Context-acceptability theories: example of family planning interventions in five African countries

The study took place in sites of varying size ranging from 14 to 114 health facilities (hospitals, health centres, and health posts) and 24 routine outreach clinics in five countries. Each study site encompassed a project on integrated delivery of FP and childhood immunisation services led by a non-governmental organisation (NGO). The study sites were Adjohoun-Bonou-Dangbo district, Ouémé department, Benin; Assosa and Bambasi woredas, Benishangul-Gumuz region, Ethiopia; Garba Tulla, Isiolo county, and Pokot West and Pokot South, West Pokot county, Kenya; Mwanza, Blantyre, and Thyolo districts, Malawi; and Karamoja sub-region, Uganda.

Unmet need for FP¹ amongst married women varied across countries according to the latest Demographic and Health Surveys (DHS) at 33.7% in Ouémé department, Benin [21]; 21.1% in Benishangul-Gumuz region, Ethiopia [22]; 12.4% in Eastern region and 20.8% in Rift Valley region, Kenya [22]; 20.3% in Southern region, Malawi [23]; and 19.7% in Karamoja region, Uganda [24]. These data, however, mask intra-regional and intra-district variations in unmet need for FP.

All project interventions had the shared aim of increasing the use of MCMs amongst postpartum women by integrating the delivery of FP services with childhood immunisations through the Expanded Programme on Immunisation (EPI). The components of the intervention varied by country based on the structure and function of the health system and on the socio-cultural context. Generally, the interventions involved a training component for health facility staff and community health workers; information, education, and communication (IEC) at the community level; and referrals from immunisation to FP services. Delivery points for FP services were health facilities (Benin, Kenya, Uganda), health posts and home visits (Ethiopia), and outreach clinics (Malawi). Where FP services were delivered at health facilities, both FP counselling and administration of MCMs were conducted by midwives and/or nurses. FP counselling and MCM administration at health posts, outreach clinics, and via home visits were through different cadres of community health workers, health extension workers in Ethiopia, and health surveillance assistants in Malawi. In addition, peer influencers were active at the community level to mobilise women to attend FP and immunisation services in Benin, Ethiopia, Kenya, and Uganda.

The interventions were designed and implemented by NGOs: Care in Benin, the International Rescue Committee in Ethiopia and Uganda, World Vision in Kenya, and Save the Children in Malawi.

This study was a qualitative interpretive evidence synthesis [25] of stakeholder interview data from a realist evaluation of integrated FP and childhood immunisation services. Whilst there are a plethora of research studies describing the prevalence of, and socio-demographic associations with, the use of MCMs in sub-Saharan Africa, the mechanisms driving the acceptability and adoption of these methods are less well understood. Resonance between concepts of acceptability and realist mechanisms suggests that a realist approach to evaluation could be useful in understanding the acceptability of FP (the spacing and limiting of births) and of the MCMs supported by the interventions (i.e. contraceptive methods excluding traditional and natural methods). The realist evaluation methods were described by Krishnaratne S, Hamon JK, Hoyt J, Chantler T, Landegger J, Spilotros N, et al: What mechanisms drive uptake of family planning when integrated with childhood immunisation in Ethiopia? A realist evaluation, unpublished. For each individual country, the realist evaluation included the development of an initial programme theory, which presented perceptions of what was driving the success of the intervention, as well as the potential barriers. These theories were developed in a workshop with programme designers and implementers from each country, approximately 15 months after the start of implementation.

Semi-structured interviews (SSIs) and focus group discussions (FGDs) were conducted in each country between October 2017 and March 2018 with stakeholders who were either directly or peripherally involved in the projects. Stakeholders with a range of perspectives and opinions on the projects were purposively selected [26] and included implementing partners, government officials, HCWs, community leaders, and women. HCWs were from project sites that were performing well and sites that were performing less well according to monitoring data. SSI and FGD guides were developed to include broad themes on workload, socio-cultural context, healthcare access, reasons for use or non-use of MCMs, barriers to MCMs use, access to FP services, and community-level acceptance of MCM use, together with questions focussed on components from an initial programme theory.

