Background
Oral cholera vaccines (OCV) are effective means to complement classical cholera control efforts in resource-constrained settings [
1]; however, little is known about their acceptability in targeted communities. Community acceptance of a vaccine is the ultimate determinant of the effectiveness of any immunisation program once safe and efficacious vaccines have become available, and functional supply chains and tailored program delivery mechanisms have been established [
2,
3]. Clarification of vaccine acceptability prior to the introduction of a new vaccine has been stimulated by experiences with stalled immunisation campaigns in many countries [
4,
5].
Cholera is an acute and highly infectious diarrhoeal illness characterised by profuse watery diarrhoea and vomiting. It can lead to dehydration, shock and death within a few hours if untreated [
6]. Since 2000, cholera incidence has been increasing worldwide: There are an estimated 3–5 million annual cholera cases, leading to 100,000-120,000 deaths per year [
7]. Recent massive outbreaks and the failure to achieve a reduction of endemic and epidemic cholera in many regions of the world increased the awareness for a need to improve cholera control efforts. At the World Health Assembly in 2011 cholera was declared a global priority (WHA64.15), with a specific role for OCV use.
Despite many efforts, prevention through the provision of clean water and sanitation has not had the desired impact on cholera incidence [
8]. As a complementary measure to improve cholera control, the use of OCVs for the prevention of larger cholera outbreaks in areas where cholera is endemic was first recommended by the World Health Organization (WHO) in 2002[
9]. Two oral cholera vaccines are currently available and prequalified by the WHO [
10].
The African continent accounts for most of the reported global cholera morbidity and mortality [
11,
12], with one of the highest concentrations of outbreaks in the Democratic Republic of Congo (DRC) [
13,
14]. In DRC, the first epidemics were officially reported in the 1970s, primarily in the Eastern part of the country, which has been heavily affected by two consecutive wars in the last two decades. Several areas where cholera is endemic are potentially eligible for the introduction of an OCV as a complementary means for cholera control.
Potential barriers to vaccine acceptance
Over the last decades many well organised vaccination campaigns have been compromised by unexpectedly low participation [
2,
15]. In Sub-Saharan Africa three explanations have been suggested. First, information about vaccines and the disease it seeks to prevent may be insufficient at the population level, and science-based concepts of prevention may not be well understood [
3,
16‐
20]. In this situation the perceived need for prevention will be shaped by local perceptions of risk and illness [
21‐
23]. Second, access problems linked to transport costs and time constraints on the user side, and operational shortcomings of vaccination campaigns on the health system side limit uptake [
24‐
28]. Third, active resistance linked to political or historical conflicts may discourage use, as documented in Cameroon [
29], Nigeria [
19,
30], and Moçambique [
31,
32].
This study sought to elucidate local perceptions of cholera and the potential acceptance of an OCV in a remote rural site and in a small town in Katanga province, DRC, where cholera is endemic. It also aimed to clarify the role of socio-economic and gender differentials, local illness perceptions, and to consider the social and cultural implications of illness for anticipated vaccine acceptance.
Methods
Study sites
The study was conducted in a rural town (Kasenga) along the Luapula River and on a fishing island (Nkolé) in Lake Mweru, Katanga Province, DRC. Kasenga has approximately 27,000 inhabitants. Electricity, water and two mobile communication networks are available. Health facilities include a district hospital, several clinics and dispensaries, and private pharmacies. Nkolé, a remote rural settlement, is inhabited by approximately 7000 people. Only one rural health centre was functional. The nearest referral centre could be reached only by boat (1 hour by speedboat, several hours by paddling).
Study design
A mixed-methods approach combined cultural epidemiology and a qualitative rapid assessment. Cultural epidemiology integrates measurement and analysis of qualitative with complementary quantitative information from semi-structured interviews, and it aims at establishing the distribution of socio-culturally shaped illness perceptions and practices. In this study a semi-structured interview was developed and administered in a cross-sectional study to a random sample of 360 adult community respondents. It enabled us to obtain representative distributions of ideas pertinent to cholera in the absence of an outbreak, and to assess the anticipated acceptance of OCVs in the two communities.
