Towards a learning healthcare community in the Bronx: evaluating the Bridging Research, Accurate Information and Dialogue (BRAID) model

Abstract

Background

Learning healthcare communities (LHCCs) have been proposed as a next-generation evolution of learning health systems that can advance health equity; however, a practical mechanism for enabling the active and continuous community engagement required for one has not yet been described in the literature. Recognizing that community-based participatory research (CBPR) could potentially meet this need, a team at the Montefiore Medical Center/Albert Einstein College of Medicine designed a novel evidence-based CBPR model – Bridging Research, Accurate Information and Dialogue (BRAID) – that initiates meaningful, longitudinal dialogues to foster bidirectional trust between researchers, clinicians, scientists and communities.

Methods

A mixed-methods cohort study of two BRAID cohorts was conducted between 2022 and 2023. Eligible participants were recruited from the Bronx, New York, United States and convened in a series of conversation circles. Multimodal data was collected from all participants, including quantitative pre- and post-series surveys and same-day conversation circle feedback forms. Surveys were administered using SurveyMonkey and descriptive statistics were completed in Excel and SPSS.

Results

A total of 42 participants were enrolled, most of whom were people of colour who had not participated in research before. Among them, 40 participants provided at least one response to a same-day conversation circle feedback form, which reflected consistently positive experiences with BRAID. This was consistent with evidence from the post-series survey, in which every one of the 36 respondents stated that they would either definitely (83.3%, N = 30/36) or probably (16.7%, N = 6/36) recommend participation in BRAID to someone like them. Of note, 91.7% (N = 33/36) had already disseminated health information learned through BRAID downstream and 84.4% (N = 27/32) indicated that BRAID strengthened their trust in science and research, highlighting unique and distinguishing features of the model.

Conclusions

Our quantitative evidence suggests that BRAID is effective, efficient and scalable, with experiential evidence supporting that it is reproducible. These factors suggest that BRAID implementation can facilitate rapid, bidirectional information sharing that builds trust between healthcare organizations and communities. This has laid the groundwork for an LHCC in the Bronx, with the potential to be adopted by healthcare organizations elsewhere.

Background

From learning healthcare systems to learning healthcare communities

Throughout the 21st century, the United States healthcare system’s top-line objectives have consistently been to improve health outcomes and address health disparities. [1]. After including these priorities in four consecutive iterations of the United States Department of Health and Human Services’ Healthy People national objectives, they have been embraced as driving forces behind much health sector activity, from policy and payment reforms to new care delivery and community engagement models [1]. This has included the development of learning health systems, and more recently, learning healthcare communities.

Learning health systems (LHS) systematically integrate internal data and experience with external evidence to enhance healthcare quality, safety and efficiency [2]. The LHS model stipulates three core requirements: (1) systems for data capture, (2) care improvement targets and (3) a supportive policy environment [2]. Adoption of the LHS has helped drive many meaningful advancements in healthcare; in particular, it has formed the basis for many value-based payment reforms and quality improvement initiatives [3,4,5]. However, LHS implementation requires significant time, resources and expertise, creating barriers to entry for many [2, 3]. This may be especially true for resource-limited healthcare organizations that predominantly serve medically and socially complex patient populations, for whom advancing health equity is paramount.

Another challenge in advancing health equity is the fact that the LHS model did explicitly include it as a top-line goal [2]. Despite this, health equity is now widely regarded as a mandate for the US healthcare system [6, 7], particularly since the coronavirus disease 2019 (COVID-19) pandemic illuminated its arguably ubiquitous inequities [2, 8, 9], which drive medical mistrust in marginalized communities such as ours in the Bronx, New York, United States [10,11,12,13] Throughout the pandemic, healthcare systems operating within the traditionally paternalistic care delivery paradigm repeatedly struggled to roll out equitable COVID-19 testing, treatment and vaccination strategies [14,15,16], disproportionately harming marginalized communities who were already vulnerable to the interlocking public health crises wrought by the virus [17, 18]. Despite this overwhelming failure, some innovative care delivery models proved more effective, with one example being those that participated in the National Institute of Health’s Community Engagement Alliance (CEAL), who saw greater advancement by prioritizing true partnership with the communities they serve [19]. This evidence warrants revisitation of the LHS model.

