Background
Vignettes are short stories about a hypothetical person, presented to participants during qualitative research (e.g. within an interview or group discussion) or quantitative research, to glean information about their own set of beliefs. They are usually developed by drawing from previous research or examples of situations which reflect the local context, creating a story that participants can relate to. Participants are typically asked to comment on how they think the character in the story would feel or act in the given situation, or what they would do themselves. As the focus is on a third person, vignettes can be advantageous in research on sensitive topics where the participant may not feel comfortable discussing their personal situation and may conceal the truth about their own actions or beliefs. They can also, through normalisation of the situation, encourage participants to reveal personal experiences when they feel comfortable to do so [
1‐
4].
Vignettes have traditionally been used in the developed world in (predominantly quantitative) research on psychology and potentially sensitive social and health issues such as sexual health, HIV, mental health, stigmatisation, violence, and in specific vulnerable populations such as children and drug users [
1,
2,
4‐
13]. Hughes, and Barter and Renolds reflected critically on the methodology of vignettes with reference to their own research and other studies conducted in the developed world, concluding that the technique can be a valuable research tool despite debates surrounding their use: primarily the extent to which vignette responses mirror social reality [
1,
4]. Studies from the developing world (including Africa) using vignettes have emerged more recently [
14‐
24], but none have critically examined the use of vignettes in such settings. These studies have mainly focussed on similar topics to those investigated traditionally in North America and Europe, such as sexual health, mental health and stigma, but also include areas such as malaria and public health campaigns.
Very little qualitative research about HIV services in this setting, particularly prevention of mother-to-child transmission (PMTCT) of HIV or drug adherence (in the push for universal testing and treatment), has used vignettes to elicit perspectives of patients (or providers) regarding service or drug use. The few examples include Varga and Brookes’ study in South Africa, based on the narrative research method of the World Health Organisation [
25]: vignettes were developed during workshops with ‘key informants’ and presented during focus groups and surveys with pregnant HIV-positive adolescents to investigate barriers to participation in PMTCT services [
24]. Bentley et al. also used vignettes to investigate perceptions of HIV-positive mothers regarding breastfeeding practices and nutrition in Malawi [
26]. Varga and Brookes discussed methodological implications of their approach, reflecting that adolescent mothers spoke more easily about their own experiences after discussing the story of another teenager, and suggesting that in-depth interviews (IDI) exploring personal experiences can be useful in verifying and understanding responses towards the vignette. However, neither paper evaluated the specific challenges nor advantages of applying vignettes in their setting, for example the extent to which respondents understood the directions they were given, or how well fieldworkers facilitated discussions or interviews containing vignettes.
Global commitments have been made to improve uptake of PMTCT services [
27] in view of the low coverage noted in many African countries [
28]. An emerging body of research is exploring reasons for low access and usage of PMTCT services: barriers include sensitive issues such as stigmatisation regarding HIV status, fear of disclosure to partners or other relatives, and psychological barriers including denial [
29].
The potential for reporting bias in studies on barriers to PMTCT service use in sub-Saharan Africa has been noted [
29], and in our study setting, under-reporting by women of other socially sensitive outcomes (e.g. number of sexual partners) was reported [
30]. We therefore expected that a number of HIV-positive women would not admit to difficulties they faced when accessing PMTCT services, or would feel uncomfortable discussing such issues during interviews. Vignettes could consequently be a valuable and under-used tool in PMTCT/HIV research and drug adherence more widely. They may also offer a contribution to the range of methods available to reduce the social desirability biases encountered with self-reporting of outcomes in HIV, sexual and reproductive health research [
31‐
33]. There is some discussion over whether responses to vignettes may also be socially desirable, particularly when respondents are asked how they themselves would act in the scenario presented. However, asking first how the fictional character would behave and why is thought to reduce the pressure to answer with socially desirable outcomes [
4]. In this paper we describe the development and application of a vignette to an investigation of barriers and facilitating factors to uptake of PMTCT services in rural Tanzania. Our objectives for using the method were 1) to create a comfortable environment for IDIs and encourage women to openly discuss difficulties they or acquaintances faced in using PMTCT or maternal and child health services, and 2) to generate data on barriers and facilitating factors to uptake of PMTCT services from the perspective of HIV-positive and HIV-negative mothers, fathers and relatives. We critique the successes and challenges associated with employing vignettes in this setting, in order to determine the feasibility and utility of using this technique in qualitative investigations more widely in sub-Saharan Africa.
