Engaging Arab/MENA Communities in Learning Health Systems: Insights and Guidance for Future Research on AI and Health Equity

ABSTRACT

Introduction

The integration of artificial intelligence (AI) tools in healthcare offers significant opportunities to improve patient care. However, underrepresented groups such as the Arab/Middle Eastern North African (MENA) community in the United States have historically been excluded in health data and conversations regarding AI tool implementation and development. We present our experience with the Arab/MENA community exploring attitudes about the use of AI in healthcare, focusing on our engagement and recruitment efforts as well as relevance for learning health systems science.

Methods

We conducted a virtual democratic deliberation session (n = 33) in partnership with the Arab Community Center for Economic and Social Services (ACCESS) in Michigan, as part of a larger study involving five sessions (n = 159). In partnership with ACCESS staff, we collaboratively developed study materials and recruited Arab/MENA community members to share their perspectives on AI in healthcare. Qualitative thematic analysis was used to identify the community's perspectives, priorities, and barriers to the use of AI in healthcare.

Results

The deliberation session highlighted four key themes related to the use of AI tools: (1) transparency in AI development was viewed as essential to building community trust, (2) human connection, with concerns that increased reliance on AI could replace empathy and weaken patient‐provider interactions, (3) the role of healthcare providers, with preference on providers using AI as a supportive tool rather than replacing direct care, and (4) representation due to concerns over whether AI systems would reflect the experiences and needs of the Arab/MENA community in healthcare.

Conclusions

Community‐based partnerships are essential for advancing responsible AI implementation in healthcare and building a learning health system. Our experiences highlight the importance of transparency, cultural sensitivity, and meaningful community involvement to build trust and address the needs of underrepresented groups as AI evolves.

1. Introduction

Artificial Intelligence (AI) uses computational methods to simulate human thinking such as learning from data, recognizing patterns, and supporting decision‐making for recommendations [1]. In healthcare, AI promises to transform patient care and healthcare systems and has demonstrated potential for improvements in diagnostic accuracy, treatment personalization, and administrative efficiency in patient care [2, 3, 4]. AI also aims to improve the care experience for providers, caregivers, and patients by optimizing communication [5]. AI tools are often trained on large datasets to generate recommendations, predictions, and improve patient care [6]. However, in communities where trust in health systems is low, AI can raise additional concerns about fairness, transparency, and how patient data is used, which can shape whether or how tools are accepted or used in care [7]. Without intentional design and meaningful community engagement, AI tools may fail to earn public trust and may not be applied in ways that reflect the values and perspectives of diverse patient communities [7, 8].

Community engaged research involves inclusive participation and mutual respect that brings together people based on self‐identified connections to place, shared interests, culture or similar circumstances [9]. Community engagement ensures that community values and needs are at the forefront of AI tool and policy development. Understanding how minoritized communities perceive AI in healthcare is especially vital, given the longstanding mistrust of medical providers and health systems rooted in historical experiences tied to these identities [10, 11, 12, 13, 14, 15, 16]. As AI transforms healthcare, health systems should develop innovative approaches to health research and healthcare delivery that value and prioritize the needs of marginalized or excluded communities.

Learning Health Systems (LHS) provide frameworks for ensuring AI is ethically integrated into health systems. At its core, LHS represents a continuous improvement cycle of collecting data to generate new knowledge that is put into practice and evaluated [17, 18]. Learning cycles are predicated on a community of interest coming together to provide a focal point for improvement [17]. Community engaged methods, such as deliberation, provide a process for including communities as stakeholders. Partnering with communities, such as the Arab/Middle Eastern North African (MENA) community, is central to building AI‐enabled learning health systems that are representative, trustworthy, and aligned with the needs of the community.

Arab/MENA communities in the United States are an underrepresented group in research and lacked distinct representation in the US Census until 2024, despite their population size. Although some research has addressed the challenges and provided guidance on best practices for engaging Arab/MENA communities, the literature on Arab/MENA health outcomes remains limited [19, 20]. Available studies suggest Arab/MENA communities experience health disparities, including a high prevalence of cardiovascular disease, low birth weight and depression [21, 22, 23]. Addressing this gap will require distinct Arab/MENA population health research and improved documentation of community health trends through the systematic collection of race and ethnicity data [21]. Historical practices in data collection have led to the absence of a distinct MENA racial and ethnic category. It was not until March 28th, 2024, that the Office of Management and Budget (OMB) published revisions to Statistical Policy Directive No. 15, introducing significant changes to improve race and ethnicity data collection and reporting [24]. These revisions included adding a “Middle Eastern or North African (MENA)” category to the US Census distinct and separate from the White category, where MENA individuals were previously classified.