We expected attitudes towards OCVs to be influenced by experiences during past cholera outbreaks. We therefore conducted twelve purposive in-depth interviews with affected persons and key informants who witnessed past outbreaks in order to better understand the current cholera-related local perceptions and practices. In addition, four site- and gender-separated focus group discussions were held with unaffected persons. Direct observation and informal conversation in the community (markets, restaurants, on the road) over the 3–4 weeks of data collection in each site complemented contextual information and common perceptions of sensitive issues such as witchcraft, or dissatisfaction with government healthcare providers. Data were collected by a team of four anthropologists.
Instruments
Different instruments were used for the cross-sectional study and the in-depth interviews and focus group discussions. The cultural epidemiology approach used for the cross-sectional study required a semi-structured interview catalogue, which produced narratives as well as quantifiable answers based on the explanatory model interview catalogue (EMIC) framework of cultural epidemiology [
33,
34]. Since unaffected adults were interviewed, the EMIC interview began with reading a vignette that described a person with cardinal symptoms of cholera. All questions referred to this hypothetical case of cholera from the vignette. The interview incorporated questions on the socioeconomic and socio-cultural context of cholera in terms of illness-related experience (somatic and psychosocial distress), perceived causes, treatment-seeking behaviour, and previous experiences with vaccines. The selection of categories was based on prior EMIC interviews as used in Zanzibar and Kenya for the study of OCV acceptability [
35,
36] and on consensus among local experts. Questions further addressed active demand for a vaccine and anticipated acceptance of an OCV. Different price levels were assumed to assess the priority assigned to a vaccine in view of competing needs (free vaccine; cost of USD 1; USD 5; USD 10.5). The maximum price corresponded with the Dukoral purchase price at the time of planning the study.
The EMIC interview follows a specific interviewing and coding procedure. Open questions are asked first, and answers are coded using a list of predefined categories. Subsequently, respondents were asked to identify the most important category among all the categories they had mentioned. This scoring facilitated analysis of the variation in cultural illness-related beliefs and practices, and of anticipated oral cholera vaccine acceptance and its determinants. Questions addressing anticipated stigma were assessed using Likert scales. The vignette and questions were translated into two local languages, i.e. ciBemba and Kiswahili.
Quantitative analysis of the EMIC interview proceeds with a deductive approach, as it requires the a priori definition of the possible answer categories. Narratives to the open EMIC questions complement the quantitative data, indicating various socio-cultural constructs. Although it enables the researcher to relate narratives to quantitative codes for specific respondents, the instrument is more limited for in-depth discussions of an emerging theme that would allow an inductive approach. We therefore conducted additional in-depth interviews to further contextualise the quantitative findings in their local background. Open questions addressed past experiences with cholera outbreaks. In this way links between the prominence of socio-cultural response categories and the framework of underlying social processes and economic forces could be addressed. Interviews were conducted and recorded in the local languages.
Sampling strategy, data collection and informed consent
Data were collected in August and September 2010 by locally recruited interviewers fluent in two local languages; they were trained for 10 days prior to the fieldwork. For the cross-sectional study, the random walk method was used to select households. Based on census data it was estimated that in Mwalimu, Kasenga, every 21st and in the village of Nkolé every 7th household had to be visited. Starting from a main place a random number between 1 and 10 was selected to identify the first household. Interviews were conducted with the head of the household or the spouse. Adults above 18 years of age who had lived in the residence for at least six months were included. If no eligible respondent was found, the next household was selected. For the interviews with key informants health professionals and local authorities were contacted and asked for consent to be interviewed. Health professionals were then asked to identify and contact persons with personal experience of cholera for an in-depth interview. Participants for focus groups were contacted in the same way and comprised non-affected persons who had however been living in the community during past outbreaks. Every participant signed an informed consent form prior to their interview. Ethical approval was obtained from the University of Kinshasa.
