REVIEW OF THE EVIDENCE LANDSCAPE ON THE RISK COMMUNICATION AND COMMUNITY ENGAGEMENT INTERVENTIONS AMONG THE ROHINGYA REFUGEES TO ENHANCE HEALTHCARE SEEKING BEHAVIOURS IN COX’S BAZAR PRODUCED BY THE SOCIAL SCIENCE IN HUMANITARIAN ACTION PLATFORM ON THE REQUEST OF THE UK FOREIGN COMMONWEALTH & DEVELOPMENT OFFICE IN BANGLADESH

This report is the first output in a body of work undertaken to identify operationally feasible suggestions to improve risk communication and community engagement efforts (RCCE) with displaced Rohingya people in Cox’s Bazar. Specifically, these should seek to improve healthcare seeking behaviour and acceptance of essential health services in the camps where the Rohingya reside. It was developed by the Social Science in Humanitarian Action Platform (SSHAP) at the request of the UK Foreign, Commonwealth & Development Office in Bangladesh. As a first step in this process, this review paper synthesises and assesses the quality of evidence landscape available in Cox’s Bazar and how the Rohingya seek and access healthcare services in Cox’s Bazar and presents the findings from key informant interviews on the topic. Findings are structured in five discussion sections: (1) evidence quality; (2) major themes and variations in the evidence; (3) learnings drawn and recommendations commonly made; (4) persistent bottlenecks; and (5) areas for further research. This synthesis will inform a roundtable discussion with key actors working for the Rohingya refugees to identify next steps for RCCE and research efforts in Cox’s Bazar to improve health outcomes among the Rohingya.


Executive Summary
This report is the first output in a body of work undertaken to identify operationally feasible suggestions to improve risk communication and community engagement efforts (RCCE) with displaced Rohingya people in Cox's Bazar. Specifically, these should seek to improve healthcare seeking behaviour and acceptance of essential health services in the camps where the Rohingya reside. It was developed by the Social Science in Humanitarian Action Platform (SSHAP) at the request of the UK Foreign, Commonwealth & Development Office in Bangladesh. As a first step in this process, this review paper synthesises and assesses the quality of evidence landscape available in Cox's Bazar and how the Rohingya seek and access health care services in Cox's Bazar and presents the findings from key informant interviews on the topic.
Findings are structured in five discussion sections: (1) evidence quality; (2) major themes and variations in the evidence; (3) learnings drawn and recommendations commonly made; (4) persistent bottlenecks; and (5) areas for further research. This synthesis will inform a roundtable discussion with key actors working for the Rohingya refugees to identify next steps for RCCE and research efforts in Cox's Bazar to improve health outcomes among the Rohingya.

What are the issues with some of the evidence?
The key informant interviews revealed a concern about biases in the available data. The quality assessment of the available research methodologies and triangulation of data pieces provides insights into these discrepancies.

Sampling bias
Overgeneralising findings from small or location-specific samples can result in the magnification of certain findings and the minimisation of others, creating a distorted broader picture. Convenience and passive sampling methodologies may also overrepresent certain voices and perspectives within a community. Some of the discrepancies in the study's findings are likely reflective of the heterogeneous and complex landscape in Cox's Bazar. Transparency about interventions mounted, expected output, samples and sampling methodology would help to elucidate differences in findings as well as the adoption of sampling strategies stratified adequately to incorporate those heterogeneities. It would also allow assessment of whether interventions have been successful and would be appropriate for further scale up.

Response bias
Accounts of positive response bias from Rohingya research participants were common. The underlying drivers of observed response bias reportedly included sociocultural factors, language, structural factors and power dynamics, and limitations in research practices and skills. Individual and community experiences, the sensitivity of the topic, who is conducting the research, the research practices and assurances in use, and the skills and mannerisms of the researchers will all contribute to response bias. Understanding the drivers of response bias could explain how data sets seemingly tell different stories.

Confirmation bias
Humanitarian aid and emergency responses around the world are under financial and political pressure to implement timebound interventions / projects, perform and demonstrate results, this can shape how data are received and managed. Confirmation bias and perceived incentives for 'positive' data or success of the project may explain some of the discrepancies in data and findings. In some instances, success appeared over projected, adding value to the health system actors.

Transparency
Unfortunately, many of the reports examined did not fully describe the methodology, tools, or limitations, making it difficult to ascertain the validity and limitations of the findings. This lack of transparency about methods and data sets can obscure limitations and context, making it challenging to fully understand potential biases in the findings. Greater transparency would be useful to provide information to the government and other partners to assess the sustainability and added value of the actions in the Rohingya camps.

Major themes and areas of conflicting information
Despite the abundance of data from Cox's Bazar, seeming contradictions and unexplained variations are commonly reported in findings on Rohingya community trust in the humanitarian response and actors, healthcare seeking behaviour, healthcare providers' behaviour, acceptance and compliance with health prevention measures, and community engagement and feedback mechanisms. This section explores the major themes and areas of conflicting data considering the biases described above to better understand identified drivers and gaps.

Trust in the humanitarian response and actors
Findings regarding trust in the humanitarian response and actors varied widely. The power dynamics at play and tendency for positive response biases about sensitive issues likely affect the reliability of findings about trust. Variations and contradictions in findings about trust, distrust and compliance with prescribed actions should also consider that the response is not uniform, and neither are the communities. Trust may vary organisation to organisation, response actor to response actor, community member to community member, and situation by situation. This is even more challenging during the public health emergencies like COVID-19 crisis when health system actors expect the individuals and communities to comply with the public health and social measures and RCCE interventions for its containment.
Healthcare seeking behaviour Data indicated a discrepancy between reported and actual healthcare seeking practices. There were also variations in healthcare seeking behaviours depending on the health issue and community group in question. These variations appeared to reflect response bias and situational and gendered differences in healthcare seeking among the Rohingya.

Rohingya patient experience
There were conflicting data about Rohingya experiences at health facilities. Some findings showed predominantly positive and satisfactory experiences, others reported majority negative experiences and perceptions, and others highlighted very mixed experiences. Even if positive experiences were more frequent, reports of negative experiences tend to spread more quickly through communities, with a negative impact on healthcare seeking behaviour and trust. Identifying the underlying factors contributing to negative experiences in health centres is essential to continuing to improve services and accountability. These factors included perceived discrimination and disrespectful treatment, unmet health needs, communication and language barriers, perceptions of unsatisfactory consultation and unavailability of specialised services, and inadequate health care capacity to provide services. In addition, overburdened health facilities with only day-care services and limited availability of 24-hour services also aggravated the situation.
Knowledge, awareness and practice While community members were reported to appear to have satisfactory levels of knowledge and awareness about health and healthcare, these did not always translate into practice. These discrepancies demonstrated that community members were receiving and understanding the messages and advice being given to them as they were able to give the 'correct' responses when asked, but for various reasons were not putting that information into practice. Explanations include community understandings of risk, sociocultural and religious norms, and structural and logistical barriers. There is a major challenge to cascade knowledge to women and adolescent girls and to translate knowledge from RCCE into practice.

Community engagement and feedback mechanisms
While in some research most community members reported knowing how to give feedback and that aid providers took their opinions into account, other studies show that most community members do not feel consulted or have negative experiences providing feedback. Positive response bias and sampling bias appear to be affecting some of the data about community engagement, particularly survey data. However, the findings also point to important experiences and perceptions of inclusion and community engagement necessary to building trust and align response efforts with community priorities and norms. Findings indicate that good experiences of engagement for Rohingya include culturally appropriate and respectful treatment, good communication, informal, open-ended consultation processes that match their comfort levels, and follow-up on reported problems.

Lessons learned and key recommendations that are often made
This section sets out some of the lessons that have been learnt in terms of communication, community engagement and accountability strategies, data collection, and coordination. It also summarises some of the main recommendations for improvement that are commonly made in the literature and by stakeholders interviewed for this project.

Communication, community engagement and accountability strategies
The risk communication aspect of RCCE, and content of messaging relating to risk, was generally considered to be appropriate in Cox's Bazar. However, what was still lacking is two-way engagement that provides the refugee population with the opportunity to engage with the material provided, ask questions, and contribute their own experiences, concerns or ideas. Only limited attempts have been made to capture the concerns, queries and misinformation prevalent in the camps through interactions by some of the RCCE partners; further efforts would help the RCCE WG to review the interventions and tailor messages accordingly. There is scope to improve community engagement, and therefore increase accountability and build trust in Cox's Bazar are by: using tailored engagement methods for specific target groups, with a focus on two-way communication and feedback opportunities; increasing the inclusion of the Rohingya in all aspects of the response; engage trusted interlocutors; and, use appropriate language and formats.

Data collection
As mentioned, the quality of the data collected was greatly influenced by the data collection method. A common theme emerging from the key informant interviews was the lack of standards in data collection and research ethics. Four key recommendations emerged: recruit Rohingya researchers wherever possible; improve researchers' skills and competencies on use of local languages; use culturally and methodologically appropriate research methodologies; identify appropriate data collection techniques, particularly in the evolving scenario of COVID-19 and for the RCCE WG to put oversight mechanisms to ensure quality are in place.

Coordination strategies in RCCE
Attempts to improve coordination appear to be working relatively well, such as coordination about the content of messaging, including its approval process. Potential avenues for further improvement include: streamlining and clarifying the coordination structure; including more diversity of voices in the working groups and decision-making processes; and, improving coordination of specific areas of RCCE, such as volunteer networks, and data collection, sharing and referral.

Structural bottlenecks
While improvements have been made to communication, community engagement and accountability in recent years, some structural bottlenecks remain. These include reports of: a complex coordination structure and limited coordination among the many response partners; lengthy approval process, inadequate investment in community engagement to reach targeted individuals and communities; limited learning and responsiveness with regard to community engagement; low levels of accountability in some response efforts; high staff turnover in the health sector across the response especially during the crisis triggered by COVID-19 that limited travel; low quality of care and poor treatment of Rohingya at health facilities; policy constraints hindering the inclusion of Rohingya in response activities; policy constraints hindering refugees' access to telecommunications and Internet connectivity; and restrictions or delays to entering the camp that hinder efforts to improve community engagement and accountability.

Sociocultural bottlenecks
Entrenched social norms on the part of both host and guest communities can translate into barriers to healthcare seeking. These are sensitive to address, and difficult to challenge although efforts are being made through inclusive provision of services, collaborative projects and sensitivity trainings for healthcare staff and the support staff. Notable challenges raised include experiences of widespread discrimination, sociocultural and language barriers, gender norms and language.