In this experience report, we describe a research project that aimed to identify attitudes about the use of AI in healthcare and to solicit recommendations about transparency practices around the use and development of this emerging technology. This study included outreach to the Arab/MENA community in Dearborn, Michigan, one of the largest in the United States. We provide a detailed account of the methods we used for engagement and recruitment as well as an overview of what we heard from MENA participants. Finally, we consider the implications of this experience for research and for learning health systems science.

1.1. Background and Experiences of Arab/MENA Communities in the US

The MENA community in the United States represents significant diversity in national origins and religious backgrounds. The MENA racial and ethnic category encompasses individuals from various countries, including countries from the Arab world (spanning 22 nations in Asia and North Africa) as well as non‐Arab countries (i.e., Iran, Turkey) [25, 26]. The religious landscape is also diverse, with Muslims (including Sunni and Shia sects) comprising the majority, and Christians (including Maronite, Chaldean, and Coptic) representing the largest minority [27]. In 2020, 3.5 million individuals identified as being of MENA descent in the United States [26]. Michigan has the second‐largest MENA population in the country, with approximately 400 000 residents [28], who are predominantly of Lebanese, Iraqi, Yemeni, and Syrian ancestry [25]. Within Michigan, Dearborn has the highest percentage of Arab/MENA residents of any US city [29].

The growth of this previously overlooked minority, coupled with a limited understanding of their perspectives and experiences in healthcare and exclusion of data about MENA communities that has contributed to underrepresentation in health research, underscores the importance of assessing their views on emerging digital health technologies. Research on MENA communities' perspectives on healthcare technology remains underdeveloped, with only one study, to our knowledge, examining the use of patient portals for communication and outreach between MENA patients and providers [30]. The recent inclusion of the MENA racial and ethnic category in the US Census provides an opportunity to more accurately represent the needs of Arab/MENA populations and better inform public policy, health research, and healthcare system interventions [31].

1.2. Barriers to Engaging With the Arab/MENA Communities

Despite the growth of MENA communities, significant barriers persist in addressing their health needs and ensuring equitable access to healthcare services. A lack of culturally sensitive practices, difficulties in accessing and utilizing health insurance, and financial struggles contribute to these challenges [32]. Practices such as ensuring cross‐gender medical interactions are limited (e.g., between patients and providers) and stigma related to sensitive topics such as mental health, sexual health or substance abuse add additional barriers to accessing medical care or participating in research. Distrust of providers and researchers is often rooted in experiences of limited empathy or understanding cultural factors [33] and may contribute to a reluctance to share information due to concerns about privacy and confidentiality [33, 34].

Language barriers represent another significant obstacle to community engagement in research. In a study evaluating the facilitators and barriers to recruiting Arab/MENA community members, limited English proficiency was identified as a key barrier to participation in healthcare research [19]. Limited Arabic proficiency within the research team also serves as a barrier, as limited language capacity can negatively affect engagement. While this challenge is often addressed by either including a team member who speaks Arabic, or utilizing translation services [35], these are difficult to implement due to limited resources. Study materials, including survey instruments, interviews, and consent forms, are typically available only in English, which limits the accessibility and comprehension. In addition, recruitment within the Arab/MENA community often requires extended timeframes to build community relationships and trust [34, 36]. Finally, historical and ongoing experiences of systemic discrimination, surveillance, and xenophobia further contribute to healthcare access barriers among Arab/MENA communities in America [20, 28, 37]. These challenges have been exacerbated by the changing socio‐political landscape in the United States and the Middle East, fostering deeper mistrust in institutions, including the healthcare system [10, 38].

1.3. Report Objectives

This report details and reflects on our community engagement experiences with the Arab/MENA community, conducted in partnership with the Arab Community Center for Economic and Social Services (ACCESS), a community‐based organization serving the Metro Detroit community since 1971. We present our experience with the Arab/MENA community in the context of recruiting, conducting, and reporting a research study that aimed to identify attitudes about the use of AI in healthcare and to solicit recommendations about transparency practices. We specifically focus on the details of our engagement and recruitment efforts as well as findings relevant for research and learning health systems science.

