Introduction

Embedded implementation research (IR) is an approach that can enhance the effectiveness of implementation and scale up of a program by promoting evidence-informed policy and practice [1–4]. Embedded IR bridges the gap between research and policy by embedding relevant decision-makers and program implementers (program managers, frontline health workers, etc.) in the process of research and dealing with issues related to health systems that have relevance in programs and policies [1–6]. This approach focuses on the questions identified by decision-makers and program implementers and ensures that knowledge generated is relevant. As an emerging area of research, embedded IR has been put forward and applied by several countries and organizations as a cornerstone to health systems strengthening and achieving universal health coverage.

While individuals of the ethnic Rohingya minority population in Myanmar have been coming to Bangladesh since 1978 [7–9], a major exodus happened in August 2017. As of January 2019, over 900,000 Rohingya refugees resided in Ukhiya Upazila and Teknaf Upazila (sub-districts of Cox’s Bazar District), while the largest single site, the Kutupalong–Balukhali expansion site, hosted more than 600,000 Rohingya [7–12].

Despite substantial progress during past 2 years, Rohingya refugees in Bangladesh remain in a precarious situation. Poor living conditions; lack of safe water, sanitation, and hygiene; high rates of child marriage; lack of support for education; poor healthcare; and lack of livelihoods are major concerns for them [13]. More than 55% of the Rohingya refugees are women and children, many of whom have faced gender-based violence, abuse, trafficking, malnutrition, and serious health problems [14–16]. Utilization of skilled maternal, newborn, child, and adolescent health (MNCAH) services is very low among the Rohingya refugees. One report found that about 78% of Rohingya babies are born in unsafe and unhygienic makeshift shelters in the district [15]. Coverage of immunization and services to prevent and treat other child health illnesses are also low in the camps [8,10–12,15,17].

Considering the overall situation, UNICEF in collaboration with BRAC University in Bangladesh, undertook an embedded IR project from 3 January to 31 July 2019 with the aim of strengthening MNCAH programs in Cox’s Bazar District. The key objectives of the project were to identify key implementation challenges of MNCAH programs, explore potential solutions, and ensure utilization of those solutions for effective implementation of MNCAH programs in the camps through engagement of decision-makers and program implementers. Part of the findings and methods used for the project were published elsewhere [18]. In addition, one of the objectives of the initiative was to document methodological and operational challenges of conducting embedded IR in humanitarian settings, which the present study focuses on.

Several articles have highlighted the challenges of conducting research in humanitarian settings in the context of an emergency [19–21]. However, despite a growing body of evidence on the value of research in emergency settings, there is still a knowledge gap regarding the challenges of conducting embedded IR in these settings [6,21–25]. In recognition of this gap, this present study explored the challenges and key lessons learned relating to the design, implementation, and management of embedded IR in humanitarian contexts based on the Cox’s Bazar experience.

Methods

This study was part of a larger embedded IR project conducted by BRAC University and supported UNICEF. Results of that embedded IR are reported elsewhere [18].

For the current study, a qualitative method was applied to capture the researchers’ experience (i.e., challenges) of conducting embedded IR in the Rohingya emergency context in Cox’s Bazar. We collected qualitative data in 2 phases to document researchers’ experience of conducting IR and to describe potential implications for embedded IR work in other emergency settings.

Study population and sampling

In January 2019 (first phase), we, the research team (ASMS, AH, and MS), visited a UNICEF-supported primary healthcare center and a health post within the Kutupalong refugee camps in Ukhiya Upazila and interviewed representatives from non-governmental organizations (NGOs), United Nations (UN) organizations, and the Bangladesh government who have been supporting and implementing MNCAH programs in Cox’s Bazar. Questions were asked to document the geographic location of the camps, living conditions of the Rohingya refugees, MNCAH service delivery modalities, and related policies and roles and responsibilities of different stakeholders.

