corresponding to the probability with which it is desired to be able to conclude that an observed change of size (P2 — P1) would not have occurred by chance; and
Z1-b = the z-score corresponding to the degree of confidence with which it is desired to be certain of detecting a change of size (P2 – P1) if one actually occurred.
a = 0.05 (Z1-a = 1.65) and b = 0.20 (Z1-b = 0.84)
To determine the sample size necessary to detect a change of 15 percentage points, for several different indicators, the initial value of (P1) was estimated at 50% (this was a conservative estimate, which yielded the biggest sample size). The design effect, D, was estimated at 2 because of the cluster design used to sample the target groups. The level of precision was set at 0.05.
2.1.3 Sample size and sampling procedure for target groups
A summary of sample size for each target group is shown by region in Annex 1.
A. Youth
The following definitions were given to in- and out-of-school youth.
i) In-school-youth (ISY) were unmarried, daytime high school students attending grades 9-12 or vocational training schools. ISY were 15 to 19 years old; students out of this age range were excluded from the survey.
ii) Out-of-school youth (OSY) were 15 to 24 years old. They did not attend day or night school, were not married, and were unemployed or employed informally. Some high school graduates were considered as OSY if they completed high school in or before the year 2000 (1992 EC) and fulfilled the other OSY specifications above. Since the age range (15-24 years) was too broad for a behavioral survey, this target group was subdivided into two narrower age groups, 15-19 and 20-24 years.
OSY (equal numbers of males and females) was surveyed in all ten regions and Dire Dawa Administrative Council (AC). Due to limited resources, ISY (equal numbers of males and females) was surveyed only in the Amhara, Oromia, Somali, Harari and Addis Ababa regions and the Dire Dawa AC.
Application of the ‘sample size’ formula yielded a sample size of 267 for each gender, and for each region and AC. Based on the information available from previous surveys, 45-55% of youth were sexually active. The higher percentage (55%) was used for the calculation of sample size for all youth. Based on this information and adjusting for estimated non-response rates of 25% (for OSY) and 10% (for ISY) the necessary sample sizes were 1220 (610 males and 610 females) for OSY and 1100 (550 males and 550 females) for ISY.
(a) Sampling procedure: in-school youth (ISY)
For ISY, a total of 6600 students were selected (1100 from each target region and AC). The sampling frames for selection of ISY were prepared in consultation with the Federal Ministry of Education, regional education bureaus and respective schools (to obtain details of classes and respective numbers of students in each grade). Probability proportional to size (PPS) was used to select classes at the first stage and then systematic sampling was applied to select students at the second stage.
(b) Sampling procedure: out-of-school youth (OSY)
For OSY, the number of individuals selected was 14,224; these were sampled from urban centers of each of the ten regions and Dire Dawa AC, and from an additional area in Borena (accessible because of existing non-governmental organization (NGO) program structures). The sampling frames for selection of OSY were prepared using the 1994 Census Report (CSA 1998).
Segmentation methods were used for larger regions and the preferred sampling frame for household surveys was the kind used by the Central Statistics Authority (CSA) census bureau, obtained from the most recent population census. This approach had several advantages, notably, the households had already been mapped and numbered within enumeration areas; moreover, the areas had population sizes associated with them that could be used as measures of size during sample selection, making control of the fieldwork easier. In most towns, enumeration areas corresponding with sub-districts (i.e. kebeles) were used as the sampling frame.
In cities such as Addis Ababa, Dire Dawa, Harar, Nazareth and Mekelle that had 20 or more kebeles, the segmentation method was applied. In order to provide the required number of kebeles (>20) in the Amhara region and the Southern Nations and Nationalities Peoples Region (SNNPR), the towns of Bahir Dar and Gondar, and Awassa and Dilla, respectively, were coupled so that the segmentation method could be applied.
In the smaller regions, i.e. Afar, Somali, Gambella and Beneshangul-Gumuz, the random walking method was applied after selecting a starting point.
B. Female sex workers (FSWs)
FSWs (15 to 49 years old) in seven urban centers (Addis Ababa, Bahir Dar, Nazareth, Awassa, Gambella, Dire Dawa and Liben-Borena) were targeted in this survey. These FSWs could be broadly divided into three categories:
i. Bar/hotel-based sex workers– women who were paid by bar/hotel owners as ‘barmaids’ but who also used the bar/hotel to pick up sex clients, and those who were not paid by the bar/hotel owners but visited to the bar/hotel to pick up sex clients.
ii. Home-based sex workers– women who worked from their homes where they were involved in commercial sex work with/without selling local alcoholic beverages.
iii. Street-based sex workers– women who went onto the streets in the evenings to pick up sex clients.