2. Engaging With Arab/MENA Community

2.1. Partnering With Community Organizations

We partnered with the Arab Community Center for Economic and Social Services (ACCESS) to engage community members and understand their perspectives about AI in healthcare. ACCESS is a nonprofit organization serving diverse communities and advancing their health, social, and economic well‐being through medical, human, and employment services and educational programs. Its vision is to foster a just and equitable society with the full participation of Arab Americans [39]. The organization currently holds 11 locations and provides over 120 programs throughout Metro Detroit, ranging from social, economic, health, and education services.

To facilitate the partnership in our study, we collaborated with a community advisory board, Deliberative Engagement of Communities in Decisions About Resource Spending (DECIDERS) [40]. DECIDERS is a community engagement initiative that uses deliberative democracy (DD) methods [41] to engage diverse communities and incorporate their voices in health research and policy discussions. The group is composed of community leaders, researchers, and stakeholders from diverse backgrounds, particularly minority and underserved communities. Members of this committee were a part of the research team for several years prior to working directly with ACCESS. Specifically, a DECIDERS steering committee member and former Senior Director of ACCESS was a part of our research team. He was embedded from the outset, participating in weekly meetings and contributing to the design and development of study materials, deliberation structure, and recruitment strategies. He also facilitated our connection with ACCESS organizational leadership to begin outreach to the Arab/MENA populations in Michigan.

2.2. Our Recruitment Methodology

To explore the perspectives of patients and the public regarding AI in healthcare, we held five deliberative sessions with community members across Michigan. Deliberation methods, used in previous studies [42, 43], bring together ~30 people at a time for in‐depth learning about new material and small group discussions about topics that are complex and relevant to policy. Additional details on the deliberation framework, protocols, and findings across all five Michigan community sessions are described in other literature [44, 45].

This experience report focuses on the session conducted with people who identify as part of the Arab/MENA community, given the demographic composition of the state of Michigan and the goal to include people underrepresented in research. To begin our recruitment efforts, we met with ACCESS leadership to explain the project and the impact this research can make on the community by engaging in this type of dialogue. To begin the collaboration, ACCESS required us to complete a Research Request Form to describe the study's elements.

We had several meetings and conversations to ensure appropriate processes were in place for the community, including cultural accommodations such as avoiding any major religious holidays like Ramadan and Eid. Based on recommendations from ACCESS staff and leadership, participants were recruited via flyers, emails, phone calls, text messages, and social media posts that included a link to an online interest form. The online interest form included a copy of the consent form and questions about eligibility, contact information, and a copy of frequently asked questions to describe the study elements.

Using information from the interest form, a research team member, who was also a bilingual native Arabic speaker (RH), called interested individuals to obtain verbal consent and enroll them in the study. To be eligible for the study among all five deliberations, participants needed to have access to an electronic device (in order to participate via Zoom), be at least 18 years of age, live in Michigan, and be able to read and speak English, due to budgetary limitations for translation services. Having a bilingual recruiter allowed us to engage with participants in their preferred language or to clarify details of the study. The recruitment period lasted approximately 6 weeks.

Enrolled participants were sent a packet in the mail 2 weeks prior to the session. The packet included a printed copy of the consent form, a booklet with an introduction to AI in healthcare, instructions for Zoom setup, and a participant discussion packet to be used during the session. The packet also included copies of the educational presentations and materials for the small group discussions. Participants completed predeliberation surveys 2 days before the event and a postdeliberation survey to evaluate their understanding and perspectives on the topics covered. Participants received a $200 incentive for completing all study components.

A total of 38 participants were enrolled in the study, with 33 participants attending the virtual deliberation. A detailed description of the participants is provided in Table 1. Most of the participants had a direct connection to ACCESS and many had a healthcare background. The majority of the participants were from Dearborn or southeast Michigan and represented a younger generation (ages 20–39) within the Arab/MENA community.

TABLE 1.

Sociodemographic characteristics of Arab/MENA session participants.

Demographic characteristics total (n = 33)
Gender
Female	23 (70%)
Male	10 (30%)
Neither of these describes me	0 (0%)
Age group
20–39 years	18 (54.5%)
40–59 years	12 (36%)
60+ years	3 (9%)
Education
Less than Bachelor's degree	10 (30%)
Bachelor's degree	16 (48%)
More than a Bachelor's degree	7 (21%)

The small group Zoom sessions were audio recorded, transcribed verbatim, and de‐identified for analysis. Two study team members (K.A.R., M.S.) coded the transcripts thematically, with another two members (R.H., D.S.) later focusing on coding the third session. Following each deliberation, participants were mailed a Certificate of Completion as well as a summary report of the results from the session.

