Study setting
Zanzibar consists of two major islands, Unguja (also named Zanzibar) and Pemba. They are situated in the Indian Ocean about 40–60 km off the coast of Tanzania and a few degrees south of the equator (Figure 1). In 1964, shortly after independence from the colonial powers, the archipelago and Tanganyika formed the United Republic of Tanzania. Zanzibar as a semiautonomous entity within Tanzania consists of five regions which are subdivided into ten districts, 50 constituencies and 296 communities (Shehias). The main islands cover approximately 2,557 km2 (Unguja: 1,651 km2, Pemba 906 km2). They are inhabited by a population of approximately 985,000 (in 2002) with Unguja having a share of 63% and Pemba of 37%. The inter-censal annual growth rates (1988–2002) varied from 2.1 to 4.5% [35]. Its mainly Islamic inhabitants are speaking Kiswahili. One-fourth of the population has not received any education and the primary education net enrolment ratio amounts to 77% [36].
The majority of the population (71%) is having access to piped water while a minority has to rely on drinking water from wells (27%) and other sources like street vendors, rainwater, spring water, and open water courses (2%) [36]. About half of the population (53%) has access to pit latrines, while more than one-fourth has no toilet facility (28%) and 12% – mainly in urban areas – are using a flush toilet.
The top three causes of admission to Zanzibar hospitals in 2007 were malaria (27.4% of all admissions), gastroenteritis (12.7%) and pneumonia (9.9%) [37]. The main causes of death were malaria (18.4%), hypertension (8.5%) and pneumonia (7.9%) with gastroenteritis ranking on fourth place (7.5%). The main sources of help consulted are primary health care units which are situated within four kilometres of households for over 90% of the population. Monthly mean per capita expenditure was TZS21,000 (ca. USD18) in 2004/5 with a 2.1% share for health-related expenditures [36]. Life expectancy at birth has risen from 47 years in 1988 to 57 years in 2002 [35].
Target areas
The SEB study will take place in the periurban Shehia of Chumbuni (population estimate: ca. 13,500) in Unguja and the rural Shehia of Mwambe (population estimate: ca. 8,500) in Pemba. These two Shehias represent core areas of the mass vaccination campaign and were selected based on epidemiological data collected from recent cholera outbreaks (Reyburn et al., unpublished data) [13]. The selection of Chumbuni and Mwambe as target areas for the SEB study was therefore based on epidemiological vulnerability criteria (i.e. attack rates) and not on socioeconomic status.
Methodological framework
The research questions will be answered by using both quantitative and qualitative methods suited for the different stakeholder levels. On levels I to III rapid assessment methods will be used [38, 39] while on level IV the cultural epidemiological approach using Explanatory Model Interview Catalogue (EMIC) interviews will be employed [40]. Focus group discussions (FGD) will also be conducted as preparation for EMIC interviews.
Level I to III: rapid assessments
Rapid assessments have been utilised to address various health issues including malaria, sexually transmitted diseases, diarrhoeal disease, water and sanitation, and nutrition to name a few. Rapid assessments are methodologically tied to the basic tenets of anthropological ethnographic research which emphasises the use of multiple sources of data to gain various perspectives on social phenomena.
The rapid assessments in this study will provide the meanings and experiences of how community leaders, health care providers and policy makers perceive the importance of cholera as a public health concern and the need and demand for a cholera vaccine. These data will be important for future public health implementation of an oral cholera vaccination programme.
Level IV: cultural epidemiology
Cultural epidemiology incorporates qualitative and quantitative methods of health research and uses culturally adapted EMIC interviews to elicit locally valid representations of illness-related experience, meaning and behaviour [41]. Like classical epidemiology, cultural epidemiological research can focus on descriptive, analytical or comparative questions with regard to control of a disease or other public health interests. Analytical studies in cultural epidemiology consider the impact local categories of distress, perceived causes and help seeking – similar to the epidemiological risk factors – are having on clinical or public health outcomes.