To complement the data of the first phase, in the second phase (July 2019, following completion of the larger embedded IR project), we interviewed 5 purposively selected members of the BRAC University research team. We asked about their experiences conducting IR in Cox’s Bazar, challenges they encountered while collecting data, and their suggestions for future embedded IR in similar settings. Table 1 shows the data collection methods and types of respondents interviewed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Data collection methods and types of respondents.

https://doi.org/10.1371/journal.pmed.1003148.t001

Results

Five key themes emerged from the observation and interview data: (1) unique and complex context with multiple actors, (2) adaptation due to a dynamic situation and changing service delivery modalities, (3) difficulties accessing the camps and research participants, (4) language and other barriers to accessing quality information, and (5) need for experienced in-country research partners.

Discussion

More than 100 organizations have been involved in providing MNACH services in Cox’s Bazar since August 2017. Despite the presence of the Bangladesh government, UN agencies and other INGOs have been playing key roles in formulating programmatic and policy decisions for Rohingya refugees. Over the study period, rapid change has been observed in delivering health services and formulating policies, as well as variation in the implementation challenges of MNCAH programs. Therefore, the approach for embedded IR had to be flexible to adapt to changes identified, with research questions and methods modified accordingly. In addition, access to the camps, reaching Rohingya respondents, overcoming language barriers in order to get quality information, and the limited availability of local research collaborators were highlighted as additional challenges by the respondents.

Every emergency is unique in terms of the severity of the problem, population affected, geographic and political situations, modalities of service provision, and the number and types of stakeholders involved in provision of services. Similarly, the need for, appropriateness of, and use of research in such settings will vary.

Embedded IR is distinct from other types of research in that decision-makers and program implementers are actively engaged in research design, data collection, analysis, and adoption of findings [1,4,29–31]. Active engagement of program implementers at the outset and throughout the research project is critical. A key question is whether it is reasonable to do embedded IR in an emergency setting such as the Rohingya refugee crisis, bearing in mind the additional commitment that active involvement in embedded IR requires of implementers who are already under significant pressure. Both practical and ethical factors must be taken into consideration in any particular emergency. As part of the planning for IR in humanitarian settings, it is therefore crucial to have an explicit understanding of the context, to identify key stakeholders, and to analyze their roles and capacity in programmatic policy decisions and service provision [1,3,4,29,32].

In the context of Cox’s Bazar, an initial visit to the camps and interviews with various stakeholders were helpful to map the health situation, geographic locations, and available resources and to assess the roles of different actors. Insights from this initial assessment helped the BRAC University research team to identify and develop appropriate context-specific research tools (interview, observation guidelines, etc.), to recruit local research assistants who could conduct interviews in the language of the target population of Rohingyas, and to know which organizations and stakeholders should be part of the research process to ensure subsequent uptake of research findings and recommendations.

The embedded IR approach that UNICEF and other organizations have used typically engages decision-makers and implementers, particularly officials from the Ministry of Health and program implementers, in the research process [33]. However, the initial assessment of implementers’ capacity and their availability to participate in the research suggested that the model for research in the emergency setting would have to diverge from the traditional model for embedded IR. In the context of Cox’s Bazar, considering the active role of UN organizations and INGOs such as Médecins Sans Frontières and Save the Children, the research teams also engaged those organizations to ensure utilization of the IR findings. Further, it was decided that instead of direct and continuous involvement in design, data collection, analysis, etc., the implementers would be periodically consulted and given opportunities to provide input at critical stages in the research project. This proved to be more appropriate to ensure their active engagement, while also respecting the level of involvement they felt was feasible for them.