Logistical bottlenecks
While there was some evidence that risk communication and messaging is working, and people are understanding the content of the messages, it was not always feasible for them to change their behaviour to comply with the advice received due to logistical bottlenecks, such as availability of limited materials, equipment and distance to health centres.

What areas require further investigation?
There is no lack of data collection in the Rohingya refugee camps in Cox's Bazar. However, according to the literature and key informants, gaps remain in the types of data collected as well as in specific areas relevant to health outcomes and community engagement efforts. It would be beneficial to write up and share best practices and learnings from the RCCE interventions as well as reviewing and exploring synergies from community engagement actions on priority topics.
Collecting the right type of data Most data collected in Cox's Bazar used quantitative methods. However, there has recently been more recognition of the importance of qualitative research to fill in the hows and whys driving the patterns being seen in the quantitative data findings. Research methods that enable participants to freely share their experiences, concerns and queries including enabling and limiting factors could be invaluable for designing appropriate and relevant RCCE strategies and interventions.

Identified knowledge gaps
While it has been noted that there are plentiful data collected in Cox's Bazar, there are still knowledge gaps in critical areas including the health status of the Rohingya communities, the underlying drivers of healthcare seeking behaviours, Rohingya experiences of healthcare facilities, how the Rohingya perceive the RCCE interventions themselves, and the extent to which these translate into improved health outcomes.

Introduction
Since 2017, more than 742,000 Rohingya Muslims have fled to Bangladesh from Rakhine State in Myanmar. They were fleeing government-sanctioned violence, human rights abuses and systemic discrimination and persecution by the Myanmar government and military. This had been going on for decades, and Rohingya refugees had been arriving in Bangladesh since 1978. The Rohingya community is the largest stateless group of people in the world, and this is one of the largest and most protracted refugee crises globally Holloway & Fan, 2018). The response has been led by the Bangladeshi government with the support of over 150 international and national organisations. Communication, community engagement and accountability face particular challenges in Cox's Bazar because of the Rohingya's long history of persecution and therefore low levels of trust of authorities or outsiders, the need to communicate in the Rohingya language, their low levels of literacy and the lack of an international recognised script for the Rohingya language, and their limited access to radios, mobile phones and Internet .
The UK FCDO in Bangladesh requested the support of the Social Science in Humanitarian Action Platform (SSHAP) in identifying operationally feasible suggestions to improve risk communication and community engagement efforts (RCCE) aimed at improving healthcare seeking behaviour and acceptance of essential health services in the camps where the displaced Rohingya reside.
SSHAP, as an external platform, has been working with WHO to better understand the existing landscape of literature on the topic, to assess the quality of available data and scrutinise conflicting areas in the data, to identify additional areas requiring investigation, and to identify lessons learnt and recommendations to inform structurally improved risk communication and community engagement efforts for health in the Rohingya response In Cox's Bazar. This review paper presents a brief account of the methodology of the synthesis, a discussion of some of the issues around quality of the evidence, an account of areas in which there is conflicting information, a discussion of the lessons and key recommendations commonly made, a discussion of persistent bottlenecks, and an overview of emerging gaps in the research. The synthesis will inform a round table discussion with the aim of establishing priority next steps for RCCE and research efforts in Cox's Bazar to improve health outcomes among the Rohingya.

Background to RCCE response in Cox's Bazar
The Communicating with Communities (CwC) coordination structure was established in early September 2017, shortly after the large influx of Rohingya that year. This built upon earlier work at the national level, with the creation of the Shongjog platform in 2015 . As part of the Inter-Sector Coordination Group (ISCG), the CwC Working Group (CwC WG) was established as a multisectoral, cross-cutting initiative whose aim was to help organisations to mainstream CwC and accountability in their operations, and to improve the coordination of messaging and feedback mechanisms across the response. Prior to the COVID-19 pandemic, the RCCE TWG was briefly co-led by WHO and UNICEF. With the onset of the COVID-19 pandemic, a large amount of the CwC work was focused on the response. To streamline and further integrate the RCCE health focused working group with the broader CwC coordination mechanisms, the chair of the RCCE TWG was transferred to the CwC chair. There are differences of opinion about the appropriateness and effectiveness of this structure, and at times disagreements about the structure and leadership have hindered an effective response. There is little involvement of Bangladeshi CSOs or local NGOs in CwC WG meetings, and direct involvement from government in the group was also reported as limited. Despite these issues there remains a high level of demand for CwC and a recognition of its importance in the response.

Methods
This review paper synthesises and assesses the quality of a body of evidence about how the Rohingya seek and access health care services in Cox's Bazar (see Annex 1: List of documents reviewed and appraised). The process comprised a rapid literature review and key informant interviews to identify what has worked well so far, what are the main bottlenecks to progress and which questions require further interrogation.
The literature review focused on synthesising the available data relevant to the Rohingya people's healthcare seeking behaviours and humanitarian response in Cox's Bazar. Academic, programmatic and policy literature was assessed along with published and unpublished documents shared by partner agencies including ACAPS, BBC Media Action, IOM, Ground Truth Solutions, UNHCR, UNICEF, and WHO. Relevant qualitative, quantitative and mixed methods studies were assessed. Data synthesis reports, grey literature and some media sources were also reviewed. Additionally, programmatic documents including training materials, community messaging, community engagement and communication strategies, and coordination Terms of reference were reviewed for context.
Quality assessment was carried out on primary data sources where methodology was transparent and traceable. Quality was assessed using customised appraisal forms developed by drawing upon several previously published tools and checklists (see Annex 2: Quality appraisal strategy and findings). Where quality assessment was not possible, including primary data sources where methodology was not transparent and syntheses and reports using secondary data sources, the findings of these documents have been triangulated with other data sources to determine reliability. Thematic analysis was applied to the data and findings from the literature review and the key informant interviews to identify themes and patterns. The trends that emerged were critically analysed and grouped. Due to the small number of key informant interviews, privacy and confidentiality is protected by presenting all information obtained in these interviews by the generic reference of 'key informant interviews' (KII). Wherever possible, additional sources have been used to verify and support findings from the interviews.

1: What are the issues with some of the evidence?
There is a substantial body of work that has been undertaken discussing perceptions of the Rohingya people regarding their health and health care. However, seemingly conflicting data complicate recommendations being carried out to improve risk communication and community engagement activities at field level. This section sets out some of the data biases identified through the key informant interviews, quality assessment of the available research methodologies and triangulation of data pieces to provide insights into these discrepancies. Quantitative, mixed methods, qualitative, syntheses, grey literature and published and unpublished reports were collated, reviewed and assessed. While research methodologies are often constrained by on-the-ground realities, understanding the ways in which these limitations may affect and bias the data is essential to contextualising and operationalising the findings and making sense of seeming contradictions.

Sampling bias
Rapid research and research conducted in emergency settings often face constraints in sampling techniques. Challenges with access to camps and the structure of the camps with makeshift housing and lack of numbering create barriers to systematic and comprehensive sampling strategies.  Many of the studies included in this assessment presented relatively small samples, did not include all camps, or relied heavily on convenience sampling. While these findings still provide useful information, it is important to understand how the sample and sampling methodology used may be affecting the data gathered and which voices and perspectives may be overrepresented, and which are not being captured.
Overgeneralising or sensationalising findings from small or location-specific samples can result in the magnification of certain findings and the minimisation of others, creating a distorted broader picture. While the Rohingya communities across the various sub-blocks and camps in Cox's Bazar face many of the same challenges and have access to many of the same resources, the landscape is far from homogenous (Humanitarian Response, 2020; ISCG, 2020). These differences can shape local experiences and inform community priorities, perceptions and practices so that data coming from one are or group may not be representative of other areas and groups. A clearer understanding of these nuances and where specifically data is generated from, its creditability and the science behind its analysis including the inferences generated can contribute to more locally tailored, appropriate and effective responses.
Convenience and passive based sampling methodologies may also overrepresent certain voices and perspectives within a community. For example, a convenience sample taken in a local market will fail to capture data from community members who have limited access to the market such as women and people with disabilities. Additionally, feedback mechanisms such as comment boxes, information hubs and Interactive Voice Response (IVR) gather data from community members that have access to those mechanisms and will actively seek to provide feedback. For example, it has been reported that while many health centres have comment boxes, many beneficiaries are unable to write and must rely on staff to write their comments. This may result in more positive feedback being recorded (SSHAP KIIs). While the data gathered from these sources are important, they often fail to capture marginalised voices within a community. Findings from the Rohingya communities in Cox's Bazar have consistently shown that the majority of community members are illiterate, particularly women, that women, the elderly, adolescent girls and people with disabilities have limited access to public spaces, that many do not have consistent access to a mobile phone and that there is not a 'feedback culture' among the Rohingya communities (Holt et al., 2020;ISCG, 2019;Translators without Borders, 2018). Due to these considerations, data sets from convenience sampling and these types of feedback mechanisms likely do not adequately capture the perspectives of women and girls, the elderly, people with disabilities and those that do not actively seek to give feedback.
Some of the discrepancies in the feedback data and studies' findings likely reflect the multifaceted and complex landscape in Cox's Bazar. Different samples and sampling methodology will likely elicit different perspectives and experiences. Transparency around samples and sampling methodology would help to elucidate differences in findings.