Using the ‘sample size’ formula, the required sample size for FSWs was 267 in each site. Since, by definition, all the FSWs had non-regular partners, it was not necessary to increase the sample size. A 25% non-response rate was assumed and thus, sample size for FSWs was 334 (i.e. 267 x 1.25) at each site; this was rounded up to 350. Accordingly, the total sample size required for sex workers was 2100 (i.e. 350 at each site). However, in Addis Ababa a larger number of FSWs was sampled as a baseline for planned interventions.
Sampling procedure
The choice of sites where the 350 FSWs were selected was not uniform because the cities included in the study contained different numbers of sites where FSWs were found. Working with relevant governmental organizations, NGOs and members of the target group in the different cities, a list of locations where FSWs congregated was established, including the approximate number of FSWs found in each site per day/night. This information helped the estimation of sample size by PPS. Once the ordered lists had been constructed, ‘time-location’ clusters were used to take into account the possibility that sex workers exhibiting different behaviors might work on different nights of the week. The expanded list of locations included ‘high activity’ and ‘low activity’ periods (days and times of each day), along with the number of sex workers that were typically found at each site on those nights.
A ‘quota’ approach was used to select sample respondents. This approach specified that a fixed (equal) number of interviews were conducted in each bar/hotel or site. The quota was determined based on the total number of FSWs selected, as well as the minimum number of FSWs estimated to be found in a particular bar/hotel or site. However, in bars where it was anticipated that only a few sex workers were present, all the sex workers were contacted. In contrast, when large numbers of sex workers were expected, a fixed number of sex workers were selected randomly. Take all and targeted snowballing methods were used to select home-based and street-based FSW respondents.
C. Farmers and pastoralists
For logistic purposes, the study examined farmers (males and females, 15-49 years old) amongst the farming populations of areas where HIV sentinel and demographic surveillance systems were in place (Butajira in SNNPR). Pastoralist (males and females, 15-49 years old) were selected from the Borena pastoral community in Oromia; this community was chosen because of its geographic location and the existing experience of Save the Children–USA in conducting surveys and services amongst the population.
Using the ’sample size’ formula, the required sample size for farmers and pastoralists was 267. However, based on the estimate that only 40% of them had non-regular partners, and recognizing that several of the survey indicators concerned only the subsample that had non-regular partners in the past 12 months, the sample size was increased to 665. Moreover, to compensate for an assumed non-response rate of approximately 25%, the sample size was increased to 850. Accordingly, the total sample size required for farmers and pastoralists was 1700 (i.e. 850 from each group).
Sampling procedure
A two-stage random sampling procedure was applied. The two stages were conducted separately for males and females. At the first stage, PPS was applied to select villages. At the second stage, the random walking method was used to select households.
D. Transport workers
As a sampling frame, the study examined three types of male transport workers (15 to 49 years old) namely, truckers, intercity bus drivers and minibus taxi drivers; each type included drivers and their assistants. Investigators contacted the Road Transport Authority in Addis Ababa to obtain a list of truck, bus and taxi owners’ associations in Ethiopia. Subsequently, the associations were contacted to obtain lists of drivers and their assistants.
Using the ‘sample size’ formula, the required sample size for each type of transport worker (truckers, intercity bus and minibus) was 267. However, on the basis of estimates indicating that 60% of them had non-regular partners, and recognizing that several of the indicators concerned only the subsample that had non-regular partners in the past 12 months, the sample size for each group was increased to 334. Moreover, to compensate for an expected non-response rate of approximately 25%, the sample size was increased further to 600. Accordingly, the total sample size required for transport workers was 1800 (i.e. 600 from each group). Nevertheless, Ethio-Djibouti corridor truckers and their assistants in Logia, Afar were over sampled, because of the logistical importance of this transport route to Ethiopia, so that data for this group would stand alone as a baseline for planned interventions.
Sampling procedure
For truckers and their assistants, a list consisting of five routes leading in and out of Addis Ababa was assembled. During the mapping process, attempts were made to estimate how many of the trucks would be stopped at checkpoints (namely, Kara (east), Burayou (west), Sululta (north), Alemgena (southwest) and Mojo (south) for trucks traveling in all direction across the country, and at the town of Logia, Afar Region, for truckers and assistants traveling to and from Djibouti). For intercity bus drivers and their assistants, a similar list consisting of five routes was assembled. During the mapping process, investigators obtained daily schedules that indicated the date/time of departure and the size of the intercity buses traveling to and from Addis Ababa at the central bus station in Addis Ababa.