Methods
Study purpose and context
The study fieldwork was conducted between May and June 2012 in Kisesa, a rural area in north-west Tanzania, to identify barriers and facilitating factors to the uptake of PMTCT services, and ways of overcoming the issues identified. Demographic surveillance and HIV sero-surveillance has been conducted in this community since 1994 [
34]. Four health facilities offer antenatal clinic (ANC) and PMTCT services in the community: a health centre in the trading centre (also including an HIV care and treatment centre), and 3 dispensaries in rural villages (providing an intermittent PMTCT service depending on availability of HIV test kits and prophylactic drugs).
Study procedures
A variety of qualitative methods were used, including participatory learning and action (PLA) group activities, and IDIs incorporating a vignette. Before commencement of the study, fieldworkers received one week of training on relevant research methods and the topic (PMTCT). Training emphasised the participatory element of the PLA activities, as fieldworkers had prior experience of and training in conducting interviews and facilitating focus group discussions, but less experience of leading participatory fieldwork. After familiarisation with the PLA protocol, fieldworkers practised the activities with volunteer participants. The protocol was revised after observing practice sessions and listening to feedback from fieldworkers, (to shorten or simplify some activities), and after conducting the first PLA activity.
Development of the vignette
The vignette was developed through PLA activities conducted with 3 groups of men and 3 groups of women from different residence areas, each group comprising 8–12 participants. Participants were selected from a sampling frame of men and women aged 15–60 who had at least 1 child. This selection was random, with the exception of a few female HIV-positive individuals (‘seeds’) who were purposively selected from the sampling frame by the principal investigator using the community HIV sero-surveillance data. Female groups included 1–5 HIV-positive ‘seeds’ (see Buzsa et al. for details of the seeded focus group method [
35]). Fieldworkers were unaware of the HIV status of all individuals on the recruitment lists and those participating in the activities. Each PLA was facilitated in Kiswahili (commonly spoken national language) by an experienced fieldworker of the same sex as participants. A second fieldworker took notes on the content of discussions, details of the role-play storyline and behaviours of characters, as well as general observations of the group dynamic. The majority of sessions were attended by the principal investigator. Activities were audio-recorded following informed consent from participants.
PLA activities included brainstorming and ranking of barriers, role-playing and group discussion (Table
1). Before the role-plays, fieldworkers facilitated a discussion to identify the central characters that would be involved in a woman’s pregnancy and delivery. Thereafter, the participants were instructed to invent a storyline of a (fictitious) woman who discovers she is HIV-positive at ANC, thinking of the issues that a real woman in their village would face and the decisions she would make when trying to use PMTCT services. Participants then acted the play to the facilitator and observers. De-briefing sessions with fieldworkers were conducted following each PLA activity, informing an initial analysis of emerging themes which was used together with PLA notes by the project investigator to draft the vignette.
Table 1
Outline of activities conducted during participatory learning and action group activities
1 (day 1) | Group discussion | Discussion focussing on knowledge of vertical transmission of HIV and the PMTCT programme |
2 (day 1) | PMTCT ‘journey’ | Arrangement of cards representing components of PMTCT services (ANC attendance; HIV testing; provision of antiretroviral drugs to mothers and infants; delivery in the health centre; infant feeding advice) |
3 (day 1) | Storyline and role-play | Character and storyline development, role-play of the story composed, followed by group discussion reflecting on the play |
4 (day 2) | Barriers brainstorm & ‘wall of challenges’ | Group discussion of barriers at each step of the PMTCT ‘journey’ and creation of cards representing each barrier; ranking of barriers by arranging barrier cards on a wall |
5 (day 2) | Hanging fruits tree | Brainstorming of possible solutions to overcome the barriers identified, with solutions represented on fruit shaped cards; fruits placed on a tree diagram, according to how easy or difficult to achieve |
To compose the vignette storyline, unifying and contrasting elements of the role-plays were identified. Discussions following the role-plays, during which facilitators discussed how realistic the storylines were, were then analysed to confirm unifying elements, or resolve differences between the stories. Themes emerging from other activities, particularly barriers deemed most important in the ranking exercise, were also considered. The final vignette also needed to be viable given the character’s profile, for example, to represent the issues that the character would face considering their residence, marital status or family circumstances. The aim was to present a story that was familiar to most participants (touching on personal experiences, or experiences of acquaintances in their community), but that also achieved the objective of making women feel comfortable to admit to any difficulties they faced (so, for example, a more extreme case of a woman failing to access several of the services was chosen). Overly emotional circumstances or events (e.g. teenage pregnancy or death of a baby) which might derail the interview were avoided.