Response bias
Accounts of positive response bias was pervasive in the literature and interviews. As Sattler and van Praag noted, "in Bangladesh, our feedback is consistently positive, perhaps telling us more about barriers to open feedback than the findings themselves" (Sattler & van Praag, n.d.). The underlying drivers of observed response bias reportedly include sociocultural factors, structural factors and power dynamics, and limitations in research practices and skills.
The dynamics at play in a humanitarian response have long been recognised as creating power imbalances that can positively bias responses. This is especially true if participants perceive that desired or 'correct' responses will ensure continued access to assistance and resources or that undesired or 'incorrect' responses will elicit punishment. One recent report noted that during consultations specifically eliciting feedback and suggestions, respondents routinely preceded any negative statement with expressions of immense gratitude (ACAPS & IOM, 2021). This tendency was notably higher among female participants in the study. Many of the participants in the study also recounted negative experiences resulting from attempts at reporting issues (ACAPS & IOM, 2021). Community members may also refrain from sharing their thoughts, perceptions and experiences if they think that what they share will be used to pressure them to seek treatments or take specific actions (ACAPS & IOM, 2020b, 2021; BBC Media Action & Translators without Borders, 2020c). In addition, they may also feel that frank and blunt feedback could jeopardise their relationships and affect the services availed from the local agencies. Conversely, a lack of perceived power may also contribute to response bias if community members do not believe that the actor they are speaking to has the power to address their complaints (ACAPS & IOM, 2021; KIIs). Actual and perceived power dynamics at play may be different for various actors across the response, which may account for some of the differences in data observed.
A bias towards positive responses among the Rohingya can also be understood in the context of their long history of oppression and abuse at the hands of authorities. There is reportedly widespread scepticism among the Rohingya about why data is being collected and how it may be used (SSHAP KIIs). Among the Rohingya an adaptive behaviour of saying the 'right thing' or hiding information has been identified as a community protection strategy. In this way, 'correct' responses are not only a way to avoid individual consequences of a 'wrong' answer, but as contributing to the safety of the larger community. As a community safety issue, the behaviour is reinforced by community norms and stigma against those who break it (ACAPS & IOM, 2020e, 2021). For example, while community members reported that they would seek care and COVID-19 testing if presenting with symptoms, research later in the outbreak revealed that potential cases were actively hidden to protect people because they feared being forcibly taken from their families and killed in quarantine (ACAPS & IOM, 2020b, 2020e, 2021; BBC Media Action & Translators without Borders, 2020c; Ground Truth Solutions et al., 2021). These fears are not groundless rumours but based in experiences of the Rohingya with health systems and authorities whilst in Rakhine, reinforced by experiences of public health measures like mandatory isolation for positive COVID-19 tests and quarantine while awaiting test results early in the pandemic in Cox's Bazar. Understanding how this history continues to inform engagement with authorities, perceptions of community risk, and community strategies for protection can help to contextualise apparent discrepancies in the data and feedback gathered.
Current tensions and social dynamics have also been identified as affecting response bias. Several key informants and studies mentioned there were reports from Rohingya people of discriminatory treatment towards them by some humanitarian and health sector workers. It was thought this may contribute to distrust and response bias when data are collected by non-Rohingya researchers and enumerators (SSHAP KIIs). These broader social dynamics mean that Bangladeshi interlocuters face greater barriers to engaging participation and establishing trust than Rohingya interlocuters (REACH, 2019). Multiple studies have found marked differences in community responses to Rohingya versus Bangladeshi enumerators using the same research tools and translation in the same communities (Holt et al., 2020;REACH, 2019; GTS/IOM, 2021 and SSHAP KIIs). Differences in responses were greatest on perception-based or sensitive topics such as questions concerning mistreatment and disrespect by Bangladeshi aid workers.
Research methodologies and practices can also affect participant responses. The Rohingya have consistently expressed a preference for face-to-face, open-ended format methodologies and are more inclined to share information when engaged in this way (ACAPS & IOM, 2021). Different methodologies may also impact who is willing to participate in a study at all, particularly if participants feel singled out (such as through random sampling) or identifiable (SSHAP KIIs). This may account for some discrepancies observed between research using different types of methodologies. Additionally, good research practices such as creating safe spaces, ensuring confidentiality and privacy and obtaining understood informed consent can be challenging to implement in crowded and emergency contexts. However, lapses, or perceived lapses, in these areas can affect the responses that various participants are willing to give. For example, there have been reported challenges of interviewing women and girls who are more apprehensive about being recorded due to safety and privacy concerns (ACAPS & IOM, 2021).
Researcher skill level and competency also affects response bias. Researchers' competency on Chittagonian and Rohingya languages of the community will affect their ability to develop confidence, gain accurate data, avoid misunderstandings and develop rapport with participants. It has been frequently reported that the majority of enumerators speak Chittagonian. While Chittagonian and Rohingya are similar, Chittagonian is not a suitable substitute for complex topics and reliable dialogues (Translators without Borders, 2018; SSHAP KIIs). Beyond being proficient in the local language, researcher competency in the appropriate meanings and customs of the community will affect their understanding of the data and the trust they are able to build. This includes understanding the appropriate local terminologies and meanings, greetings and titles. Additionally, if researchers are unskilled in active, unbiased and respectful listening, and building trust with participants this may limit the responses they are able to obtain.
The assessment of research (from the document review and the key informants) in Cox's Bazar has highlighted that response bias is widespread. How forthcoming and open a participant will be during a given research interaction will be shaped by several factors. Individual and community experiences, the sensitivity of the topic, who is conducting the research, where is the research conducted, who else are in the sampling frame, the research practices and assurances in use, and the skills and mannerisms of the research will all contribute to response bias. Understanding the drivers of response bias can improve understandings of how response bias may be impacting data and why different data sets seemingly tell different stories.

Confirmation bias
Humanitarian aid and emergency responses around the world are under financial and political pressure to perform and demonstrate justifiable need and results. These pressures are reportedly high in Cox's Bazar (SSHAP KIIs). This can be a large driver of confirmation bias in how data is received and managed. Several of our key informants reported that there tends to be a greater acceptance of data findings that are in line with expectations or are considered 'positive' without the additional scrutiny applied to more negative data findings (SSHAP KIIs). It can also encourage agencies to accept the data on face value in the interest of time and funding. Inconsistencies or lapses in data oversight and transparency were also identified as contributing to seeming inconsistencies in the findings. Additionally, there have been reported instances of researcher and enumerators forging data that they believe is what will their managers want them to report (SSHAP KIIs). Confirmation bias and perceived incentives for 'positive' data may explain some of the discrepancies in data and findings.

Transparency
Unfortunately, many of the reports examined did not describe the methodology, tools, or limitations, making it difficult to ascertain the validity and limitations of the findings (see Annex 2: Quality appraisal strategy and findings). Some key informants also noted that datasets are not consistently made available which limits the ability to double check the validity and rigour of the analysis and findings (SSHAP KIIs). Lack of transparency and accountability in these areas can obscure limitations and context making it challenging to fully understand potential biases in the findings. It also limits awareness of partners and the government authorities and the potential to explore synergies and work for coherent actions due to the scarce information on interventions, and their outcomes.

2: Major themes and areas of conflicting information
A large body of literature along with humanitarian response evaluation and community feedback reports have sought to shed light on the Rohingya people's health and healthcare related perceptions and behaviours. Despite this abundance of data, seeming contradictions and unexplained variations are commonly reported in findings on Rohingya community trust in the humanitarian response and actors, healthcare seeking behaviour, acceptance and compliance with health prevention measures, and community engagement and feedback mechanisms. This section explores the major themes, and in particular, areas of conflicting data in light of the issues and biases set out in section one to better understand some of the identified drivers and gaps. However, variations and contradictions in findings about trust and distrust should also consider that the response is not uniform, and neither are the communities. As noted in one study, 'participants commonly stated that generalising and discussing an entire group (those providing them with assistance) as all the same could be misleading, as not everyone is the same' (ACAPS & IOM, 2021). Trust may vary over time, organisation to organisation, response actor to response actor, community member to community member, and situation by situation. Examples of these variations can be seen in different rates of negative experiences reported with non-Rohingya humanitarian staff versus Rohingya volunteers, requests for foreign doctors and Rohingya volunteers at health facilities because they are more trusted than Bangladeshis, and the widespread hiding of COVID-19 symptoms from actors across the response (ACAPS & IOM, 2021; Holloway & Fan, 2018). As these examples demonstrate, the underlying drivers for lack of trust may be context or person specific and tied to historical and current experiences of discrimination and disrespect, communication barriers, lack of consultation and inclusion, and poor treatment and services (ACAPS, 2020; ACAPS & IOM, 2020d; Holloway & Fan, 2018). Further rigorous and transparent research processes that unpack these drivers are essential for understanding variations in patterns of trust across communities in Cox's Bazar.

Healthcare seeking behaviour
Data findings indicate a discrepancy in the reporting of healthcare seeking practices and the reality as well as variations in healthcare seeking behaviours depending on the health issue and community group in question. Research findings showed 89% of the key informants in one study reported that members of their community would seek in-person care if presenting with symptoms of COVID-19 (Ground Truth Solutions, 2020). However, even though research indicated that at least one wave of COVID-19 had gone through the camps, the actual rates of people accessing healthcare services from the facilities dropped significantly in the early months of the pandemic compared to 2019 (Ground Truth Solutions, 2020; ISCG, 2020). Similar observations were noted with routine immunisation. Qualitative research conducted in the community elucidated these discrepancies with findings that communities were actively and intentionally misleading the response and hiding potential cases of COVID-19 during the early phase of the response (ACAPS & IOM, 2020b). Although rates of healthcare seeking have reportedly recovered (for some services even beyond baseline levels), these various findings from the initial response reflect a change in healthcare seeking behaviour and participant reporting in response to the pandemic and the response measures.
Seemingly contradictory data in healthcare seeking patterns also predates the pandemic. According to both the 2020 and 2019 Joint Multi-Sector Needs Assessment (MSNA), the vast majority of the population reported seeking medical care when needed (94% in 2020 and 97% in 2019) and most reported seeking care at the free NGO run health clinics (64% in 2020 and 79% in 2019) (ISCG, 2019(ISCG, , 2020. However, these data contradicted studies that found that Rohingya households preferred paid health services and high rates of reported negative coping mechanisms such as paying for and going into debt for health care(ACAPS, 2020; BBC Media Action & Translators without Borders, 2020d; Ground Truth Solutions, 2020; ISCG, 2019, 2020). While data bias may account for some of this discrepancy in data, 'this pattern of seeking additional, or alternative, paid healthcare, when free health care is readily provided, strongly suggests that Rohingya either do not trust the health care they are receiving in the camps, do not perceive it to be appropriate or adequate to meet their needs', or face barriers in accessing it (ACAPS, 2020). However, as there are over 80 organisations engaged in the provision of healthcare in the camps with varying degrees of resources, capacities and community trust, healthcare seeking behaviour for NGO facilities are variable (Health Sector Cox's Bazar, 2020b; SSHAP KIIs). Other reported sources of care accessed included private health clinics (26%), pharmacies or drug shops (20%), government clinics (6%) and traditional or community healers (1%) (ISCG, 2020). The expectation for specialised health services have always led to unmet needs among the Rohingya Refugees in the camp.
Data patterns on healthcare seeking behaviour also vary according to health care service. For example, while rates of health care utilisation have rapidly improved in nutrition and immunisation, they have been improving at lower rates for antenatal care and sexual assault cases (Health Sector Cox's Bazar, 2020a; ISCG, 2019; SSHAP KIIs). Gendered differences in healthcare seeking behaviour have also been recorded. A recent study found that among adolescent participants in camps, boys were significantly more likely to go to NGO clinics than girls (40% and 25% respectively) (Guglielmi et al., 2020). Understanding the drivers of both the contradictions and variations in healthcare seeking behaviour is critical to designing health programmes that meet community needs and improve health outcomes.
Understanding discrepancies and variations in healthcare seeking data must take into account community understandings of disease and illness, accessibility and sociocultural norms, and trust and experiences with the health systems. How community members understand a particular disease or illness will inform their healthcare seeking behaviours. For example, some views that COVID-19 is a punishment from God may result in stricter adherence to religious customs that limit women's access to health clinics (ACAPS & IOM, 2020a). In Rohingya communities women and girls practice purdah (strict gender segregation) and it is considered undignified to be exposed to men outside of their family or be seen in public. This is particularly true for young and unmarried women (ACAPS & IOM, 2021;Holloway & Fan, 2018;ISCG, 2020). Coping mechanisms to maintain dignity reportedly include limiting time outside the home and use of essential services, including health services (ACAPS & IOM, 2021). Women and girls may also need permission from men before seeking healthcare services (Guglielmi et al., 2020). It has been noted that a lack of privacy and insufficient numbers of female attendants during medical consultations along with limited confidence of privacy of the personal medical history mentioned by the health facilities are large drivers of the lower rates of healthcare seeking among women and girls.(SSHAP KIIs) Additionally, trust and previous experiences with health systems inform individual cost-benefit analysis for accessing health care services and engaging with the response and may differ for different issues (ACAPS & IOM, 2020e). For example, traumatic experiences with human rights violations, including forced detainment and government sponsored population control efforts, have reportedly shaped community willingness to present to health facilities with COVID-19 symptoms or to seek sexual and reproductive care services (ACAPS & IOM, 2020d; Ahmed et al., 2020;Parmar et al., 2019). For any given situation, community members may decide that the risk to themselves or the community is too high, that the care is inadequate or inappropriate, the wait too long or the treatment too poor. Other services may be more convenient, easier to use, or perceived to be more effective (BBC Media Action & Translators without Borders, 2020d). As with most communities, healthcare seeking behaviour may differ from person to person and situation to situation. Further exploration of drivers of healthcare seeking behaviour in relation to various groups and issues is necessary to contextualise and understand contradictions and variations in healthcare seeking behaviour data.