Learning healthcare communities (LHCC) were designed as a next-generation evolution of the LHS. Positing that actively and continuously engaging communities can better account for and ultimately address systemic health inequity, the LHCC retains the same core LHS elements, but adds an important fourth: active, continuous bilateral community engagement [2]. LHCC theorists argue that transforming LHSs into LHCCs can help healthcare organizations more readily foster shared accountability for improving patient and public health, thereby advancing health equity in a meaningful and sustainable way [2]. Returning to the COVID-19 pandemic example, whereas a LHS might have implored a hard-hit community to get vaccinated via public service announcements handed down by public health experts, a LHCC might instead engage community members regarding what health messages would resonate most in their communities and who is best positioned to deliver them, similar to the community engagement approaches used by the NIH CEAL networks [19]. In this way, LHCCs can reject paternalistic models of care that attempt to generate retroactive so-called buy-in around health system-identified solutions, and instead aim to nurture real, proactive, longitudinal community engagement, partnership and investment.

Thus, formation of an LHCC is predicated upon having a purpose-built community engagement apparatus. However, a model for one has not yet been defined in the literature, likely because the LHCC is itself a nascent concept. Encouragingly, healthcare organizations already employ various methods of community engagement, including community health needs assessments and resource mapping, community advisory boards, community forums and community-based participatory research (CBPR) [20]. In fact, these activities are often required of healthcare organizations seeking to retain US government funding [21, 22].

Among these familiar strategies, CBPR in particular holds substantial potential to enable active, continuous community engagement while redressing the lack of trust that has long hampered it [23,24,25]. In conducting research with, rather than on, communities, CBPR seeks to illuminate issues and develop effective, enduring interventions that address them. Moreover, research shows that CBPR models that prioritize reciprocity, authenticity and sustainability can also build much-needed trust among marginalized communities [25]. Taken together, these strengths can enable health systems to first develop a community engagement apparatus necessary for LHCC enablement, and then move that apparatus along the International Association for Public Participation (IAP2) Spectrum of Public Participation to progress from community-informed action toward real community empowerment [26, 27].

Bridging Research, Accurate Information and Dialogue (BRAID): a theoretical model

Bridging Research, Accurate Information and Dialogue (BRAID) is one such CBPR model that was specifically designed to address systemic health inequity by (1) fostering meaningful, longitudinal engagement between community and health system leaders and (2) building community trust. It therefore provides a natural infrastructure to support the formation of LHCCs, which is predicated upon having a mechanism to support active, continuous community engagement that advances health equity.

The BRAID model was cocreated by clinicians, researchers, staff and community members at the Montefiore Medical Center/Albert Einstein College of Medicine, which is the largest healthcare provider in the Bronx, New York, United States [18, 28]. Montefiore/Einstein serves a demographically diverse and socially and medically complex community. Of the 1.4 million people living in the Bronx, 44% identify as Black/African American and 57% identify as Hispanic/Latino. Emblematic of the many entrenched issues facing this population, more than 27% live in poverty and more than 8% lack health insurance [29]. These realities have necessarily shaped health system strategy, generating a substantial interest in and appetite for innovative approaches to addressing systemic health inequity in the Bronx. This, in turn, potentiated support for BRAID, particularly in light of its prospective role in LHCC enablement.

BRAID builds upon the traditional CBPR models by integrating best practices from validated community engagement, trust-building, public participation and experienced-based co-design models [26, 30,31,32,33]. The model, an overview of which is depicted in Fig. 1, convenes clinicians, scientists and members of the community in a series of semi-structured dialogues known as conversation circles.