Information was also obtained on the number and routes of minibus taxis in Addis Ababa; data were collected from key informants (including drivers, assistants, and minibus owners’ and taxi drivers’ associations) and to a large extent from observations made by the BSS team during the mapping procedure. Two-stage sampling was carried out for each category of transport workers. At the first stage, PPS was used to select taxi routes and time-interval selection was used at the second stage.
E. Factory workers
National or regional samples of factories and their workers were not surveyed because of resource limitations. Instead, one factory, the Akaki Textile Factory, was identified for the BSS and male/female workers (15 to 49 years old) were selected.
Using the ‘sample size’ formula and making the same assumptions as for transport workers, the sample size required for factory workers was 600 (proportional representation was used when deciding the number of males and females to be interviewed).
Sampling procedure
A list of eligible male and female factory workers (15 to 49 years old) was obtained from the pay roll. This list was used as a sampling frame to randomly (simple random sampling) select 600 factory workers.
F. Uniformed services
Men (15 to 49 years old) from two branches of the uniformed services were targeted: these were ground forces stationed in the eastern part of Ethiopia (Dire Dawa and Awash) and air force personnel stationed in Debre Zeit.
Using the same formula and making the same assumptions as for transport and factory workers, the sample size required for the uniformed services would have been 1200 (i.e. 600 from each group). However, subsequent to the planning stage, at the request of the Ministry of National Defense (MoND), the number of individuals surveyed was increased and consequently, was greater than the sample size necessary for the study.
For selection of men from the uniformed services, a two-stage random sampling procedure was applied. Lists of eligible personnel (15-49 years old) were obtained from the commanding officers in Debre Zeit and Harar. These lists were used as sampling frames to randomly select (simple random sampling) the respondents.
The study was designed and conducted by the MoND.
2.1.4 Data collection and analysis
Data were collected using a standardized pre-coded, pre-tested questionnaire. For each target group, the FHI core questionnaire for BSS was used. Definitions of BSS indicators are summarized in Annex 6. The questionnaire covered the following topics:
- Socio-demographic characteristics of the respondents;
- Sexual behavior;
- Knowledge and use of condoms;
- Knowledge about STIs and STI treatment-seeking behavior;
- Knowledge, attitudes and opinions about HIV and AIDS, including misconceptions;
- Stigma and discrimination;
- Perception of risk; and
- Exposure to HIV prevention interventions.
The questionnaire was adapted (to make it suitable for use in Ethiopia and acceptable to the target populations) and translated into the appropriate local languages.
Interviewers were selected from the 10 regions and one administrative council, the uniformed services, and Butajira and Borena for the farmers and pastoralists, respectively. The BSS team of supervisors selected the interviewers, with input from the regional HIV/AIDS secretariats. All interviewers had completed high school and had some previous experience of collecting survey data. Within each area, equal numbers of men and women were included in the team of interviewers.
Interviewers were given a one-week training course by the BSS investigators and supervisors, with technical assistance from FHI. The training included: objectives and design of the study; administrative issues (e.g. length of interview, number of interviews/day, reporting systems and salaries); details of the questionnaire; role play; and field practice. At the end of the training period, the questionnaire was pilot tested in Addis Ababa.
All problems recognized during the pilot testing of the questionnaire were addressed, including those relating to selection of respondents, feasibility of conducting and completing interviews with the selected respondents, and timing. During the pilot study, it was observed that the interviewers needed to work together with the respective local (kebele) representatives in order to identify respondents in the selected target groups. The kebele representatives acted as guides for each team and introduced the interviewers to the respondents during the survey.
The supervisors were graduates of public health and demography, and post-graduate students. They were trained for a week on the objectives, methods and instruments of the survey, including the use of mapping for developing sampling frames for each target group. Roles of the supervisors included: assistance with the mapping process; selection and training of interviewers; coordination and planning of day-to-day activities; contact with concerned officials and individuals; making spot checks during interviews; monitoring the performance of interviewers; editing of errors on completed questionnaires to maintain data quality; and management of survey finances.
The BSS investigators also visited the interviewers and their supervisors, and discussed problems that arose in each site during the data collection.
Data were entered and cleaned using the Epi-Info (Version 6; EPI6) statistical package. Various statistical packages (EPI6, SPSS and STATA) were used for data analysis.