Rohingya patient experience
There is a spectrum of data about Rohingya experiences at health facilities. Some findings showed predominantly positive and satisfactory experiences, while others reported majority negative experiences and perceptions, and others highlighted very mixed experiences (ACAPS, 2020; ACAPS & IOM, 2020d; Guglielmi et al., 2020;Masud et al., 2017). The mechanism and methodology used to obtain the data, the sample of the population engaged with, who conducted the research and how, which health centre was accessed, and which physician was seen can all shape the types of responses elicited. Improvements in the health services and community engagement from lessons learned may also be reflected in changes in the data over time. However, based on the issues affecting data, it is likely that the findings here generally tend to be skewed towards the positive. Importantly, even if positive experiences are more frequent, negative experiences tend to spread more quickly through communities with negative impacts on healthcare seeking behaviour and trust (ACAPS & IOM, 2020e; BBC Media Action & Translators without Borders, 2020d). Understanding the underlying factors contributing to negative experiences in health centres is essential to continuing to improve & scale up the services and accountability.
Factors reported as contributing to negative experiences included discrimination and disrespectful treatment, communication barriers, perceptions of unsatisfactory consultation, and inadequate health care capacity. Both the literature and several key informants highlighted that Rohingya who reported negative healthcare experiences frequently described discriminatory or disrespectful treatment by health care and support staff. Reported grievances include being turned away by staff without explanation, staff members being rude or yelling, long waiting time, not being greeted in a kind or appropriate way and not being listened to (Guglielmi et al., 2020;Holloway & Fan, 2018;SSHAP KIIs).
In addition, Rohingya have reported being treated better when foreigners visit the health centres to assess the services of the health facilities and observe conditions and treatment (SHHAP KIIs).
Communication barriers were also commonly reported. Research underway by Translators Without Borders is finding that a large percentage of health care workers believe they are communicating more effectively than they actually are and many don't understand key words of more complex issues. SSHAP KIIs). Communication barriers resulting in patients not being able to describe their symptoms adequately or to clearly understand the diagnosis, prescribed treatment procedures and instructions also contributes to frustrating experiences and distrust in the care and guidance provided (ACAPS & IOM, 2020d, 2021Humanitarian Response, 2020). Dissatisfaction with the consultation, including not getting to see a doctor and lack of expected testing and physical examinations, also contributed to scepticism about diagnoses and medical advice (ACAPS & IOM, 2021; SSHAP KIIs). Additionally, while there is a standardised minimum package of essential health services expected to be provided at each health centre which is overseen by the government and health sector (Health Sector Cox's Bazar, 2020b), the wide variety in resources and capacity between service providers and gaps between community health needs and services persist (SSHAP KIIs

Knowledge, awareness and practice
Findings about community knowledge, awareness and practice of health prevention behaviours also presents seeming contradictions. For example, data collected on COVID-19 prevention measures showed that community members generally had some awareness and understanding about the prevention measures and that awareness and knowledge improved over time with the response. However, practice of prevention measures remained low in the communities (BRAC University, 2021). While data from the 2020 Joint Multi-Sector Needs Assessment found that 98% of households reported wearing a facemask to prevent themselves from getting COVID-19 and a WHO survey reported 83% of respondents indicated that they always wear a mask everywhere outside of the home, observed mask wearing contradicted these numbers (ISCG, 2020; UNHCR & WHO, 2020; WHO, 2020). These discrepancies demonstrate that community members are receiving and understanding the messages and advice being given to them as they can give the 'correct' responses when asked, but for various reasons are not putting that information into practice. Bridging this gap by responding to barriers to behaviour change and motivating communities to put the lessons conveyed in the messages into practice is necessary for more effective health promotion in Cox's Bazar.
Identified underlying factors contributing the observed contradictions between knowledge and awareness and practice include community understandings of risk, sociocultural and religious norms, and structural and logistical barriers. Risk perceptions and understanding the need for and importance of adhering to public health measures behaviours inform community behaviour. While many Rohingya reportedly understood that cleanliness and physical distancing were important for disease prevention, and these messages align well with Islamic teachings, gatherings at mosques for prayers were not understood to have the same transmission risks as other activities. Mosques are seen as clean places and prayers were viewed as protective against infection (ACAPS & IOM, 2020c). In addition, Rohingya women reportedly view their traditional face covering (particularly the nose) as sufficient to protect them from infection (ACAPS & IOM, 2020c; WHO, 2020). Structural and logistical barriers can also result in discrepancies between knowledge and behaviours. For example, not having access to the resources needed to comply with health measures (BBC Media Action & Translators without Borders, 2020b;Ground Truth Solutions et al., 2021). Mandatory mask policies that did not account for the lack of masks in the early stages of the outbreak reportedly resulted in community members sharing masks at distribution sites and increasing transmission risks (SSHAP KIIs).

Community engagement and feedback mechanisms
Findings related to community engagement and feedback mechanisms also present seeming contradictions. Most community members report knowing how to make suggestions (64%), having no barriers to providing feedback (96%), and that aid providers take their opinion into account when providing aid (80%) (Humanitarian Response, 2020; ISCG, 2020). In contrast, a large qualitative study by ACAPS and IOM (2021) reported that participants in approximately half of the focus group discussions related negative experiences when trying to report problems and issues in the camps and women frequently reported not knowing where to go to provide feedback (ACAPS & IOM, 2021). Another study found that some Rohingya are reluctant to raise concerns at the Information Centres as they fear their aid may be stopped if they complain . Survey data also indicated that almost half of households reported that they rarely or never felt consulted about their needs and preferences (ISCG, 2020). The Rohingya community in general has not been systematically included in needs assessments and women and vulnerable populations, including transgender and monolingual community members, have been found to be less likely to be consulted and face greater challenges to providing feedback (ACAPS & IOM, 2021;Heward & Carrier, 2020;Humanitarian Response, 2020). Community and religious leaders, volunteers engaged with the response, literate community members, and elders were identified as more likely to be included in consultations and decision-making, while women and girls, youth and illiterate people were excluded (ACAPS & IOM, 2021). Positive response bias and sampling bias appear to be affecting some of the data around community engagement, particularly survey data. However, the findings also point to important experiences and perceptions of inclusion and community engagement necessary to building trust and aligning response efforts with community priorities and norms. Community inclusion in various levels of the response has also been identified as an important aspect of community engagement, self-reliance and agency, and dignity. While findings highlight that community members are more comfortable interacting with Rohingya volunteers, and better data is often obtained when engaging Rohingya interlocutors, findings also point to frustration with the limited capacity Rohingya volunteers have to enact positive change and address issues and problems along with answers to follow up queries (ACAPS & IOM, 2021). In focus group discussions, many community members linked this to the fact that Rohingya are not placed in management or decision-making roles. The expressed frustrations were compounded by lack of transparency and understanding about how decisions within the response are made and persistent bottlenecks and barriers that limit which problems can be solved (ACAPS & IOM, 2021).

3: Lessons learned and key recommendations that are often made
Recent accounts point to improvements in terms of community inclusion, community engagement, information provision, investments in translation, training of volunteers, coordination of messaging and feedback, and mainstreaming of CwC and accountability mechanisms into partner operations, however, there continue to be areas that need improvement. (Sida & Schenkenberg, 2019) This section sets out lessons learned in communication, community engagement and accountability strategies, data collection, and coordination. It also provides some of the key recommendations for improvement that are commonly made in the literature and by stakeholders interviewed for this project.

Communication, community engagement and accountability strategies
The "risk communication" aspect of RCCE, or one-way messaging, was generally considered by respondents to be performing well in Cox's Bazar.(SSHAP KIIs) In the context of COVID-19, evaluation studies also show that people are receiving and understanding the messages disseminated, through community volunteers, "miking" and even social media (Ground Truth Solutions et al., 2021; UNHCR & WHO, 2020; WHO, 2020; SSHAP KIIs). However, there is a general agreement among the humanitarian sector and refugees, that one-way information provision is not enough. What is needed is two-way engagement that provides refugees with the opportunity to engage with the material provided, ask questions, and contribute their own experiences, concerns or ideas. Community engagement should be further invested in and prioritised (Sida & Schenkenberg, 2019;SSHAP KIIs). Key recommendations for the improvement of community engagement, leading to increased increase accountability and build trust in Cox's Bazar include: use of appropriate engagement methods for specific target groups, with a focus on two-way communication; increase the inclusion of the Rohingya in the decision-making, planning, roll-out and monitoring of the response; engage trusted interlocutors; and, use appropriate language and format.