Data management
Interviews were recorded in original language. Narratives were then transcribed and translated into French by the interviewers using f4 V. 4.0. Coding for thematic content was done with MAXQDA 10. Categorical data from EMIC interviews was double entered in EpiInfo V. 3.5.1 and converted for statistical analysis in SAS V. 9.2.
Approach to analysis
The analysis was conducted in two stages triangulating different epistemological and methodological dispositions. Statistical analysis of the EMIC data provided distributions of illness-related experience, meaning, help-seeking behaviour, and attitudes towards prevention. It also enabled assessment of determinants of anticipated OCV acceptance. While this first part of analysis takes a deterministic stance and aims at answering testable questions, in a second part, narratives were coded and interpreted in order to better understand the local meaning of cholera and its prevention.
First the quantitative analysis is presented. Frequencies and prominence values (importance assigned to an item) of response categories for questions about the perceived causes of cholera, its prevention, treatment seeking, and the potential social impact of cholera, are presented graphically.
Anticipated OCV acceptance at different assumed cost levels (free, USD 1, USD 5, USD 10.5) was calculated. To examine socio-cultural determinants of anticipated acceptance of OCVs, categories answering the above questions about the perceived causes of cholera, its prevention, treatment seeking, and the potential social impact of cholera, are considered as potential explanatory variables. For response categories related to the same underlying concept, indices were calculated. The following items were selected for inclusion in a cumulative index: a hygiene index combined ‘hand washing’, ‘clean water’, ‘clean/safe food’, ‘safe garbage disposal’, ‘safe disposal of stool’, and ‘health education’. A social impact index combined ‘fears of being isolated’; ‘fears to infect others/blame’; and ‘interference of cholera with social relationships’ while items, which related to felt or internalized stigma such as ‘feeling shame’ were assessed separately. Cronbach Alpha was used to test internal reliability of the indices and sum scores were calculated.
Due to the near-universal anticipated acceptance of an OCV at no cost and a low variation at a low cost of USD 1 logistic regression could only be conducted for an assumed expenditure of USD 5 (medium) and USD 10.5 (high). For both outcomes (anticipated OCV acceptance in case of costs of USD 5, and anticipated OCV acceptance in case of costs of USD 10.5) univariable logistic regressions were conducted with SAS V. 9.2 for the following factors: cholera-related experiences (somatic and psychosocial distress and stigma), perceived causes of cholera, treatment-seeking behaviour, previous experiences with vaccines, and sociodemographic factors including sex, age, education and main source of income. Each response category was tested for interaction with sex and with site. Following the univariable analysis, every response category that related to the same question was included in a separate intermediate model (not shown) if either the p value for being associated with the dependant variable was <0.2, or if the p value for interaction with sex or site was <0.1. Intermediate models were additionally controlled for socio-economic covariates. From the intermediate models all determinants with p values <0.2 for main effects or p values <0.1 for interactions were included in a comprehensive multivariable model. Variables were retained in the comprehensive model if their p value was below 0.2.
While the comprehensive model allowed assessing the effect of each answer category and did not include the index variables, we were also interested in assessing a cumulative effect of the social impact and hygiene variables. Therefore the multivariable analysis was repeated, replacing the hygiene and social impact categories with the respective index. Based on the same criteria as for the individual answer categories a comprehensive model containing the indices (index models) was created for each outcome (acceptability at USD 5 and USD 10.5).
Approach to qualitative data analysis
Narratives from the open questions of the EMIC interview for the 360 respondents, the four focus group discussions and twelve in-depth interviews were transcribed and imported into MAXQDA. Narratives of the EMIC were initially grouped with reference to a specific question eliciting the narrative. The narratives were read several times and coded for thematic content. The codes were then grouped into larger thematic areas pertinent to cholera and vaccine acceptability to achieve theoretical saturation. Findings were then compared with the results from the quantitative analysis to elaborate and triangulate with reference to the framework of underlying social processes.