A first concept of this research approach was already proposed 20 years ago to the benefit of health planning within the context of diarrhoeal illnesses and ORS promotion [42]. An early validation of this approach was later done in central Thailand when EMIC interviews were used to describe local diarrhoeal illnesses for generating public health policy recommendations [43]. In this study, EMIC interviews will be used to describe the cultural epidemiology of cholera and shigellosis and to estimate determinants of OCV acceptance.
A repeated cross-sectional cultural epidemiological study will be conducted in the two target areas (Figure 2). After a preparatory phase, where focus group discussions will be held, a baseline survey using EMIC interviews (phase 1) will be done in a random sample from each Shehia. After the intervention new samples will be drawn randomly based on the vaccination status. The same instrument will then be used again but with minor changes to account for the post-intervention status of the communities (phase 2).
Focus group discussions
Focus group discussions will be used to describe the context of cholera and shigellosis in the target areas. Results from the FGD among community residents will help to guide the planned quantitative research, i.e. the FGD guideline follows the topics and items that will be elicited in the EMIC interviews described below.
EMIC interviews
The interview will enquire about the sociocultural context in terms of illness-related experience, meaning and help-seeking behaviour and about topics related to vaccination. Under experience, different aspects of the illness like physical symptoms and psycho-emotional and social problems and also financial issues that can have an impact on patients will be elicited. Under meaning, respondents will be questioned about their views and opinions on why and how they think one can get the illness with regard to various categories of perceived causes including biological, behaviour-related, social and traditional/magico-religious factors. To find out more about sources of help seeking, respondents will be asked to identify and assess all health care providers, i.e. locally available allopathic and traditional sources, patients suffering from cholera or shigellosis will likely consult. The possible options for self-treatment (at home, not at a health-care provider) will also be elicited.
Variables to be elicited will include:
Sociodemographic factors
Categories of diarrhoeal illnesses, name of condition
Perceived severity of cholera and shigellosis episodes
Perceived vulnerability of cholera and shigellosis
Illness experience: operationalised as patterns of distress (PD) related to cholera and shigellosis
Illness meaning: operationalised as perceived causes (PC) related to cholera and shigellosis
Illness behaviour: operationalised as help seeking (HS) within and outside the household related to cholera and shigellosis
Stigma (index) of cholera
Prior episodes of cholera
Prevention of cholera
Experience and perception of vaccinations
Barriers to vaccination (only phase 2)
Anticipated vaccination status (only phase 1)
Willingness-to-pay for OCV
Actual vaccination status, i.e. vaccinated/unvaccinated, not to be elicited but recorded during vaccination campaign
Sample size calculation for EMIC interviews
The following calculations are based on a 95% significance level and 80% power. The prominence means of the different categories of cultural epidemiological variables will be compared in bivariate analyses (by site, gender, anticipated vs. actual vaccination status) and tested for significant differences using the Wilcoxon rank-sum test. To detect a difference of 0.5 between prominence means with equal standard deviations of 1.5, a sample size of 164 for each group is required. The calculation is based on a two-sample t-test assuming a worst-case scenario, i.e. no underlying distribution in the data, which requires that the sample size derived from the t-test (n = 142) be divided by 0.864 which equals 164.4 [44]. Hence, the overall sample size (two groups per phase times two phases) will be 656, with a sample size of 328 per phase.
Data collection
Level I to III
Strategy
Rapid assessment individual interviews and focus groups will collect qualitative information on policy makers' considerations and need for information for planning the introduction of a cholera vaccination programme. This phase of the study will also involve data collection with health care providers and community leaders on their perceptions of cholera, its importance to community residents, and their interest in the OCV. This research will explore policy issues and processes, as well as barriers and enabling factors that are likely influences in the introduction of a vaccine against cholera. The descriptive data of the rapid assessment interviews and focus groups can be related to and compared with level IV data. The approximate duration of this phase of the study is four weeks. It will be conducted following the mass OCV campaign in order to ground perspectives on the acceptance of an OCV in real world contexts and experiences.