Using appropriate engagement methods
There is a sense that current strategies, predominantly visiting people's houses to deliver information or administer surveys, fall short of effectively and meaningfully engaging Rohingya communities. Studies and evaluations suggest that not enough is being done to understand the cultural, linguistic and religious dynamics and nuances of communicating with the Rohingya or adopting engagement strategies that makes them feel comfortable and respected, and which invite open dialogue (Sida & Schenkenberg, 2019;SSHAP KIIs). It is also important to be aware of communities' priorities and provide information that is relevant to them, and to frame information in a way that makes sense in the context of an evolving scenario (e.g. COVID-19) (SSHAP KIIs). This may mean, for example, framing health information from a religious perspective and in line with Islamic teachings to achieve better compliance with circulars issued by the Government and disseminated through the partners ( Using face-to-face communication became more challenging during COVID-19 necessitating more reliance on other methods such as "miking" and provision of audio recordings through high volume speakers (BBC Media Action & Translators without Borders, 2020a). However, the lack of more twoway communications meant that while communities received the same messages repeatedly, there were few ways for them to ask further questions. This has reportedly lead to rumours that quickly spread through the community (SSHAP KIIs). Some studies have advocated for making use of social media platforms and mobile phones (BRAC University, 2021;Joarder et al., 2020). However, there is limited access to mobile phones and the Internet, as well as a low level of literacy among the Rohingya (Ground Truth Solutions et al., 2021). Interactive Voice Response (IVR) is a relatively new communication mechanism that provides pre-recorded messages in local languages and collects community feedback through calling campaigns, and appears to have been well accepted according to feedback received (Holt et al., 2020;SSHAP KIIs). Pictorial and video messages have also been found to be useful for providing information for people with low literacy levels (BRAC University, 2021; Heward & Carrier, 2020). However, these methods still face access issues. Diversifying communication channels has been recommended to reach the greatest audience, using channels such as radio, TV, art and theatre groups (Heward & Carrier, 2020).
Engagement of vulnerable groups, such as women, the elderly, trans people and people with disabilities, present additional challenges and may require targeted communication strategies. Their mobility and access to public spaces, levels of education and literacy, communication preferences and access to different communication technologies need to be assessed and taken into account when designing strategies (ACAPS & IOM, 2020c; BRAC University, 2021). Stigmatisation of disability adds another dimension, potentially influencing people's willingness to attend community engagement activities (SSHAP KIIs). Mobile outreach teams have been recommended as a way to reach these audiences, to understand their healthcare seeking barriers and enablers and engage them in decision-making (Guglielmi et al., 2020).
With regard to women, the practice of purdah can have implications for women's ability to access information and engage in activities outside the home, as well as their access to healthcare (Ripoll, 2017). Women in the camps generally have limited mobility and rely on male family members to make decisions about healthcare seeking, and often to provide access to information about health (BRAC University, 2021). Many women have been found to prefer face-to-face dialogue within their own homes, and have called for humanitarians to make more frequent block visits (BRAC University, 2021; Ground Truth Solutions et al., 2021; SSHAP KIIs). Some alternative communication tools such as transparency boards and women-friendly spaces are thought to have had some success in reaching women (Heward & Carrier, 2020). Another suggestion is women's groups, and mother and daughter activities (Aelbers et al., 2018;Ripoll, 2017). Mobile outreach teams and community health workers are also recommended for engaging married and unmarried adolescent girls (Ripoll, 2017). UNICEF data shows that 70 to 75% of the users of its Information and Feedback Centres are women, and that most of these women use the centres for health referrals , however, it is not clear what percentage of women are using the centres in total. Some evidence has been presented that women rely on information shared by community leaders such as Mahjis and Imams more than men do (UNHCR & WHO, 2020), however, it has also been noted by partners that women feel less comfortable than men in communicating through these actors, and that they receive information through them indirectly (SSHAP KIIs). Their reliance on these actors may be because they have fewer other options for information access. In any case, it is recommended to engage these influencers when wishing to share information with women, whilst also looking for alternative communication channels. It should also be noted that, they only have information issued by government authorities and might not have answers to respond to queries.
Women were reported to generally feel more comfortable speaking to other women, and female religious teachers are also now being engaged as community leaders (SSHAP KIIs). It has also been recommended to hire more female staff and community outreach workers (BRAC University, 2021; Heward & Carrier, 2020). While the employment of women has created some social tension, Imams have had some success in shifting social norms that act as a barrier to women's involvement in volunteer or paid work. Imams who are supportive of women's ability to work outside the home have been able to explain to their communities that from the perspective of religious teachings, there is no impediment to women working (ACAPS & IOM, 2020c). Framing this from the perspective of Islamic teachings was much more effective than if humanitarians had framed it from an equal rights and gender perspective (SSHAP KIIs).
For their own health literacy, but also as important decision-makers and gatekeepers with regards to women's health and healthcare seeking practices, men and boys also need to be engaged in discussions about reproductive and sexual health (Parmar et al., 2019). It has been found that adolescent boys find it difficult to attend youth activities due to work commitments, so attempts should be made to design strategies that suit this groups needs and availability (Ripoll, 2017). Schools, community spaces and dedicated adolescent-friendly spaces have been suggested as possibilities (Guglielmi et al., 2020). Studies have found that while women prefer face-to-face communication, including focus group discussions, men and young boys prefer to receive information through the workplace and messages through TV, megaphone or mobile phone, as well through face-to-face interactions (Ripoll, 2017).

Increasing the inclusion of Rohingya refugees in the response
There is growing recognition that all aspects of the response will be more effective if Rohingya are involved as central actors and consulted on the decisions that concern them. As discussed above, there is emerging evidence that Rohingya trust, feel more comfortable with and respond differently when asked questions by Rohingya than by Bangladeshi or other humanitarians (BRAC University, 2021). It is thought that the use of Rohingya refugee community health workers by UNHCR was vital in tackling outbreaks such as the Diphtheria outbreak in 2017.(UNHCR, 2019) Likewise, Bangladesh Red Cross Society volunteers are thought to be very well trusted, as they are from the community and have a strong network (SSHAP KIIs).
With a view to improving health outcomes and accountability, Rohingya could be incorporated into many aspects of the response, including awareness raising, outreach, data collection, monitoring and evaluation and delivery of health services (ACAPS, 2020; SSHAP KIIs). Rohingya refugees have requested more Rohingya volunteers working with humanitarian organisations and at health facilities (ACAPS, 2020; ACAPS & IOM, 2020b), and have stated that there are many willing, qualified and educated Rohingya who could fill most positions in the camps (ACAPS & IOM, 2021). As well as building trust in the health system, this would enable Rohingya volunteers to contribute to making these services more socially and culturally appropriate for Rohingya, through suggested structural changes and through training of healthcare workers on culturally sensitive service provision (ACAPS, 2020; Guglielmi et al., 2020). Note that attempts should be made to also engage female Rohingya volunteers (Guglielmi et al., 2020).
Potential ways to foster this engagement could be to support already-existing community structures, or to promote the development of Rohingya-led community-based organisations (Sida & Schenkenberg, 2019;SSHAP KIIs). There are structural challenges associated with recruitment of Rohingya, which will be discussed in a later section.

Engaging trusted interlocutors
Beyond engaging Rohingya as community engagement actors, some lessons have been learnt about other specific groups of trusted or important community actors who have been or could be engaged to enhance health-related communication with Rohingya. There is some evidence that religious leaders, such as Mullahs, Imams and Mulvis, are trusted and well respected as sources of information (Aelbers et al., 2018;SSHAP KIIs). There has recently been more of an effort to engage with religious actors by organisations such as UNHCR, in communication activities similar to those conducted by volunteers and other partners. Religious leaders are provided training on communication and community engagement, and given equipment such as speakers and projectors to share information they consider relevant with their communities and visitors to mosques (SSHAP KIIs). It should be noted, however, that it is not conclusive as to whether men and women trust these actors equally (UNHCR & WHO, 2020).
There is conflicting evidence as to whether Mahjis are trusted among the community, as they were not elected, but appointed by the Bangla military during the influx (Ground Truth Solutions et al., 2021). Nonetheless, they continue to be an important source of information for Rohingya, who often depend on them to represent their interests. In some camps, there are ongoing attempts to reform this system through community representative elections. This is thought by partners to be a positive change but takes time to implement due to 'political complexities' (SSHAP KIIs). In the meantime, it was recommended to ensure Mahjis have access to accurate information in order to fulfil their roles, whilst also using other interlocutors as appropriate (Aelbers et al., 2018;Ground Truth Solutions et al., 2021).
Community Health Workers/Volunteers (CHW/Vs) have reportedly been very effective in disseminating messages and improving uptake and coverage of some health services. Male and female CHW/Vs are engaged to cover issues related to health and hygiene promotions, refer community members to services, liaise with other volunteer groups to ensure referrals to services are streamlined, conduct community surveillance and handle sexual and reproductive health issues and education (UNHCR et al., 2018). CHW/Vs have been credited with improving acceptance and uptake of maternal health services which have currently reached 70% in the camps (Community Health Working Group, Cox's Bazar, n.d.; SSHAP KIIs). In addition, each CHW/V covers roughly 750 population. This allows for extensive one-on-one coverage of the population. CHW/Vs have recently been engaged in immunisation efforts to reach children who had missed their vaccinations, resulting in greatly improved coverage (SSHAP KIIs). As part of the COVID-19 response, CHW/Vs were engaged by health sector partners in community outreach initiatives to disseminate messages on COVID-19 and essential health services. These joint efforts are thought to have resulted in increased daily COVID-19 testing rates from less than 10 to several hundred per day during peak periods (Health Sector Cox's Bazar, 2020a). CHW/Vs have also been instrumental in surveillance and community feedback mechanisms (SSHAP KIIs). It has been noted by key informants that CHW/Vs are an incredibly effective resource that could be scaled up and expanded (SSHAP KIIs).
Alternative healthcare service providers, such as ayunku (people providing medicines or treatments without formal medical training), Rohingya doctors, pharmacists and Traditional Birth Attendants, are also important sources of information for some Rohingya, and should be engaged as important health influencers (Parmar et al., 2019;Ripoll, 2017;SSHAP KIIs).
As discussed, preferred interlocutors may be different for men and women. Women hafés (women who have memorised the Quran) and midwives may be appropriate interlocutors for women, especially around sexual and reproductive health. It is also important to consider that mothers-in-law hold considerable influence in household decisions, especially around maternal health, and are an important group to engage if wanting to reach and influence the behaviour of younger women of reproductive age (Ripoll, 2017).