Discussion
Past experiences with cholera outbreaks created high awareness and fear of cholera in this remote area in South-Eastern DRC. Health education and the provision of clean water have curbed cholera incidence to some extent, but resources were lacking to maintain protected wells and boreholes and diarrhoeal illness continues to impinge on the population. This is the background to the high acceptability of an oral cholera vaccine in this cholera-endemic area. The high willingness to pay may somewhat overestimate respondents’ actual ability to pay (hypothetical bias) [
37]. Cost-related barriers have been repeatedly reported to negatively affect vaccine acceptance [
16,
24,
27,
38‐
40]. That costs may play a role in the study sites as well is supported by the fact that respondents who considered the loss of income in case of cholera a problem were
less likely to anticipate OCV acceptance if a higher cost would have to be met. Efforts to reach the most vulnerable people may therefore be required if equity in access is to be achieved. Factors linked to a vaccination campaign itself, such as opening hours, could not be assessed in absence of a campaign but may affect uptake as well, as a similar study conducted in Zanzibar showed [
41,
42].
Our results equally showed that respondents who feared the social impact of cholera, such as interference with social relationships, were less likely to anticipate OCV acceptance This may seem counter-intuitive at first sight because we would expect a person who fears the consequences of cholera to be more interested in prevention. But the fear of losing social support because of cholera needs to be understood in this context of broader social insecurity. In African settings access to resources is usually mediated by group membership, mainly the family, or clan. Where formal social security is lacking, group membership is often the only way for the poor to mobilise material support during a crisis such as illness [
43]. Investments in social networks are important coping strategies [
44]. The lower anticipated OCV acceptance of respondents who stress the social implications of cholera may reflect their material insecurity and weak social networks rather than their reluctance to use a vaccine. Prior research has come to similar conclusions. Cassell et al. (2006) reported from Gambia that mothers with a weak social network were less able to access childhood vaccination [
20].
In our study fishermen and their families were more likely to anticipate social and financial implications of cholera. In the unadjusted analysis they were also less likely to anticipate acceptance of an OCV at the highest cost, even if their monthly income was not lower as compared to others. Fishermen often live under difficult conditions. Already in the 1990s van Bergen had described a ‘poverty-complex’ of poor living conditions, frequent migration, disrupted social structures, and problems around alcohol consumption and sexual transactions providing a breeding ground for both HIV and cholera in the fishing camps of lake Mweru [
45]. An agency report from 2009 confirms that for example poverty-related fish-for-sex exchange in Kasenga and the surrounding fishing camps are locally perceived a problem and a cause of HIV in the area [
46]. HIV and poverty have been described to be a problem in fishing villages in other settings in the region as well [
47‐
49]. The lower acceptability of a vaccine at a high cost among fishermen in the unadjusted analysis is therefore likely to be explained through greater social and economic vulnerability in the multivariable models. The identification of vulnerable subgroups such as the fishermen and fish traders may help to improve equity in access to a vaccination campaign, because there is often a predisposition to neglect difficult-to-reach people out of logistic reasons especially if herd protection may already be achieved by a relatively low coverage as was shown to be the case for cholera [
50].
A process of social marginalization can be reinforced if a feared disease is thought to be the problem of a particular group of people. To explain the effects of both negative impact and concern about stigma of cholera on anticipated vaccine acceptance, one may consider reluctance to accept a vaccine as a kind of anticipatory coping with dreaded social exclusion and stigma by denying vulnerability to cholera. Negative attitudes of health professionals towards poor and marginalized people may additionally compromise acceptance of a vaccine. In other contexts a patronising and disrespectful treatment by health professionals has been shown to discourage especially poor and marginalised persons from using (childhood) vaccination services [
20,
27].