Fig. 1

Schematic overview of the BRAID model

Before conversation circles begin, community members are recruited to BRAID on the basis of their potential to serve as “community experts”. The process of identifying community experts is largely outsourced to community-based organizations in an effort to honour community expertise, and likewise divest the health system from its role in gatekeeping participant selection. With that in mind, the role of community experts in BRAID is twofold. First, as in the conceptually similar Community Engagement Studios model, BRAID’s community experts participate in the grassroots identification of community health needs and co-production of solutions during phase one of the model [32, 33]. However, their role extends beyond this in phase two, during which community experts serve as trusted messengers, a concept analogous to Popular Opinion Leaders [34, 35]. In this way, BRAID’s community experts not only provide vital input into the BRAID model, but also share its outputs.

Phase one of BRAID focuses on building trust and co-designing health messages. The research team first proposes a set of trust-building, evidence-based guiding principles to establish a safe space for honest dialogue between community experts and health system representatives [30]. All participants are encouraged to revise and amend these guiding principles on the basis of their identities, experiences and priorities. Next, conversation around a salient health equity topic is initiated. To date, topics have been selected on the basis of health system-identified public health and research priorities (e.g. vaccine hesitancy, clinical trial diversity and sharing research findings back to the community); however, in the future, both community and health system partners would ideally propose ideas. During the conversation circles, research team members use motivational interviewing-aligned facilitation techniques to ensure that the conversation remains centred on participants as they share experiences, concerns and questions on behalf of their communities [36]. This empowers the community experts, rather than the research team or invited guests, to drive the conversation. For example, information unearthed at each circle often guides the selection of future invited guests, including clinicians and scientists, who can in turn provide attendees with information and resources tailored specifically to their needs. As this process continues, trust is built, creating an environment conducive to the co-production of health messages that capture what matters to the community [37].

During phase two of BRAID, community experts lean into their role as trusted messengers as they transition into the role of “BRAIDers”. Having originally been recruited due in large part to their social capital, BRAIDers willingly disseminate co-produced health messages throughout their already-robust social networks in a process called BRAIDing. The reach and impact of this process is sizeable in theory, with efforts underway to measure it in practice.

Furthermore, BRAIDers can help identify new community concerns in phase two, feeding that input back into the clinical and scientific response loop to inform future action at the individual or institutional levels. For example, after voicing their church’s desire to learn more about Alzheimer’s disease in the Bronx, one BRAIDer was connected with physician–researchers who presented to and shared informational resources with their congregation. In the future, this process could inform larger-scale health system action – including ideas for new BRAID circles themselves.

In alignment with the Spectrum of Increasing Public Participation, the power within BRAID resides firmly with the community rather than the research team or invited guests [26]. As depicted in Fig. 2, elements of the BRAID model reflect several modes of public participation in an effort to promote true community–health system partnership [26]. This further distinguishes the BRAID model from its contemporaries, which more commonly embody only one of these strategies.

Fig. 2

Alignment of BRAID along the spectrum of public participation

Bridging Research, Accurate Information and Dialogue (BRAID) in practice

The theoretical BRAID model has now been applied in practice with three distinct cohorts. The first was a pilot held during the COVID-19 pandemic that aimed to build COVID-19 vaccine confidence (BRAID-CVC) within our Bronx communities, findings from which have already been discussed by Stephenson-Hunter et al. (2023) and Gutnick et al. (2023 and 2024) [18, 28, 38]. Through this early pilot experience, BRAID identified significant opportunities to enhance its model.

BRAID-CVC initially sought out community members to represent the community’s voice in conversation circles, but did not pre-screen these individuals to understand their level of expertise on and influence within their communities; in other words, it recruited community members rather than community experts. Participants therefore had varying degrees of confidence in their ability to serve as a representative voice for their community and disseminate learnings downstream through their social networks. In subsequent cohorts, BRAID ameliorated this by developing a systematic but nimble process for selecting community experts, which compensates community-based organizations (CBOs) for identifying potential trusted messengers.