Using appropriate language and format
Although Rohingya language and the local language of Chittagonian are linguistically similar, there are key differences in some terms, and ample possibilities for misinterpretation between the two. Partners have recommended that all verbal communications with Rohingya should be in the Rohingya language, and that materials should be developed using Rohingya terminology (SSHAP KIIs). It has also been recommended that Chittagonian speakers hired to act as interpreters for Rohingya refugees be tested for Rohingya language competency, and provided with training in interpreting skills, as well as in specific Rohingya terminology and in social and cultural awareness relevant to communicating respectfully with Rohingya (Translators without Borders, 2018). Written translation to Rohingya should be done by local community members who have an in-depth understanding of local meanings and nuances, and Rohingya community members should review the text to ensure the translation is accurate (SSHAP KIIs).
Given the low levels of literacy among Rohingya and the fact that Rohingya does not have a universally accepted script, written text materials are less effective than audio or visual materials, especially for women, who are less likely to be literate (Internews, 2017;SSHAP KIIs). Other suggestions for potentially effective communication formats are traditional poems or music, such as Tarana songs. It could be useful to identify musicians or poets in the camps who could be engaged to work on messaging in these formats (Ripoll, 2017).

Data collection
As mentioned in Section 1, the quality of the data collected is greatly influenced by the data collection method. A common theme emerging from the key informant interviews was the lack of standards in data collection and research ethics. In particular, four key recommendations emerged: use Rohingya researchers wherever possible; improve researchers' skills; use appropriate research methodologies; ensure quality oversight mechanisms are in place.

Rohingya researchers
The COVID-19 pandemic forced partners to rely more on people living in the camps to conduct research. Due to recent findings about response bias, there is a growing recognition of the importance of using local researchers, and of the advantages in terms of language, history and trust. Some agencies and organisations are now investing in restructuring their data collection mechanisms to account for this.
Where it is not possible to make large-scale shifts in the short-term, it is recommended that Bangladeshi researchers be given additional training in the nuances of the Rohingya language and in culturally appropriate behaviour and communication methods, such as respectful greetings, to build trust and improve data collection outcomes with the Rohingya.

Research skills trainings
All researchers, whether Rohingya or Bangladeshi, would benefit from ongoing, experiential-based training to improve their research skills. Methods that are considered to have worked well include weekly training sessions to researchers, focusing on interviewing, listening, facilitating, meaningful informed consent and creating safe spaces for interviewing.

Appropriate methodologies
There is a growing understanding that research methodologies need to be adapted to the community's needs and comfort levels. It has been found that the data provided is different when respondents are allowed to remain anonymous and not provide their name or address, when interviews are conducted in the places and at the times communities prefer, when researchers take the time to fully explain who they are and what the data is intended to be used for, when researchers invest time sitting and chatting with participants in a relaxed manner, and when the focus is on listening and learning rather than providing information. It has been recommended to focus more on participatory approaches to gathering feedback, or opportunistic approaches, such as having social mobilisers record the concerns and questions they hear about during their daily activities (Aelbers et al., 2018;Heward & Carrier, 2020). During the COVID-19 pandemic, efforts were made to continue and introduce multiple methods for collecting feedback, including radio listener groups, other community groups, compilation of rumour bulletins, WhatsApp and social media. UNHCR have started using IVR (Interactive Voice Response), which is considered to have been a highly successful mechanism for providing information, but also allows participants to leave a voice message with their feedback. Although designed to solve the problem of restricted access during the pandemic, this mechanism has the advantage of being able to reach people who may not be able to access information hubs, such as the elderly or people with disabilities. However, it is restricted to those who have access to phones, which excludes many women.

Quality oversight
It is recommended to ensure quality oversight mechanisms are in place, including reviewing of transcripts, reviewing and questioning data with a view to identify sampling or response biases, and listening to the audio to assess whether true informed consent was provided. Further, it is recommended that programme evaluations should use a community-based, mixed-methods approach that incorporate ongoing community feedback from a broad representation of the community (Parmar et al., 2019).

Coordination strategies
There are ongoing efforts to improve coordination of partners working on RCCE in Cox's Bazar. In some areas this appears to be working relatively well, such a coordination of messaging. Recommendations made for further improvement include: streamlining and clarifying the coordination structure; including more diversity of voices in Working Groups and decision-making processes; and, improving coordination of specific areas of RCCE, such as volunteer networks, data collection and sharing and referral of data and feedback.
Streamlining and clarifying the coordination structure While some partners point to the successes of the coordination mechanisms of the CwC and RCCE WGs, others mention that the structure continues to be complex, disorganised or confusing. Partners reported being unclear on roles and structure, being unsure when different entities were meeting for which purpose, and not feeling confident that feedback they provided would reach the right place and be acted upon. Previous studies have found that partners agree on the current arrangement of the CwC Working Group as a free-standing, cross-sectoral platform, but views diverge as to where the leadership of this group should lie. This divergence of opinion is grounded in the group's history of changing composition, which created discord among some partners that continues to resonate. It was also mentioned that the group is too governed by UN agencies and should be led or co-led by an NGO or INGO, while UN staff often felt that a UN agency (i.e. their own) should lead.
It is a challenge for the CwC Working Group, as a multi-sectoral body, to maintain a balance between the different entities and stakeholders, including the different sectors, and to enable equal participation of all parties. Consulting many stakeholders, each with a different focus, can take time. It is also thought to be important to forge stronger relationships with different stakeholders, such as the government and Civil Surgeon (the health manager for the government in each district).(SSHAP KIIs) Key informants interviewed consider that some aspects of coordination are working well. The overall CwC Working Group strategy for COVID-19 included the participation of all members, was well consulted and reviewed by the relevant sectors and endorsed by the government.(SSHAP KIIs) Some implementing organisations coordinate closely with the health sector, liaising with them to access information about health gaps to guide their activities (SSHAP KIIs). There has also been some encouraging regional coordination, with a RCCE Technical Working Group established in the Asia Pacific region. During the COVID-19 pandemic, there has been ongoing discussion and sharing of information about challenges and lessons learnt from the different countries. There was also ongoing communication and information sharing with the CwC Working Group in Myanmar, led by UNICEF (SSHAP KIIs).

Including more voices in WGs and decision-making processes
There are differences of opinion about the number of partners that should be included in the RCCE Working Group. Currently, a small number of organisations with specific technical expertise in the area are included, while previously a larger number were involved. Some interviewees feel it would be appropriate for more of the implementing partners to be involved, to provide feedback to the group on what is and is not working on the ground and in different contexts, and to ensure a direct link between decisions made and implementation. Partners thought to be missing from these spaces include INGOs and NGOs, CSOs and Rohingya community members, and government partners (SSHAP KIIs).
There were also conflicting perceptions of the levels of government ownership and support for CwC and RCCE. While some key informants noted the lack of government staff in meetings and indicated government support had been declining, others credited the continued support of the government for enabling the rapid response to COVID-19 and other emergency situations (SSHAP KIIs). Although government partners rarely attend CwC or RCCE WG meetings (their office is thinly staffed), all RCCE messages are submitted to and signed off on by the Civil Surgeon prior to dissemination in the community to enable accountability, gain legitimacy and avoid duplication of messages. However, it was reported that government involvement and inclusion in two-way communication mechanisms is limited (SSHAP KIIs). Improved mechanisms for feeding back and presenting community perceptions to the government have been suggested as an avenue to advocate for more effective and evidencebased programming and policies (SSHAP KIIs). This could be done by having a point person from the RCCE TWG or CwC WG provide the government with regular briefings on key issues, supporting data and proposed solutions emerging from the working groups (SSHAP KIIs).
Few national NGOs participate in the Working Groups, and language has been cited as a major barrier to the inclusion of Bangladeshi nationals, as all meetings, presentations and handouts are in English . It is thought that more could be done to take advantage of the strong skills of some Bangladeshi organisations in Disaster Risk Reduction and preparedness and other areas, as well as investing more in capacity building of national organisations.

Coordinating on messaging
Attempts are being made to coordinate between agencies on messaging, and materials are being developed in line with government messages, which is proving successful (SSHAP KIIs). The CwC Working Group provides partners with guidelines on dissemination mechanisms, and the partners develop the materials. The CwC Working Group aims to bring all partners into line with their agreed strategy. However, with over 150 organisations working in the response, in some cases agencies that are not partners of the CwC Working Group produce materials that are not in line with the strategy, and they have been assisted to adjust the messages (SSHAP KIIs). It has been noted that COVID-19 specific modules have been helpful in ensuring cross partner consistency in message dissemination and suggested that producing additional modules for various topics could be helpful (SSHAP KIIs). As mentioned above, the CwC and RCCE WG mechanism requires approval from the Civil Surgeon for health messages, so it is important to foster a good working relationship with them to ensure a fast turnaround time (SSHAP KIIs). Coordination with partners is key when testing messages with communities, finalising with local experts representing appropriate sector and agency and securing final approval from the Civil Surgeon before its dissemination from all available platforms (SSHAP KIIs). In some cases, key informants have felt that messages are hastily signed off by the Civil Surgeon Office, without being reviewed by the whole RCCE Working Group particularly during public health emergencies where RCCE interventions need to be rapid (SSHAP KIIs).