Instrument
In-depth interviews and group discussions are open-ended and semi-structured to both cover critical topics and allow the respondents to identify and discuss what is important to them regarding cholera and the use of a cholera vaccine. For each stakeholder level specific interview and discussion guidelines have been created, translated and orally back-translated to ensure their intelligibility and appropriateness. Two experienced field researchers will conduct the research either in English or Kiswahili depending on the respondents. Each session will last approximately one hour.
Sampling
The following groups will be selected purposively:
Policy makers (n = 10–20) including national and regional policy makers (e.g. from MoHSW, international non-governmental organisations);
Allopathic and traditional health care providers (n = 6–8 per island) working in the target areas and district hospitals;
Formal and informal local government and community leaders and teachers (n = 6–8 per island) from the target areas (e.g. Shehas, secular and Islamic teachers).
Level IV: preparatory phase
Strategy
Community residents will be approached on both islands and information collected in focus group discussions. Data will be collected to complete or expand the lists of variables of the different EMIC interview sections mentioned above. The approximate duration of this phase is two weeks with one team working per site. This phase must be completed before phase 1 starts.
Instrument
A FGD guideline was drafted which covers all the relevant issues of the EMIC interviews. Each team will consist of a moderator and a note taker.
Sampling
Sample: adults (≥ 18 years) from both communities;
Sampling frame: registers from Shehias adjacent to the target Shehias;
Sampling method: stratified purposive sample, by gender and age group (18–45 years, >45 years);
Exclusion criteria: belonging to other stakeholder level I to III;
Sample size: n = 48–64, eight FGD with each having ca. 6–8 respondents (4 FGD per site)
Level IV: phase 1 (pre-vaccination)
Strategy
Community residents will be approached on both islands and information collected through EMIC interviews. Data will be collected i) to clarify the local features of cholera and shigellosis; and ii) to identify social and cultural determinants of anticipated OCV acceptance. The approximate duration of this phase is about ten weeks. Three teams each per site will be working simultaneously. This phase must be completed before the social mobilisation and vaccination campaign starts.
Instrument
An EMIC interview for phase 1 was drafted and, once it has incorporated information from the preparatory phase (FGD), will be pilot tested and finalised before implementation. Since the general adult population – and not cases – will be interviewed, it was decided to introduce the topic to the interviewees by using slightly varying gender-specific vignettes which describe cardinal physical symptoms of cholera and shigellosis. This approach ensures that the interviewees will respond with regard to the clinically relevant conditions and not with regard to local notions of diarrhoeal illnesses labelled by the respective local terms for cholera and shigellosis.
A ten-day workshop will be conducted to train the fieldworkers in interview and data entry skills, to familiarise them with the instrument and to pilot it under field conditions. Each team will consist of an interviewer and a note taker. Narrative data will be translated and typed on a daily basis by the teams.
Sampling
Sample: adults (≥ 18 years) from both target areas;
Sampling frame: census and Geographic Information System databases;
Sampling method: stratified random sample with gender ratio 1:1;
Exclusion criteria: belonging to stakeholder level I to III;
Sample size: n = 164 per site, n = 328 per phase
Level IV: phase 2 (post-vaccination)
Strategy
Community residents will be approached on both islands and information collected through EMIC interviews. Data will be collected i) to clarify the local features of cholera and shigellosis after vaccination, ii) to compare these features with pre-vaccination data, iii) to identify social and cultural determinants of actual OCV acceptance, and iv) to identify barriers to vaccination. The approximate duration of this phase is about ten weeks. Three teams each per site will be working simultaneously. This phase will be implemented after the end of the vaccination campaign.
Instrument
The EMIC interview will be exactly the same as in phase 1 except that a special section will enquire about the experience with the vaccination campaign and reasons against getting vaccinated among both unvaccinated and vaccinated respondents.