As IR deals with real-time implementation problems, so the method chosen to carry out any particular study should fit the purpose and should address the relevant and real-time questions identified. It should provide evidence that can be used for real-time program improvement in a dynamic and nonlinear way [1,4,29,34,35]. Like in other emergency settings, the overall situation of Rohingya refugees has changed rapidly, as have the barriers to implementing MNCAH services [36]. The scope of the response itself has evolved from addressing the urgent needs of refugees, to considering the needs of host communities, the risks of a “crisis within a crisis” posed to both groups during the monsoon season, and medium-term planning [8,9,12,37]. Given such an unstable situation, and to align IR with existing MNCAH programs, a relatively easy and flexible research method that could produce quick, real-time data was pivotal. Therefore, instead of using a complex experimental design, a simple descriptive design with both qualitative and quantitative approaches may be more feasible and appropriate in other such settings [19–21]. In addition, an adaptive and iterative approach involving relevant stakeholders may be more appropriate to understand the situation in each stage of the emergency, learning and adjusting the methodology accordingly, as required [38]. In some cases, analysis of routinely collected secondary MNCAH data may be helpful for answering questions related to implementation barriers, in which case collection of primary data may not be necessary.

Gaining access to certain areas and reaching potential respondents are common research challenges in humanitarian settings. In Cox’s Bazar, due to several recent incidents in the camps, the government of Bangladesh put access restrictions on research organizations. Before going to the camps, every research organization needed government approval. Moreover, the overall condition of the camp settlements in Cox’s Bazar was very challenging. Almost 70% of locations were accessible only by footpath, which created an extremely challenging situation for the delivery of humanitarian services [36]. Being a hilly and coastal area, the sites were highly vulnerable to rain, floods, cyclones, fire, and landslides, a situation made worse during the rainy season [8,10,36]. In our study, the research team had to visit the sites multiple times to get respondents during the rainy season, which cost additional time and money. Further, the safety of the research teams is an important consideration in humanitarian settings as difficult terrain, weather, and violence are often problems, as has been highlighted in other studies [19,21,39,40].

Further, language is often a challenge to researchers working with linguistically and culturally diverse communities [41]. The language barrier may result in failure to understand the true situation of certain communities or may result in failure to include certain groups in a research project, especially when this barrier proves insurmountable. This barrier can be mediated through the use of an interpreter or translator or bilingual researchers [21,39]. However, it becomes difficult to collect high-quality information when a bilingual researcher does not have sufficient research skills. In our study, the BRAC University research team had to spend substantial time on training to build the capacity of the local bilingual research assistants to make them familiar with the interview guidelines. Moreover, during the interviews, bilingual research assistants had to be supported by senior researchers. These factors increased the amount of both time and money needed to complete the research.

Challenges of conducing IR in humanitarian settings can be reduced with the involvement of researchers or research institutions with adequate research capacity and understanding of the context and research approach. Although the BRAC University research team was familiar with the context and with the IR approach, they still had challenges collecting data due to difficulties accessing the camps, finding respondents on time, and finding good bilingual researchers to do interviews with the Rohingya participants. In general, research capacity within humanitarian contexts can be a challenge, particularly in low-income countries that lack researchers or research institutes with the required research skills and knowledge of local settings [42,43]. This research therefore often depends on international researchers or research organizations despite the fact that this can require more funding and time than working with a local research institution. Further, IR is still a relatively new area for many researchers in low-income settings. However, investing in collaborations with in-country researchers or research institutes provides opportunities to build their capacity for IR in humanitarian contexts, which may help to institutionalize IR in emergency settings. This type of collaboration may also reduce research costs and time and increase the likelihood of obtaining reliable data that are useful for policy and programmatic decisions in the local context.

Supporting information

Acknowledgments

The authors would like to acknowledge colleagues from UNICEF’s Cox’s Bazar office (Yulia Widiati, Abdullah Bin Akhtar, ASM Mainul Hasan, and Helen Chakma) for their support in the field during data collection, and colleagues at the UNICEF Bangladesh Country Office (Maya Vandenent and Juanita Vasquez Escallon) and ROSA (Samuel Bickel) for their useful feedback on the proposal. The authors would also like to thank the study participants and participating agencies in Cox’s Bazar.