An analytical framework was developed to guide the analysis and was based on realist and acceptability concepts. First, the framework conceptualised relationships between context and intervention components in triggering empirical mechanisms of acceptability. Our analytical framework assumed that the prevailing context, that is the conditions in which the interventions were introduced [18], could directly influence mechanisms of FP acceptability amongst women and the community (including acceptability and non-acceptability); that the context provided the conditions in which interventions triggered mechanisms; and that the intervention could influence and change the prevailing context (Fig. 1). The framework presents context-actor-mechanism-outcome (CAMO) configurations and context-intervention-actor-mechanism-outcome (CIAMO) configurations [27] as the triggers of the empirical mechanisms of acceptability directly by the prevailing context and by the intervention within the prevailing context, respectively.

Second, the analytical framework matches empirically derived mechanisms of acceptability to the constructs of Sekhon et al.’s Theoretical Framework of Acceptability (TFA) [17], ultimately producing context-acceptability theories (CATs). The TFA consists of seven constructs of acceptability relating to participants’ perceptions and feelings about the intervention: affective attitude (general feelings about the intervention), burden (perceptions of effort needed to take part), ethicality (fit with their value system), intervention coherence (understanding of the intervention and how it works), opportunity costs (what must be given up in order to participate), perceived effectiveness (perceptions of the intervention’s ability to achieve its aim), and self-efficacy (feeling that they are able to do what they need to do to take part). The TFA was chosen as the theoretical basis for the development of CATs as it provided clearly defined building blocks for theorising through its seven constructs [28], potential to explore mechanisms triggering constructs of acceptability, and links between triggers of acceptability constructs and context. The CATs developed through this analytical framework are middle-range theories. Middle-range theories enable the consolidation of hypotheses and empirical realities and allow these to be grouped into theories that can be further empirically tested [29].

Translated transcripts were imported into NVivo 11.2 for coding and analysis. Data was anonymised and labelled by type of stakeholder. A coding framework was developed which included context (economic/political, community/social, and organisational/team), actors (community members and HCWs), interventions (integration of immunisation and FP, community mobilisation, HCW training, and supply chain systems), mechanisms (HCW decision-making, community decision-making, women’s decision-making), and outcomes (expected and unexpected/unrelated or additional). This deductive coding was supplemented by the inductive addition of sub-codes, which expanded the depth of information within each code. Coding was undertaken for each country individually.

An initial thematic narrative analysis [30] was conducted for the realist evaluation for each of the five countries using the individual country coding and further analysis to construct context-mechanism-outcome configurations (CMOs) [18]. The CMOs for each country were used to construct a revised programme theory to represent the context- or intervention-related triggers of identified mechanisms, which were perceived by stakeholders to drive intermediate outputs and ultimately the outcome of FP uptake in each country. CMOs and revised programme theories were then discussed and validated by co-authors (JKH, JH, and SK).

For the context-acceptability synthesis, the realist evaluation CMOs were re-examined to identify those that included mechanisms of acceptability, and expanded to either CAMOs or CIAMOs. CAMO configurations were used in this study to analyse the influence of the prevailing context on mechanisms of acceptability and CIAMO configurations to analyse the influence of the intervention on mechanisms of acceptability. Identified empirical mechanisms of acceptability were then reviewed against and matched to TFA constructs of acceptability. Adaptations to the TFA constructs were made as required.

Finally, CATs that summarised constructs of acceptability triggered for specific actors in specific contexts were developed and examples of interventions that could be used to trigger acceptability for these actors were described. The Standards for Reporting Qualitative Research [31] and reporting which incorporate the role of theory were used to ensure rigorous reporting of the study (see additional information).

Ethics approval was obtained from the Comité d’Éthique de l’Institut des Sciences Biomédicales Appliquées, Benin; Benishangul-Gumuz Regional Ethical Review Board, Ethiopia; African Medical and Research Foundation, Kenya; National Committee on Research in the Social Sciences and Technology, Malawi; Mildmay Uganda Research Ethics Committee, Uganda; and the ethics committee of the London School of Hygiene & Tropical Medicine.