The interview teams will attend a second week-long workshop where they can refresh their skills and where the new section of the interview will be explained. A second pilot testing phase will take place to adapt the instrument with regard to the new section.
Sampling
Sample: adults (≥ 18 years) from the target areas;
Sampling frame: vaccination campaign data (vaccinated people), census database after exclusion of vaccinated people (unvaccinated people);
Sampling method: equally stratified random sample among confirmed vaccinated (two doses of OCV) and unvaccinated (one to two doses) people with gender ratio 1:1;
Exclusion criteria: belonging to stakeholder level I to III, already interviewed in phase 1, member of household already interviewed;
Sample size: n = 164 per site, n = 328 per phase
Data management and analysis
Level I to III: rapid assessment data
The rapid assessment interviews and focus group discussions will generate lengthy textual material which will be tape-recorded, followed by transcription and translation. This information will be organised and analysed using the qualitative analysis software of Ethnograph 6.0. This programme will allow the data to be multiply coded, segmented and searched according to important and emerging variables and themes.
Level IV: qualitative data
Information from focus group discussions will be tape-recorded, followed by transcription and translation. Narratives from EMIC interviews will be transcribed verbatim and translated. Typing will be done in a word processor software while MAXQDA 2007 will be used for managing the textual data and to facilitate analysis regarding findings from quantitative data.
Level IV: quantitative data
Information will be double-entered and cleaned in Epi Info 3.4.3 software. Descriptive statistics and bi- and multivariate analyses will be computed with the statistical analysis programme Stata 10.
Regarding the cholera vignette, the variables related to illness experience (PD), meaning (PC) and help-seeking behaviour (HS) will be coded with a value of two after a spontaneous response, a value of one after a probed response and a value of zero for no response at all to reflect the response style. A value of three will be assigned to the summary variables (i.e. most troubling, most important and most useful). For each category of such a variable, a total prominence will be computed, ranging from zero to five. Thematically similar individual categories will be grouped under specific headings (e.g. "physical symptoms" among PD variables) to enable the analysis of broader concepts of experience, meaning and behaviour. Grouped frequencies will be computed by adding the maximum value based on the response style of each respondent for every single category falling under each group. Calculation of the grouped prominence will follow the same procedure as with the individual variables.
For questions related to the shigellosis vignette, only frequencies of the categories mentioned in relation to illness experience, meaning and help-seeking behaviour will be computed.
Tabulations of the above-mentioned variables will be done after each phase followed by descriptive, bivariate and multivariate analyses to address the research questions.
To test for significant differences (p ≤ 0.05) between two groups, the t-test will be used for normal data and the Wilcoxon rank-sum test for nonparametric data. The Kruskal Wallis test will be used for comparing more than two groups of a nonparametric variable. The Chi2 test and the Fisher's exact test will be applied for comparing two proportions. To compare paired data, i.e. variables between cholera and shigellosis vignettes, the McNemar's Chi2 test, the paired t-test and the paired Wilcoxon test will be used.
Determinants of anticipated and actual OCV acceptance (outcome variables) will be identified in a twofold approach. First, bivariate tests (see above) will be done to see which variables are related with the outcome variables. Then, variables having a suggestive bivariate relationship (p ≤ 0.3) with the outcome variables will be retained as potential explanatory variables for computing stepwise logistic regression models.
Narrative accounts will be used to clarify and substantiate the relationships identified in the above analyses.
Data safety and storage
A password-secured data management system will be established where the responsible researchers on both islands can up- and download qualitative and quantitative data. Electronic data can only be accessed with permission from the co-investigator(s). Raw data (filled-in forms, narratives) will be kept under appropriate climatic and safety conditions in the Public Health Laboratory, Pemba. The data will be destroyed two years after the completion of the study (hard copies by incineration and soft copies by digital erasure).
Translation of instruments and informed consent forms
The English versions of all the instruments and informed consent forms will go through cultural adaptation and translation into Kiswahili. Back-translation into English will ensure the validity of the translation and that no ethical alternation is introduced in the informed consent forms.