After completing the five deliberation sessions, participants across Michigan were invited to a virtual town hall to review the overall findings and engage in a collective discussion. Eleven participants, or approximately one‐third of the initial deliberation group, from the Arab/MENA deliberation session attended the Town Hall. The session provided participants with a summary of findings, including quotes from the deliberation sessions. We also brought in three leaders in AI and health on the topics to share their thoughts on the participant discussions and how they might inform policy and practice.

3. Descriptive Findings on Health Data Equity: Transparency, Human Connection, Role of Providers, and Representation

The deliberative session with the Arab/MENA community highlighted four key themes related to the use of AI tools in their healthcare: (1) transparency, (2) human connection, (3) the role of healthcare providers, and (4) representation.

Transparency was one of the key themes that was seen as foundational to building trust in AI tools in healthcare. Participants highlighted the need for open and honest communication to foster trust within the community, emphasizing transparency about how these tools were created and by whom. One participant said, “I think as patients, if an AI tool is being used in our healthcare services, I feel like we should know how that AI tool was developed, and by who.” Many highlighted that AI use‐related disclosures should reflect cultural sensitivity, be available in the language of the community being served and be explained in lay terms to ensure accessibility and comprehension.

Concerns about maintaining human connection were also raised. While some recognized the potential of AI to assist with administrative tasks, data analysis, and early diagnosis, they emphasized that AI can never replace empathy, creativity, and the personal touch that healthcare providers offer, especially to those with mental health needs and the elderly. Participants from this community highlighted that their trust in the healthcare system is built around human relationships, something they feared AI would never be able to offer. “The human factor should always be there,” a participant stated, especially for emotionally sensitive care.

This is closely connected to how participants viewed the role of providers in an AI‐integrated system. Most felt AI should support and not substitute the provider's role, with one participant sharing, “I'm not against AI if it's an assistive system. I'm against AI taking over.” Other related concerns were raised about accountability, training, and the industry's influence on AI‐generated recommendations. One participant asked, “What becomes the role of a healthcare professional in AI‐driven care?”

During the session, representation as an issue of health data equity was a recurring concern about AI tools in healthcare. Participants highlighted how Arab/MENA populations have long been misclassified as White in US Census data, despite having their own needs and experiences, leading to underrepresentation in healthcare and research. This lack of inclusion raised fears that AI tools trained on biased datasets would fail to serve their community. As one participant shared, “If the data is collected from primarily Caucasian people… it might perpetuate existing biases in the healthcare system.” They felt that while the introduction of the MENA category in the Census is helpful, potential AI bias and lack of cultural sensitivity may still negatively affect Arab/MENA patients in the healthcare system. They expressed a strong desire for diverse stakeholders, especially those representing their community, to play a role in shaping the regulatory framework and oversight of AI tools and to ensure cultural sensitivity and equity concerns remained as top priorities when they sought care.

Participants emphasized that taking such an approach would build trust in new initiatives within the healthcare system. Following the deliberation session, several key reflections and implications emerged regarding the Arab/MENA community's perspectives on AI tools in healthcare. The subsequent section discusses these thoroughly.

4. Discussion

Our work with ACCESS and the Arab/MENA community offers several insights for engaging communities of interest to investigate a specific problem or initiating the first step in a learning cycle [46]. In this section, we review our key processes and consider their implications for research and learning health systems.

4.1. Partnership and Recruitment

Community‐based methods are often described along a continuum ranging from studies that do not engage the community in research, conducting research “on” community, to studies in which community members are partners and co‐leads in research, or research “with” community [47]. Our work with ACCESS and the Arab/MENA community was intentionally designed to engage community partners at multiple stages of the process including the recruitment and ensuring that the materials and methods were culturally appropriate. Community partners were an integral part of the larger research team, and the ongoing involvement of our community research partnership with ACCESS facilitated a key connection with the Arab/MENA community that made our research possible.

As our work continues, we expect the collaboration to advance along the continuum of community engagement as we further develop a relationship that offers opportunities to identify questions that research can help answer. This is likely to also support participation from Arab/MENA community members given previous research that suggests Arab Americans are more likely to participate in healthcare research studies when the studies work with community‐based research organizations, provide compensation, promote healthcare diversity by meeting the needs of the MENA community, and improve health literacy [20, 48].