Five major prevailing contexts were found to directly influence the mechanisms of acceptability of FP by the community, men, and women (Table 2). These were (1) FP was believed to be incongruous with Islamic (Ethiopia and Kenya) and Christian (Kenya) religious principles, teachings, and values; (2) traditional desire for large families and the assumption that FP aims to reduce the family size (Kenya, Uganda); (3) stigmatisation of MCM use (Benin); (4) some male partners who are non-accepting of FP or MCM use (all countries to differing extents); and (5) experience or rumours of side effects of MCMs (all countries). Each of these contexts directly led to mechanisms of non-acceptability of FP or MCMs amongst specific actors.

Nine CIAMOs representing the empirical mechanisms of acceptability triggered by intervention components in specific actors within each context were identified, as was one CIAMO representing an empirical mechanism of non-acceptability (Table 3).

Empirical mechanisms of acceptability were aligned to one or more constructs of the TFA and varied by context and actor (Table 3). Mechanisms that were assessed to be strongly aligned to a specific construct were termed dominant, and mechanisms that were more tenuously aligned to other constructs were termed additional. Ethicality was a dominant mechanism of acceptability of FP in contexts where there was a strong belief in religious values and a traditional desire for large families. Ethicality was accompanied by opportunity costs such that leaders understood that they would not need to compromise on their religious values to accept, promote, and encourage the use of MCMs. In contexts where a traditional desire for large families was prevalent, mechanisms of acceptability were aligned with these same constructs of ethicality and opportunity costs but in the community more generally, including amongst women. That is, using MCMs to achieve healthy birth spacing was no longer equated with compromising on the desired family size. Acceptance of birth spacing also aligned with the construct of intervention coherence and was extended beyond the TFA definition to include ‘believing information on how it worked’.

The TFA was adapted to include an additional construct: ‘unintended consequences’ which was defined as ‘the extent to which the intervention leads to consequences beyond those for which it was designed’. The elevated risk of stigmatising women was perceived as an unintended consequence of the use of referral cards from childhood immunisation to FP services in Benin, due to women’s fears of being seen accessing FP services by their peers. Providing access to MCMs in private after the clinic closed was a reactive adaptation of the intervention led by HCWs to the stigmatisation of MCM use. This mechanism of women feeling confident to accept MCMs when they were not seen by others aligned with the construct of self-efficacy, as women were confident they could do what was needed to take part in the intervention without the risk of unintended consequences.

Intervention coherence was a dominant TFA construct aligned to the empirical mechanisms driving acceptability of FP and MCM use amongst male partners. At intervention sites in Kenya, male partners believed the information they received about the advantages of birth spacing, in terms of financial, educational, and health benefits for their families, because this information was provided by trusted and relatable male role models. Male partners also acknowledged the value of using MCMs to space births, which was perceived effectiveness.

Seven CATs were developed by linking the empirically derived mechanisms of acceptability to TFA constructs. Three of the theories involved ethicality mechanisms of acceptability, three involved self-efficacy (with or without unintended consequences), two included opportunity costs, and one involved intervention coherence and unintended consequences. In Table 4, each CAT includes an identified context and the mechanism that will lead to the acceptability of FP (and MCM use) within this context. Thus, a CAT specifies what exactly will trigger an actor to accept FP and therefore highlights the response in the actor that an intervention needs to bring about. Conversely, a CAT also indirectly suggests what interventions are not likely to be successful. For example, in contexts where there is a strong belief in religious values and where there are powerful religious leaders, an intervention that only teaches religious leaders about how FP works and about its benefits to women and families is unlikely to be effective as it will not trigger acceptability mechanisms aligned with ethicality, i.e. religious values (theory 1). Similarly, telling women that they may experience side effects when using MCMs without teaching them how to manage these should they occur is unlikely to be effective as acceptability mechanisms aligned with self-efficacy will not be triggered (theory 7).