Coordinating on volunteer networks
Many humanitarian agencies in Cox's Bazar engage Rohingya volunteers for different purposes, including RCCE. There is no coordination body for these different volunteer networks, and there is a call for systematic mapping & rostering for the development of a coordination mechanism for these networks . Better coordination of their activities would reduce duplication of efforts and allow for standardisation of training, standards and recruitment terms. A suggestion has been made to establish a sub-group for volunteers, but this has not yet occurred (SSHAP KIIs).
Coordinating on data collection, analysis and sharing Relevant, quality data is only useful if it is used appropriately. To make the best use of the data collected and to avoid duplication of efforts and research fatigue, partners need robust mechanisms for sharing data and aligning their efforts to collect feedback. One positive development from the COVID-19 pandemic has been a necessity for partners to improve coordination in terms of data collection and sharing, because access to the target population was restricted. As such, organisations have begun to work together to share priority questions, to share data, and to coordinate data collection exercises. Since 2018, BBC Media Action and Translators Without Borders have been working together to collect, collate and analysis feedback from communities from a range of sources and organisations. They also conduct interviews with the Rohingya to gain a better understanding of the issues raised from the data. The "What Matters?" bulletin is published on a regular basis using this information and aims to provide a snapshot of feedback received from Rohingya to assist sectors to better plan and implement relief activities with communities' needs and preferences in mind. The publications are available in both English and Bangla. In addition, organisations such as IOM, UNHCR and their partners are beginning to use a common system to collect feedback, rather than each using a different system and data list. This means data can be analysed across, rather than only within agencies. However, many agencies and organisations continue to use different feedback collection mechanisms, with methodologies that are not comparable, making the different data sets incompatible (Sida & Schenkenberg, 2019).
There is an Information Management Working Group in Cox's Bazar, with a technical working group that has been looking at trying to increase the coordination levels of data assessment and analysis and to develop standards. The idea is to map who is doing what, where and when in terms of data collection to coordinate assessments and analysis and to avoid duplication of data collection efforts. This is working relatively well; however, there continue to be limitations in standard setting, with much of the data collected being of uncertain quality, as discussed above. More coordination and agreement are needed to introduce minimum standards for data collection and analysis.
Mechanisms for referring feedback to other sectors are also ad hoc, and need to be systematised to make the best use of the vast amount of data being collected (Internews, 2017;SSHAP KIIs). Some organisations such as BBC Media Action and ACAPS communicate regularly with partners to ensure data needs, the data itself and communication strategies are aligned and responsive. BDRCS share information via the RCCE structures and follow up to promote action based on the feedback (SSHAP KIIs). Although a massive amount of feedback is being collected, there is a need for a formalised system to deal with multi-sectoral feedback, with clarity on who takes responsibility for following up on which feedback (Aelbers et al., 2018). Further, there is a need for organisations across the response to share lessons learnt for data collection and community engagement (SSHAP KIIs).

4: Persistent bottlenecks Structural bottlenecks
While improvements have been made to communication, community engagement and accountability in recent years, some structural bottlenecks remain. These include: a complex coordination structure and lack of coordination among the many response partners; a lack of adequate investment in community engagement; a lack of learning and responsiveness with regard to community engagement; low levels of accountability; high staff turnover; low quality of care and poor treatment of Rohingya at health facilities; policy constraints hindering the inclusion of Rohingya in the response; policy constraints hindering refugees' access to telecommunications and Internet connectivity; and camp structures that prohibit the improvement of community engagement and accountability.

Coordination
Although there have been moves toward better coordination, partners continue to point to a complex and confusing coordination mechanism, as well as a lack of effective coordination about communication and community engagement, data collection and analysis, data sharing, and referral of feedback to other agencies or sectors. Interviewees note that at times the different roles of the CwC coordination mechanism are not clear, and that the RCCE Working Group is not able to effectively feed its inputs into the CwC Working Group without slowing down the process.(SSHAP KIIs) It has been highlighted that there are numerous different types of information hubs or centres run by different agencies, and poor coordination between them. There is also a lack of coordination of the many different volunteer networks and a need for standardised training of staff and volunteers (Aelbers et al., 2018). There are differences in policies, processes and standards between agencies (UNHCR, 2018b). It has also been noted that the response does not work closely enough with the governments, and that government is not present in CwC meetings; however, they are well briefed on the context for the development of the public health messages and the expected outcome when an approval is sought. There is a high turnover rate of staff within some organisations, which makes continuity, coordination and collaboration a challenge (SSHAP KIIs).

Investment in and support for community engagement
There is a sense that money continues to be spent on healthcare equipment and personnel, but that not enough is invested in community engagement (SSHAP KIIs). Additional targeted investment is required from organisations and donors for dedicated and specialised communications, community engagement and accountability staff and specialists (Internews, 2017), as well as on capacity building for local volunteers, translators, interpreters, and media. More investment needs to be made in understanding how meaningful and effective community engagement can be fostered in the context of Cox's Bazar. In most cases, the same donors remain active and synergies between agencies for coherent RCCE interventions could be explored to deliver better engagement and sustainability.

Learning and responsiveness
While there is a recognition that it takes time to build trust and create an environment in which people feel comfortable speaking honestly, there is frustration that the response in Cox's Bazar has been slow to learn these lessons and to make changes accordingly. It is considered that lessons learnt and ways forward are not always incorporated into operational changes, which maintains the status quo (SSHAP KIIs). The CwC Working Group has made important inroads, but since it does not have a large operational presence on the ground, has not yet managed to place the community engagement agenda at the very heart of the response (SSHAP KIIs). The response has needed to evolve from an emergency one focused on supply and logistics, to a long-term, people-centred and participatory one, and this is where it has stagnated (Sida & Schenkenberg, 2019).

Accountability
There is still a weak feedback loop, with few organisations sharing the data they collect with the Rohingya community, or informing people what they are doing with the information, how it is influencing service delivery and how their questions can be answered Internews, 2017;SSHAP KIIs). In addition, many organisations do not share feedback with their headquarters or coordination chairs, meaning that this data is not being used to inform broader programming decisions at partners end except immediate actions for containment of the public health emergency (Aelbers et al., 2018;SSHAP KIIs).

Quality of care and treatment at health facilities
A structural bottleneck that it may be beyond the scope of CwC partners to tackle is the state of health service provision in the camps and accessible to Rohingya beyond the camps. Local Bangladeshi government structures have limited resources for health, and reports indicate unclear mandates with regard to health service provision for Rohingya refugees (Guglielmi et al., 2020;Sida & Schenkenberg, 2019). However, the government and the WHO work to ensure that minimum health packages and disease specific guidelines are met by each health facility; periodic monitoring and evaluation visits and data collection are done by the health sector to maintain the compliance (SSHAP KIIs). There are complaints of lack of available services, with inadequate services in the way of RSH and emergency obstetric care, NCDs, eye and dental care, and age-, gender-and disability-responsive care (Guglielmi et al., 2020).
These shortcoming are compounded by the previously discussed perceptions of poor treatment some refugees have described receiving at the hands of health workers at facilities (ACAPS & IOM, 2020b; SSHAP KIIs). While not all experiences are negative, the greater impact on healthcare seeking, adherence to treatment procedures and trust necessitates a systemic approach to addressing the issue. An Accountability to Affected Populations (AAP) framework for health actors is currently in development with an anticipated rollout in 2021 (SSHAP KIIs). In addition, some organisations are taking active steps to address the issue and some sector wide trainings have been provided to healthcare workers on respectful communications (SSHAP KIIs). However, these efforts should be scaled up and expanded, especially in light of the high turnover of healthcare staff (SSHAP KIIs). Recommended efforts should involve strict codes of conduct about treatment of and communication with Rohingya at health facilities (ACAPS & IOM, 2020b). This would include training of healthcare workers on patient/doctor relations, culturally appropriate communication, respectful use of language and accountability mechanisms. In parallel raising awareness amongst Rohingya of usual consultation protocols would reduce misunderstandings about the intentions of healthcare workers. As in all aspects of the response, attention also needs to be paid to language in health facilities, as good healthcare provision and trustbuilding are impossible if patient and doctor cannot understand one another (ACAPS & IOM, 2020b). A health specific glossary of key terms has been suggested as a helpful tool (SSHAP KIIs). Gendersensitive attention should also be available, meaning female staff and adequate privacy for women during consultations (Ripoll, 2017).

Policy constraints hindering inclusion of Rohingya
The Bangladeshi government currently prohibits the employment of Rohingya refugees. This means that although they are allowed to be engaged as volunteers, they do not have contracts or benefits and cannot be promoted to positions of management and decision-making (SSHAP KIIs). Even the space for engaging Rohingya volunteers is constrained as new regulations that prioritise engaging Bangladeshi staff and volunteers are affecting the number of Rohingya versus Bangladeshi an organisation can recruit (SSHAP KIIs).

Policy constraints hindering access to communication channels
It is not legal for Rohingya refugees to possess SIM cards for Bangladeshi mobile phone networks, although most of them do in practice Internews, 2019). Currently, the Bangladeshi government agency, RRRC, is allowing response partners and refugees to use SIM cards only for the purpose of implementing the IVR system, but for no other purpose (SSHAP KIIs). Among those refugees who do have phones, most are male, and usage of phones is usually controlled by male family members which further restricts available strategies for engaging vulnerable and hard to reach populations (Holt et al., 2020;SSHAP KIIs).
Lack of Internet access in the camps, as restricted by the government, is also a barrier identified by response partners (BBC Media Action & Translators without Borders, 2020a). This leads to communication delays between Cox's Bazar, where coordination and message production takes place, and the camps. During the recent fire incident, for example, partners had to return to their offices outside the camp to receive messages about support available to the affected population, before returning to the camps to disseminate these (SSHAP KIIs). For some disasters, such as cyclones, content is prepositioned with relevant people within the camps so that information can be made available, as needed, in an emergency. In addition, few Rohingya have access to radio due to poor broadcast coverage in the camps, as well as lack of radio sets. This is a clear missed opportunity, due to the appropriateness of local radio as a channel for communicating with communities with low levels of literacy .
Camp structures prohibiting improved community engagement and accountability The current system of army appointed Mahjis acting as representatives of Rohingya is perceived as a challenge, since it requires people to depend on these actors who are often not considered representative or trusted community members and have limited knowledge of the humanitarian architecture and mechanisms in place. Efforts are ongoing to replace this system with one of elected camp committee members (Sida & Schenkenberg, 2019); however, while anecdotally successful, this process is slow, and hindered by the need for political coordination with the government and site management (SSHAP KIIs).

Sociocultural bottlenecks
Certain social norms on the part of both host and guest communities can translate into barriers to healthcare seeking. These are sensitive to address, and difficult to challenge. Notable challenges raised include experiences of widespread discrimination, gender norms and language.