2.2 Qualitative study
2.2.1 Purpose and objectives
The qualitative study was not a separate study but a part of the BSS that aimed to substantiate and complement the quantitative survey. The major purpose was to explain target groups’ perceptions, misconceptions, attitudes, behaviors and knowledge about HIV/AIDS. The objectives were: i) to explore knowledge gaps identified and discover reasons for some of the results obtained in the quantitative survey; ii) to triangulate the major findings of the quantitative survey; iii) to provide an additional angle for viewing the findings of the BSS and hence create an opportunity for initiating further research in some areas; and iv) to provide additional information for consideration in subsequent rounds of BSS.
Qualitative data were collected between August and September 2002, after quantitative data had been collected and most quantitative data analysis was complete.
2.2.2 Methods and target groups
Focus group discussions (FGD) and individual in-depth interviews (IDIs) were the major qualitative methods used. The target groups were the youth (ISY and OSY), factory workers, FSWs, farmers, pastoralists and transport workers (truckers, intercity bus drivers and minibus drivers) (see Annexes 5A to D for numbers and characteristics of the FGD and IDI participants).
2.2.3 Questions
The questions were formulated to address the following topics:
Knowledge
- Knowledge about modes of transmission of HIV infection, preventive methods, relationships of STIs with HIV/AIDS and voluntary counseling and testing (VCT);
- Misconceptions about HIV/AIDS, modes of transmission and condoms, and impacts of these misconceptions; and
- Influence of knowledge, education and religion on behavior related to HIV/AIDS.
Attitudes and behavior
- Stigma and attitudes towards people living with HIV/AIDS (PLWHA);
- Perceptions and practice/behavior related to the prevention of HIV infection;
- Risk groups for HIV infection and circumstances and factors facilitating engagement in unprotected sex; and
- Perceived risk of behaviors and behavioral change in communities.
Interventions
- Sources of information about HIV/AIDS, and usefulness and preferences by target group;
- Participation in anti-AIDS activities and clubs; and
- Suggestions of ways to avert the spread of HIV/AIDS.
2.2.4 Tools and procedures for data collection
Detailed guidelines on how to organize and conduct FGDs and IDIs were prepared and discussed with the supervisors. Discussions and the interviews were guided with semi-structured questionnaires, developed after a series of discussions with investigators and stakeholders following the national dissemination workshop on 24-25 June 2002. Supervisors of the BSS quantitative data collection and other similar surveys were involved in the qualitative study team.
Participants were selected randomly from the target groups described in the quantitative methodology, with consideration of age, gender, religion and educational level. The regional HIV/AIDS secretariats’ offices assisted in the selection and organization of the qualitative study, as it formed part of the BSS. Participants in the FGDs were not involved in the IDIs.
Moderators of the FGDs and IDIs were trained to facilitate discussions and group dynamics. Male and female moderators interviewed and discussed issues with male and female participants, respectively; this arrangement created favorable conditions for the discussions and interviews. Between 8 and 12 participants were included in each FGD. The discussions and interviews were tape-recorded after obtaining consent from the participants. FGDs and IDIs were conducted in places where there were no frequent interruptions.
All the FGDs and IDIs were transcribed into the local language, Amharic, before the data were analyzed.
2.2.5 Analysis and write up
Consultants amongst the BSS team of investigators produced the final write-up of the data. Tapes, full transcriptions, observation notes and summary sheets were sources for the write-up.
2.2.6 Presentation of results
Although, the two parts of the BSS are complimentary, the results of the qualitative study are presented separately from those of the quantitative study. This layout was chosen: to maintain integrity of the quantitative and qualitative methodologies; to retain the narrative qualities of the qualitative data; and to ensure that the results in each section corresponded clearly with the objectives of the quantitative and qualitative parts of the BSS. Some of the major qualitative findings are discussed with corresponding quantitative survey results in Section 5.
2.3 Ethical considerations
The BSS protocol was prepared in cooperation with the HIV/AIDS Prevention and Control Office (HAPCO). Investigators ensured that methodologies and procedures included were in accordance with Addis Ababa University (AAU) and the Ethiopian Science and Technology Commission (ESTC) rules governing health research. Ethical clearance was obtained from ESTC. The Family Health International (FHI) Protection of Human Subjects Committee also approved the study protocols for the target groups.
Participation of respondents in the survey was strictly on a voluntary basis. Measures were taken to ensure the respect, dignity and freedom of each individual participating in the study. During training of the interviewers, emphasis was placed on the importance of obtaining informed consent (orally) and the avoidance of any kind of coercion. Complete confidentiality for study participants was also emphasized. For example, names of respondents were not recorded anywhere on the questionnaires.