Once developed, the vignette and associated questions were incorporated into an interview discussion guide, along with open-ended questions about the personal experiences of the respondent during pregnancy, delivery and infant feeding. As conceived in the original study design, fieldworkers then received an additional day of training on the concept and use of the vignette, including examples of other studies employing this technique [
24], and on confidentiality (particularly if participants disclosed their HIV status during the interviews). This additional training session was intended to give fieldworkers the chance to familiarise themselves with and discuss the vignette developed from the PLAs, and to ensure the associated methods were fresh in fieldworkers’ minds prior to commencing the interviews. Fieldworkers were asked to review the vignette, and comment on how well it reflected the role-plays and major themes identified from the PLAs (no amendments were suggested). They were instructed to probe for whether responses to the vignette (what participants thought the character in the story would do) reflected real life in their community. After training, fieldworkers practised the questionnaire among themselves and with volunteer participants.
Use of the vignette
Twenty-one IDIs with HIV-positive (n = 16) and HIV-negative mothers (n = 5) who had recently delivered a child (since 2009) were conducted in Kiswahili by the same fieldworkers that facilitated the PLAs. Mothers were recruited purposively for interview from the PLA activities (and had therefore not necessarily attended clinic-based services, n = 11), and from each of the 4 health facilities in Kisesa by clinic nurses (n = 10). On completion of the PLA activities, each participant was asked to come forward, separately, to receive their travel compensation (5000 Tanzanian shillings, or approximately 3 USD), and asked if they were interested in being contacted for personal interviews in the future. Facilitators only scheduled specific appointments for interview with selected HIV-positive and negative participants, based on coded lists prepared by the principal investigator using community surveillance data. Facilitators were unaware of the HIV status of participants at the time of recruitment. For the clinic-based recruitment, each nurse was asked to invite and schedule interview appointments for at least two HIV-positive women who were pregnant or had recently given birth, during private consultations with their clients at antenatal or child follow-up clinics. Researchers did not have access to any clinic data for the recruitment.
Three interviews with partners/relatives of HIV-infected mothers were also conducted: women who had disclosed their HIV status during the IDIs were asked if their male partner, or otherwise a female relative, could be contacted for interview.
The same vignette was used in all interviews, and was read out to participants. Interviews lasted between one and three hours, and were audio-recorded after obtaining consent from the participant.
Critical analysis
Critical analysis of the vignette methodology was guided by the following key questions:
1.
Was the vignette method developed and implemented as intended? This includes how well the vignette was developed for the study context, delivered by interviewers and received by participants, in order to assess the feasibility of the approach. To answer this evaluation question, we: (a) reflected on the successes and challenges in developing the vignette; and (b) assessed IDI transcripts for any difficulties in interpretation of the vignette by the participants or fieldworkers, including confusion, misunderstandings or delays during the vignette section of the interviews, and whether participants considered the final vignette to be realistic. In analysis of the transcripts (audio-recordings were transcribed verbatim, translated into English, and the resulting data managed using NVIVO 9), codes were created to capture the way participants responded to the vignettes, and how fieldworkers dealt with their answers.
2.