The acceptance of FP alignment with religious values and the promotion of this alignment in the community created a new prevailing context for men in Kenya as they came to perceive FP as fitting with their values (ethicality). Subsequently, acceptance of FP amongst male partners became the new prevailing context in which women felt they were supported and therefore able to accept and use MCMs (self-efficacy). This can be seen as a context-acceptability cascade where mechanisms and outcomes of acceptability in one group of actors, or an intervention, change the prevailing context for another group. This new prevailing context then in turn triggers the second group’s mechanisms of acceptability (Fig. 2). The CAT for each population group within this cascade differs.

The mechanisms of acceptability that were identified from the empirical data and that were directly triggered by the prevailing context centred on contexts with a negative or potentially negative influence on the acceptability of FP/MCMs. For each of the five categories of prevailing contexts, intervention components were identified that triggered positive mechanisms of FP acceptability despite the negative contextual influences. Some intervention components targeted specific actors and others were aimed at the wider community. Interestingly, there were instances where it was not only the intervention itself that triggered a mechanism of acceptability, but also the actor who delivered it. For example, where male role models were used to promote the financial, educational, and health benefits of using MCMs for birth spacing, mechanisms of trust in, and relating to, the male role models who delivered the messages contributed to the mechanism of belief in the messages. Findings from this study support the idea that context is not only multi-layered, but is also non-static. Context may change during the implementation of an intervention due to external factors or due to components of the intervention itself. This change in context will lead to new mechanisms of acceptability.

Knowledge of empirical mechanisms is of practical value in both the design and ongoing implementation of interventions. For instance, during the design of an intervention, identifying that in some strong religious contexts FP and the use of MCMs are believed to be misaligned with religious values highlights what the trigger for acceptability will be. The identification of these triggers, and of their nature, indicates what interventions should focus on and which population group they should target to be effective.

Matching empirical mechanisms of acceptability with TFA constructs allows for abstraction and transferability. The middle-range CATs identify which mechanism of acceptability must be triggered in a given context. For example, in contexts of strong religious values, ethicality is the mechanism of acceptability that needs to be triggered. Therefore, the intervention that is chosen must be able to convince actors that FP fits with their values. Showing alignment of FP using religious text was an example of this. Similarly, there may be other interventions that can convince key actors that MCM use does not compromise their values. However, it is unlikely that an intervention that only focusses on the effectiveness of FP and/or how it works would achieve an increase in acceptability of FP in this context. This approach to identifying dominant constructs of acceptability may be useful in increasing the acceptability of other public health interventions beyond FP.

Our CATs identified five major contexts impacting the acceptability of FP with nine intervention components triggering mechanisms of acceptability in these contexts. Whilst these include relatively recognised FP strategies (e.g. birth spacing messaging), further evidence on intervention outcomes and processes would be valuable for other, more novel, strategies. The relationship between religious values and FP varies across religions and in denominations and sects within these religions [33, 34]. Christian texts and Islamic texts without direct statements can lead to varied interpretation in both support and opposition to FP [35] and may be driven by the views of individual local religious leaders [36]. This adds support for the potential of well-considered interpretation of religious texts, and of exploration into the most effective processes for disseminating these interpretations.

Privacy and covert use of MCMs fit naturally within the women’s empowerment agenda and associated frameworks [37, 38] and research; however, the majority of the studies on women’s sexual and reproductive empowerment have been conducted outside of sub-Saharan Africa [39, 40]. Recent studies in a number of countries of East and West Africa have investigated women’s reasons for covert use, challenges to covert use, and consequences of discovery [41‐44]. Realist approaches to interrogating decision-making offer opportunities to deepen the understanding of women’s covert use of MCMs within currently developed women’s empowerment frameworks of the existence of choice, exercise of choice, and achievement of choice [40]. Management of side effects of MCMs is a challenge to continued covert use [41] and is an important component of FP counselling. An understanding of the influence of advice given on side effect management and of the ability to manage side effects on the continuation of MCM use amongst women who are covertly using MCMs would be important in supporting women who are unable to openly use MCMs.