4.2. Listening to the Community

In addition to our direct research findings, we heard frustration at the historical and contemporary lack of representation rooted in a long history of categorizing MENA individuals as “White” in national datasets and the US Census, which has led to their invisibility in the healthcare system and research [19, 48, 49, 50]. Participants connected this structural inequity with ongoing barriers to participation in health research. They also noted several recruitment and participatory challenges such as institutional mistrust for disclosing personal information and language barriers, emphasizing the importance of education around AI that is delivered in a manner that is understood to make informed decisions, using translated study materials, and culturally relevant communications.

4.3. Returning to the Community

Sharing the research findings back to the community when a study is completed is a key component to capacity building within a community, as it provides the community members with insights that they can use to advocate for change and support local initiatives [51]. In our research, we accomplished this in the short term by mailing participants summaries of the findings from their session. In the long term, we conducted an online webinar to present the study's results. During that final session, we offered our study reflections and opportunities for the community to provide feedback, while also sharing results with people in leadership positions who could speak to the ways in which both findings and community feedback could be made actionable in health and research institutions. Our team also collaborated with ACCESS in presenting a poster at the 10th Arab Health Summit: Advancing Health Amidst Conflict and Crisis [52], which included their Board of Directors and people from the Dearborn area. By returning the results to community organizations and participants, opportunities for ongoing and future engagement become more likely. This process builds trust and contributes to sustainable, long‐term relationships.

4.4. Implications for Learning Health System Science

Our work offers an example of community members forming a learning community to initiate a learning cycle to understand how AI should be implemented ethically and equitably in healthcare. Notably, the formation of the learning community was a critical first step to exploring the issues surrounding health AI. In so doing, community stakeholders could define the problem in their own terms, develop parameters in which issues are discussed and co‐created potential solutions to ethical AI issues that are culturally appropriate. This engagement approach removed barriers to the community's participation allowing for more equitable attendance and representation combined with the tailored education supports informed decision‐making on the use of AI in healthcare [53]. Most importantly, the focus was on learning with the community members, improving our understanding of health data inequity and underrepresentation in AI and highlighting the needs and expectations for accountability transparency among the Arab/MENA community.

This report demonstrates how LHS core values of inclusiveness and transparency [54] can be operationalized by engaging members of minoritized communities, such as the Arab/MENA community, as co‐experts in developing an ethical framework for the implementation of AI in healthcare. More specifically, to build an LHS, learning should shift more often from relying exclusively on existing health system data to partnering with Arab/MENA communities to set priorities, interpret findings, and define meaningful outcomes [55]. Investments should be made into ensuring that LHS can not only leverage the data that we already have, but also the data we need and should have. Sharing the results back with community members can sustain engagement across iterative learning cycles and ensure health systems are trustworthy partners in AI development efforts. Continued collaboration with the community will be essential to advance health equity and ensure that this work remains responsive and sustainable within community priorities [56]. To support this, LHS must enable communities to conduct routine data collection and feedback processes while ensuring the necessary information and cultural infrastructure are in place [55].

4.5. Acknowledging Limitations

As with any research project or partnership, our collaboration had limitations, which we were able to discuss and acknowledge in the context of our relationship with ACCESS and with our community research team member. Time and budget constraints are persistent. Translated materials, for example, would have facilitated broader participation, particularly for those facing language barriers and others for whom the presence of materials in Arabic would have signaled our commitment to including Arab/MENA communities. In our experience, including members of the research team who have similar backgrounds to the study participants was a feasible way to ameliorate this limitation and has been shown to promote inclusion and willingness to participate in the study, especially when discussing sensitive topics [34]. We also found it valuable to communicate with community partners about potential barriers to recruitment or participation such as current events, religious holidays, and culturally sensitive research topics.

5. Conclusion

Our experience provides a model for community‐based efforts as one approach for learning health system scientists, highlighting both the outcomes and processes of research. These lessons apply not only to engaging with Arab/MENA communities but also other underrepresented communities, each of which has significant and unique health needs. Working locally helps to understand these differences and offer more nuanced and more accurate descriptions of research findings and possible health systems interventions. Disaggregating Arab/MENA data, for example, will enable better identification and methods for addressing needs specific to this population. Community‐led efforts and partnerships can be used to efficiently identify the problems of interest that motivate learning health systems.

Funding

This work was supported by the National Institute of Biomedical Imaging and Bioengineering (5R01EB030492).

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.