Did the vignette method achieve its intended objectives? To this end, we gauged from transcripts whether the vignette helped to: (a) make participants comfortable during the interview, e.g. to discuss their personal experiences with ANC/ PMTCT services and HIV status; and (b) generate useful findings (data) about barriers and facilitating factors to PMTCT uptake, analysed through a framework approach which included thematic analysis to develop the coding scheme for barriers to PMTCT service uptake. We considered data quantity and quality, including any difficulties in interpretation of the data during analysis.
Ethical approval
This study was approved by the Lake Zone ethical review board of Tanzania, the Tanzanian Medical Research Coordinating Committee, and by the London School of Hygiene and Tropical Medicine ethics committee.
Discussion
This is, to our knowledge, the first methodological paper to critically examine the development and use of vignettes in Africa, and one of few studies to apply this technique within the context of HIV research in Africa. Overall, the development of vignettes through participatory group work, and use of vignettes within IDIs by locally trained fieldworkers, was feasible and useful in this setting. We believe it could be a valuable tool for future qualitative research in the field of PMTCT and other health or social issues in Africa.
Storyline development and role-play was a practical way of generating ideas for the vignette, although a simplified, more structured approach was required. It is possible that the more open-ended approach used by Varga et al. in South Africa was feasible in their study because participants had a greater knowledge of PMTCT services and a higher level of education: respondents included health workers, and eligibility was based on having “experience and knowledge of the health issue”, while our study included rural community members with no experience of the programme. Alternatively, differences in the experience level of fieldworkers may explain the variation in success of this approach. While our fieldworkers were generally experienced in qualitative methods, including focus group discussions and IDIs, they had less experience of participatory methods and no experience of using vignettes. In addition, they had been involved in HIV research, but were less familiar with PMTCT specifically. Intensive training was provided, but more practical experience, including more time for practice and piloting prior to the fieldwork, may have been needed to better facilitate the storyline development and role-plays.
While our approach to creating the storylines was simpler for the facilitators and participants, decisions of what to include in the final vignette were not straightforward. However, triangulation with results from other activities and discussions during the PLAs facilitated this ‘merging’ process. It could also be reasoned that the final vignette does not need to represent the majority of women in the community, but should at least be a realistic example of some women, so that it can successfully be used to build discussion in the interviews and encourage women to admit to their own experiences.
Most interviewees appeared to understand the concept of the vignette. This may partially reflect the fact that roughly half of the interviewees had participated in the PLA activities used to generate the vignette, although a similar proportion of interviewees recruited by other means comprehended the task. Prior participation in the PLAs may also have affected responses to the vignette more generally, for example coloured by views of other PLA participants regarding PMTCT service use, though interviewees would only have recognised small elements of their role-play in the vignette. Believing the vignette was realistic or anticipating the next section of the vignette may also reflect comprehension of the story, while this also generally facilitated the discussions.
The minority of cases where the story had to be repeated, or instructions had to be clarified, may have been due primarily to unclear questioning by interviewers, language barriers (poor command of Kiswahili), general shyness, or lack of readiness for the interview, rather than difficulties with the vignette technique itself. Encouraging the participant to think of the character as a woman in their community was especially helpful in enabling them to grasp the concept. Allowing time for the participant to digest information in the vignette and to seek clarification before proceeding with any questions or discussions may also be beneficial, particularly with longer vignettes [
1]. A few participants misinterpreted questions about the vignette, thinking they were being asked what Flora
should do, while fieldworkers also found these unexpected responses confusing (discussed further with regard to data interpretation below).
Use of the vignette achieved the main objectives. Firstly, we hoped that the vignette would encourage participants, particularly HIV-positive women, to feel comfortable and freely discuss their own situation and any difficulties they faced in accessing maternal health or PMTCT services: several respondents offered examples from their own experience, or that of family or friends, or commented on what they would do in Flora’s situation. Renolds noted that respondents were encouraged to voice more extreme concerns when the story was real [
2]. While our vignette did not give a biographical account of one person, it was based on discussions with the community and their own stories, and was considered realistic by the majority of interviewees. This emphasises the utility of developing the vignette through participation of community members. While the expression of personal experiences was a benefit, interviewers occasionally struggled to deal with this and were reluctant to digress from the discussion guide and Flora’s perspective. Fieldworker training should therefore stress the importance of drawing out the respondent’s own experiences, before returning the conversation back to Flora. Ideally, transcripts should be analysed during the course of fieldwork to identify and deal with these issues immediately. The use of questions such as “does this really happen in your community?” was particularly effective in drawing out personal experiences.