On the individual level, education and information about vaccination are known to influence vaccine acceptance [
51,
52]. In our study education influenced OCV acceptability at a lower price as well, while at a higher price material insecurity became more important. But cholera-related knowledge influenced vaccine acceptability irrespective of the price. In Zanzibar, where a similar study was conducted, individual-level barriers to OCV acceptance during a mass vaccination campaign included unawareness of the infectious pathways and symptoms of cholera as well [
42]. It was notable that local illness beliefs, like witchcraft or the breach of a taboo, were not associated with OCV acceptability in DRC. Other research in Ghana found as well that traditional practices had no influence on the readiness to use vaccines [
20]. Hence local traditional health beliefs and practices do not necessarily compete with science-based approaches. Nonetheless there is a possibility of underreporting of traditional beliefs and practices in our study, and delayed treatment-seeking because of witchcraft beliefs might still be a problem even if the acceptability of a vaccine is not compromised.
Besides traditional beliefs, Christian religious practices were common, without being mutually exclusive. In contrast to traditional beliefs, faith-based practices – namely praying for healing and the belief that God was responsible for cholera outbreaks – did show an effect on anticipated OCV acceptance, however in a contradictory way. The literature mentions an ambiguous influence of religion on health-related behaviour as well. Religiosity has been associated with a higher sense of control over one’s health [
53]. Conversely, prevention such as vaccinations may be perceived as interfering with God’s plans [
32]. Similarly ambiguous were the findings in this study. Prayers for healing, which are positively associated with OCV acceptability, may on the one hand indicate more active coping with health problems at the individual level and reflect active church membership and a strong social support network. On the other hand, respondents who mentioned God’s will to be at the origin of cholera were less likely to anticipate OCV acceptance. Persons who lack agency and autonomy in particular have been shown to ascribe to fatalistic positions [
54]. They may be less optimistic about their participation in a vaccination campaign, which too, may play a role for vaccine acceptability.
Limitations of the study
Household selection took place at random, and only two persons per site refused to participate in the study. Nonetheless, a potential underreporting or misreporting of local practices for treatment and prevention due to the negative connotation of these practices within the health system cannot be excluded (desirability bias). But witchcraft beliefs for example, which were common in the area, were not negatively associated with anticipated acceptance, suggesting that traditional medical beliefs and practices are not necessarily a barrier to vaccine acceptability. Nonetheless these results should be interpreted with caution. Witchcraft narratives were often linked to local authorities and politicians, suggesting limited trust in government or other authority. As the cross-sectional study did not include questions related to quality of care or trust in healthcare providers this could not be further investigated.
Conclusions
The town of Kasenga and the fishing island of Nkolé are prototypical sites in DRC where cholera has remained endemic, despite attempts to improve the sanitary infrastructure. Respondents demonstrated active interest in a vaccine and anticipated OCV acceptance was nearly universal. Nonetheless, there are several risks. In this setting, social insecurity and costs linked to vaccination may jeopardise the use of OCVs especially for socially marginalized individuals and groups, and social mobilization is likely to be key for the success of any vaccination campaign. Collaboration with local authorities who know the dynamics in their communities is therefore important. Furthermore, in order to better meet the expressed needs of the population for an improved water and sanitation infrastructure, any vaccination campaign should be combined with other cholera control activities, such as water and sanitation rehabilitation and the provision of health education.
Acknowledgements
We gratefully acknowledge the funding of this study by the Bill & Melinda Gates Foundation. We further thank Christian Schindler, Leticia Grize, Neisha Sundaram, Anicet Yemweni, Tarcisse Elongo and Laurent Akilimali for their scientific support and contributions to this study.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
CS, MGW and SM conceived of the study, with the support of BL, CM, CC, and RH. SM and CM conducted the study and analyzed the data, with the support of CS and MGW. SM drafted the manuscript. All authors commented on earlier drafts and read and approved the final manuscript.