Social tensions and negative perceptions of Rohingya
While many have noted the accommodations and efforts of the Bangladeshi to aid the Rohingya refugees, it has also been frequently reported that the influx of Rohingya into Bangladesh has contributed to tensions and resentments with the host community as a result of the impact and competition for available resources (UNHCR, 2018a; SSHAP KIIs). The government seeks to address some of these tensions and the strain on the host community by ensuring that resources made available to the refugees are also available to the host communities to some extent. For example, health messaging efforts are also required to be provided in Bangla so that host communities can also benefit from awareness raising efforts (SSHAP KIIs).
Fortunately, findings on social cohesion between the Rohingya and host communities has shown improvement over the years. However, less than half (43%) of Bangladeshi respondents feel there is harmony between themselves and the Rohingya and only 41% have indicated willingness to socialise with Rohingya (Ground Truth Solutions, 2020). Multiple stakeholders further indicate that negative perceptions of the Rohingya persist and many Rohingya report disrespectful treatment (SSHAP KIIs). In one study, Bangladeshi community members have attributed the tension to Rohingya employment competition as Rohingya unofficially work in the local economy (Ground Truth Solutions, 2020). However, policies that exclude Rohingya from education and legitimate forms of employment in the camps also contribute to criminality and reinforce discriminatory views of the Rohingya as 'uneducated criminals' (SSHAP KIIs). From the Rohingya side, there have been reports of higher tensions with the host communities in areas the populations are more mixed and reports of discriminatory treatment by health workers and responders (Ground Truth Solutions, 2020; Guglielmi et al., 2020; SSHAP KIIs). Guglielmi et al (2020) provide examples where Rohingya described that they have been treated like dogs, teased and mistreated at health facilities. As noted in section two, these experiences, compounded by a history of mistreatment and discrimination in Myanmar, leads to a cycle of mistrust and misunderstanding (SSHAP KIIs). Tackling entrenched discrimination and building trust between communities will be a protracted process. Within humanitarian organisations, policies and accountability mechanisms to address this issue need to be implemented, and employees and volunteers trained on respectful intercultural communication.

Gender norms
Women and girls practice purdah which restricts mobility outside the home, and require permission from their husband or mother-in-law to seek healthcare, including maternal and reproductive care (Guglielmi et al., 2020;SSHAP KIIs). In Myanmar, being able to maintain purdah was associated with a higher socioeconomic status. In this context, it may be additionally confronting and undesirable for women to leave to home or to engage in work or volunteer work, creating a barrier to the engagement of female community leaders and volunteers (Ripoll, 2017). While attempts are being made to engage female religious teachers and other volunteers to communicate about health, these teachers are also restricted in their mobility, rarely leaving their shelters. Visiting and engaging with the religious leaders themselves is a challenge, particularly given the difficult landscape in the camps. Meanwhile, female volunteers have faced difficulties supporting humanitarian work due to the entrenched view that women should not work and should not receive incentives. In some cases, there have been security and privacy threats against female Rohingya volunteers (SSHAP KIIs). Trusted leaders can play a critical role in shifting social norms around women's work. In addition, women have lower literacy rates and less access to mobile phones than men (UNHCR, 2018a). Research continues to show that female-headed households have less awareness about available aid and services than male-headed households (Ground Truth Solutions, 2020). On the Bangladeshi side, family planning, pregnancy-related services, post-abortion care (mensuration regulating services) are not approved for provision to unmarried women (Aelbers et al., 2018).

Language
The difference in language between host and refugee community remains a persistent challenge, and even where attempts are made to deliver materials in the appropriate language, quality assurance is a challenge. Translations are often done by Bangladeshis who may not fully understand the linguistic nuances of the Rohingya language all though they are generally well-versed with the Chittagonian language (SSHAP KIIs). In most cases, knowing Chittagonian language provides misguided confidence about communicating well in Rohingya language.

Logistical bottlenecks
There is some evidence that risk communication and messaging is working, and people are understanding the content of the messages. However, they are not always able to change their behaviour in line with these messages due to logistical bottlenecks. These include lack of materials and equipment, costs and distance to health centres. In FGDs, when told to use masks or hand sanitiser, they ask to be provided with these materials (SSHAP KIIs). Cost of care or medicines have also been cited as barriers to seeking care. As illustrated by an adolescent refugee, 'the hospital doctors tell [us] to buy [medicine] from outside [of the medical facility]. We go there because of our financial problemsif we have to buy medicines from elsewhere, what is the need of this hospital?' (Guglielmi et al., 2020). Transportation costs for travel to hospital in cases of more serious illness, including to Cox's Bazar, Chittagong or even Dhaka are prohibitive (Guglielmi et al., 2020). A study on adolescents with disabilities in one of the camps founds that 100% of respondents experienced barriers in accessing health services due to lack of accessible health facilities, long distance to health providers, physical barriers due to topography, and lengthy wait times (Guglielmi et al., 2020).

5: What areas require further investigation?
There is no lack of data collection in the Rohingya refugee camps in Cox's Bazar. However, according to the literature and key informants, gaps remain in the type of data collected as well as in specific areas relevant to health outcomes and community engagement efforts.

Collecting the right type of data
Most data collected in Cox's Bazar is quantitative. However, there has recently been more recognition of the importance of qualitative research to fill in the hows and whys driving the patterns being seen in the quantitative data findings (Sida & Schenkenberg, 2019). Surveys with closed-ended questions are limited in terms of the richness of the data that can be collected. On the other hand, research methods that enable participants to freely share their experiences, concerns and queries on topics that are most important to them provide an additional layer of understanding and nuance, which can be invaluable for designing appropriate and relevant RCCE strategies and interventions.

Identified knowledge gaps
There are still knowledge gaps in critical areas including the health status of the Rohingya communities, the underlying drivers of healthcare seeking behaviours, Rohingya experiences of healthcare facilities, how the Rohingya perceive the RCCE interventions themselves, and the extent to which these translate into improved health outcomes.
Health status of the Rohingya Partners and literature have stated that more needs to be known about the actual health status of the refugees. Existing literature has indicated that certain non-communicable diseases like hypertension and diabetes are prevalent in Rohingya communities in Cox's Bazar, however, the evidence is limited (Joarder et al., 2020). While the WHO produces weekly epidemiological summaries (WHO Cox's Bazar Data Hub https://cxb-epi.netlify.app/#dashboard), it was indicated by key stakeholders that further research on the prevalence and distribution of disease and illness and their determinants is necessary to establish good baseline data for understanding needs and designing effective responses.

Healthcare seeking behaviours
Gaps in knowledge about Rohingya healthcare seeking practices need to be identified and researched, rather than continuing to gather data on areas that are already well understood, or that may not be relevant (ACAPS, 2020). For example, partners have stated that more needs to be known about Rohingya's actual and perceived needs, priorities and challenges. A more nuanced understanding is also needed of the Rohingya community's processes and rationale for decision-making around healthcare seeking, and the ways these link to culture and context, with particular attention to vulnerable groups within the population. For example, the ways in which gender norms and historic experiences with repressive reproductive health policies and sexual violence shape sexual and reproductive health service usage (Parmar et al., 2019). Social science data and community perceptions need to be contextualised within the data from the health sector, to highlight which perceptions are "rumours" and which perceptions reflect deficits in services to address all health needs.

Experiences of health services
Perceptions on safety, availability and credibility of health services is showing some improvement. However, more research and standardised indicators so that patterns across camps and facilities can be compared and assessed is needed. Further research should examine perceptions of recent health care experiences and explore how factors such as gender, language, education, etc. shape these experiences (ACAPS & IOM, 2020d;Parmar et al., 2019).

Impact of RCCE strategies
Ongoing impact evaluations of RCCE strategies are necessary to validate strategies and the effectiveness of messages and identify where they are working and where they are not. It is essential to understand where, with whom and why specific strategies and messages work and how behaviour change is adopted to ensure that appropriate strategies are used in each context. Findings from the COVID-19 response have also highlighted the need to further explore how health promotion messages are understood by various community members, including women and girls, the elderly and people with disabilities (ACAPS & IOM, 2020c; BRAC University, 2021). Agencies are usually funded by the same donors; hence, there is a need to explore and identify the synergies between health and non-health RCCE interventions, share best practices and showcase where there are cost effective, sustainable and pragmatic interventions which can be continued with fairly minimal supervision and funding (SSHAP KIIs).

Effective methods of engagement
Additional research is needed to understand effective methods of engagement for various groups within the Rohingya population in Cox's Bazar. These include religious leaders, Imams, women religious teachers, women's groups, youth, older people, people with disabilities, transgender people, etc. Understand appropriate and respectful ways to engage with members of these groups, how, when and where are best for engaging with them, and how to appropriately include their voices in decision-making is critical for advancing community engagement strategies, improving accuracy of data, improving trust, and creating programs that are appropriate to the communities they seek to serve. This includes the validations of messages being disseminated, identification of appropriate messaging platforms for various groups and subgroups within the communities, and appropriate and effective feedback mechanisms (SSHAP KIIs).

Community solutions
Community solutions to the issues they face in the camps and how they are and would like to manage health threats like COVID-19 have yet to be fully explored. Community members are the experts on what is appropriate and acceptable for them and what resources and support structures they have ready access to. As such, explorations of how communities work to solve their own problems can provide valuable insights into effective strategies for the response.

Annex 2: Quality appraisal strategy and findings
The documents were appraised for quality using customised appraisal forms developed by drawing upon several previously published tools and checklists. Different appraisal forms will be used for qualitative and quantitative studies and grey literature.

Qualitative studies
For qualitative and mixed methods studies, the rigour of studies will be assessed using an approach adapted from Bangpan et al. (2019) according to criteria of reliability and usefulness, as follows: • Reliability criteria: 1) sampling; 2) data collection; 3) data analysis; 4) the extent to which the study findings are grounded in the data.
• Usefulness criteria: 5) the extent to which the findings contribute substantially to answering the research questions.
Reviewers will answer yes or no to a question relating to each of the 4 reliability criteria. The study will then be given an overall reliability score of low, medium or high. Next, a score of low, medium or high will be given against the 5th criteria, which will serve as the study's usefulness score.

Quantitative studies
Quantitative studies will be assessed according to criteria of selection bias, detection and performance bias, attrition bias and reporting bias.
We have chosen to follow an appraisal method that assesses the credibility and robustness of research, while not being arduous in gathering unnecessary data about the study. The method also places importance on the relevance of the study and the contribution it can make to the review. Since this review aims to capture a comprehensive understanding of the existing evidence, a judgement may be made to include studies that would ordinarily be excluded from a formal systematic literature review.

Grey literature
Journalistic reports and other formats such as programmatic reports and synthesis will be appraised using a more limited checklist, including apparent partiality or impartiality of the author, the extent to which findings are supported by evidence, and the extent to which the findings contribute to the research question.