An unusual and interesting feature of this study was the knowledge of participant HIV-status by the principal investigator, thus allowing exploration of whether the technique may have encouraged or hindered disclosure of positive results. Most women disclosed their positive status during the vignette or personal experiences section of the interview, although they may have disclosed their status regardless of whether or not the vignette was included. A few women disclosed their status to the interviewer before the vignette discussion, in which case the vignette may not have offered any extra benefit. Some participants chose not to disclose their status (or they had not received their results, or research testing results were false positives). Therefore it is possible that presenting the case of a woman diagnosed with HIV who faces difficulties accessing PMTCT care, is sometimes insufficient to encourage disclosure. However, in such situations, the vignette at least enables discussion of the topic in a non-threatening way in the third person. While other factors will influence disclosure during the interview (such as the interviewer, the environment in which the interview is conducted, the respondent’s disclosure history and their willingness to disclose to strangers), the majority of HIV-positive respondents disclosed their status during or following the vignette discussion, suggesting that the vignette may have contributed to creating a comfortable atmosphere for the interviewee.
It is worth noting that a few participants said they had ‘failed’ to answer questions, suggesting that they perhaps felt ‘tested’ by the questions. This has been described in vignette studies from North America and Europe, particularly when respondents felt the story outcomes differed from what they had anticipated [
1]. In order to avoid making participants feel nervous, it may therefore be important to reiterate that there are no ‘wrong’ answers in the introduction: an approach adopted in one study in Ghana [
16].
Secondly, the vignette facilitated the discussion of barriers to using PMTCT services by focussing on a third person: most participants spoke freely about potential challenges that Flora would face. This meant that barriers could be discussed in all the interviews, including those with HIV-negative mothers, partners and relatives who had no direct experience of PMTCT services, but who were able to contribute useful information based on experiences of acquaintances, or their own experiences of maternal and child health services (into which PMTCT services are usually integrated). This advantage has been noted previously in developed world studies [
2], and also contributed to boosting the quantity of data generated.
The direct comparison between the vignette and personal experiences section of the IDIs was another strength of this study, contributing to an understanding of the extent to which responses to the vignette (what Flora would do) reflected participants’ own actions and thus data quality. Participants’ responses to the vignette often appeared to mirror their own experiences. This suggests that the vignette can be a useful tool to capture (indirectly) the perceptions and actions of shy respondents, for example HIV-positive individuals who do not wish to reveal their status or personal experiences. It also suggests that vignettes may be a valuable method for reducing the social desirability biases associated with self-reporting in HIV and reproductive health research.
Some participants gave advice to Flora and stated what she
should or
must do, and it was only through further questioning that what she
would do, or likely difficulties that she would face, came to light, if at all. This may reflect the respondent’s own actions, or illustrate a social desirability bias (what they think they should have done themselves). This distinction between ‘belief’ and ‘action’ is important and is one of the most common problems reported in developed world studies when using vignettes and interpreting their findings [
1,
2]. However, Renolds and Finch argue that the process of the discussion is more important than the stated outcome/action, and that vignettes can still yield useful information, particularly when integrated with other methods such as interviews [
2,
3]. None-the-less, interviewers should be prepared for and probe further when ‘
should’ responses are given, to determine if the answer is realistic. While this was included in training for our fieldworkers, further emphasis and practice may be required.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AG designed the study, managed the fieldwork, analysed the data and wrote the manuscript. AG and GM conceived the idea for this paper. GM gave advice on fieldwork materials and procedures, and assisted with fieldworker training and de-briefings. IB advised on fieldwork materials and gave extensive feedback on early manuscript drafts. IB and GM helped interpret the findings. GB facilitated female PLA activities and interviews, and helped with recruitment. MU (Kisesa cohort study director) facilitated coordination of daily activities. MU and BZ (senior technical advisor for Kisesa cohort activities) provided overall guidance. All authors reviewed and agreed to the final version of this manuscript.