Multiple theories have been used in FP studies; however, to our knowledge, the TFA has not been used. For example, the Theory of Planned Behaviour, the Health Belief Model, and Social Cognitive Theory were each used in evaluating a decision aid for postpartum women in Kenya [11] whilst the TFA was not. We found the TFA to be highly applicable to developing middle-range theories for mechanisms of acceptability of FP. We found one adjustment was needed to the wording of one of the TFA constructs, which was the addition of ‘…and believe what they are told about how it works’ in reference to the Intervention Coherence construct (Fig. 3). We considered ‘understanding’ the intervention and how it works as weak in terms of triggering a decision that would drive an outcome [18] similarly to the adage that knowledge does not necessarily trigger behaviour change [45]. To accommodate the mechanisms that we identified empirically, we added a new construct to the TFA. In their preliminary TFA, Sekhon et al. included the construct ‘Ethical Consequences’ defined as ‘associated side effects with the intervention’. During further development, this construct was removed. We added the construct of ‘unintended consequences’ to include both the health-related side effects of MCM use and the negative consequences of being exposed to FP-related stigma.

Complex multi-component interventions necessarily take different forms in different contexts as they interact with people, institutions, and other systems [46]. Decision-makers are tasked with understanding whether interventions that work elsewhere will work in their setting. Here, we recommend that for interventions to be accepted, the context and not the intervention should be the first consideration, with the intervention then designed or selected based on CATs. That is, rather than asking ‘will this intervention work in this context?’ the approach should ask: ‘given this context, which intervention do we need to apply?’

Examples of interventions achieving expected outcomes in one geographic site yet failing to do the same when implemented elsewhere are prevalent in both FP and reproductive health more widely [47, 48]. This non-transferability of effectiveness is most often linked to differences in contexts. That is, the mechanisms that drive outcomes, such as acceptability, are not triggered by the intervention in the prevailing context. We suggest that there is significant potential to reduce the prevalence of unsuccessful interventions and resulting waste of resources through firstly defining CATs and subsequently using them to select appropriate interventions.

A practical approach to using CATs to select and implement appropriate interventions in a particular geographic setting would involve two complementary approaches, which are (1) generating and testing further CATs and (2) defining the context in a project/programme site and selecting an appropriate intervention based on these CATs. Our seven CATs provide the basis for a database of context-linked mechanisms of FP acceptability based on the expanded TFA constructs. Further research to expand the current CAT database as a validated resource for decision-makers would involve testing the CATs in other geographic settings, reviewing the literature for evidence to support or refute each of the CATs, and expanding the database to add new CATs as they are developed from empirical studies. The database would serve as a resource to projects/programmes obviating the necessity to develop new CATs in each site. A simple assessment of context could enable the determination of dominant contextual factors influencing the acceptability and uptake of FP. Such an assessment could take the form of individual interviews or a participatory workshop with key informants that ask what influences the acceptability of FP in the defined sites amongst providers and their clients. This simple empirical investigation of context coupled with reference to the CAT database would then inform the selection of interventions.

Finally, our study focussed on CATs specifically for acceptability and uptake of FP; however, similar CATs could be developed for practical application with other public health interventions using the approach outlined in this manuscript.

Our study made a number of novel contributions, each demonstrating the value of realist approaches in implementation research. We showed that despite the complexity of context, there were a limited number of major contextual factors that influenced mechanisms of FP acceptability in key actors. We highlighted the usefulness of abstracting context-linked empirical mechanisms to constructs of the TFA in predicting what intervention should be implemented where, and critiqued and suggested adaptation of the TFA. We defined context-acceptability cascades which are key to understanding the need to adapt implementation through the life of a project. In terms of the field of FP specifically, whilst the mechanisms of FP acceptability included recognised strategies, they explained why these strategies worked and where they would and would not work. The need for evidence on other more novel strategies such as triggering ethicality in religious leaders and the question of covert MCM use were also identified.