As BRAID planned to scale its network of BRAIDers, action was taken to address the associated constraints on research team members’ time and resources. Fortunately, BRAID is integrated with the Montefiore Office of Community and Population Health, which is already charged with executing the health system’s community engagement strategy. This allowed BRAID to make use of existing personnel, relationships, technology and systems. For example, a project manager began overseeing operations, a community liaison took on relationship management and medical students and pre-health pipeline students helped collect and analyse qualitative and quantitative data. In this way, BRAID’s small core team could flex its support to meet the demands of the work.

With these changes in place, two additional cohorts focused on diversity in clinical trials generally (BRAID-CT) and mental health clinical trials specifically (BRAID-CT-MH) were planned and executed between 2022 and 2023. This manuscript leverages quantitative data from those cohorts to explore the efficacy, efficiency, scalability and reproducibility of the BRAID model in view of its potential role in LHCC enablement.

Methods

In this mixed-methods cohort study, we present quantitative survey data from two distinct BRAID cohorts (BRAID-CT and BRAID-CT-MH) that took place between 2022 and 2023. The first cohort, BRAID-CT, focused on diversity in clinical trials, and the second, BRAID-CT-MH, focused more specifically on diversity in clinical trials regarding mental health. These topics were selected on the basis of their relevance to both our community in the Bronx and the scientific and research community. The study was approved by the Institutional Review Board at the Albert Einstein College of Medicine. BRAID is supported by the National Center for Advancing Translational Sciences (NCATS), National Institutes of Health, through Clinical and Translational Science Awards (CTSA) award numbers (see UM1 or UL1 numbers below). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

For BRAID-CT, which is further funded by Genentech, participants were recruited by contacting community organizations known to the Montefiore Office of Community and Population Health, which in turn identified community experts. For BRAID-CT-MH, which is further supported by a grant from the American Psychiatric Association Substance Abuse and Mental Health Services Administration (SAMHSA) Minority Fellowship Program awarded to a member of our team (N.G.), participants were recruited from three local mental health clinic patient advisory boards and a mental health club house program. Eligibility criteria for each cohort are available in the implementation toolkit included in Appendix I. Informed consent was obtained from all participants.

Once enrolled, each participant attended a series of four virtual conversation circles. A mix of quantitative and qualitative data was collected before, during and after the session series. A pre-survey gauged participants’ baseline trust in healthcare, research and clinical trials, and a post-survey evaluated those items as well as their overarching perceptions of BRAID. After each meeting, a brief post-conversation circle feedback form assessed participant experiences with BRAID that day. All surveys were conducted using SurveyMonkey, and descriptive statistics were completed in Excel and SPSS. Furthermore, each conversation circle was recorded and transcribed for qualitative analysis. Qualitative data analysis is not presented here and will be explored in future papers.

For this study, the BRAID research team included individuals affiliated with the Albert Einstein College of Medicine (A.P., B.R., D.G., J.C., K.W., N.G., N.Y., S.J. and S.M.) and Montefiore Medical Center (B.R., D.G., D.W. and N.G.), where BRAID was developed. All research activities were completed in the Bronx, New York, United States. Aligned with best practices in community trust-building, all BRAID participants were also compensated for their time and expertise.

Results

There were 42 total participants across BRAID-CT and BRAID-CT-MH, most of whom were people of colour who had not participated in a research study before. Demographics were similar across BRAID-CT and CT-MH, with most participants being non-Hispanic Black (20/42, 47.6%) or Hispanic (16/42, 38.1%), reflecting the diversity of the Bronx. Most (24/42, 57.1%) had previously never been asked to participate in research, and even fewer (14/42, 33.3%) had actually done so. There was one key distinction between the groups: all BRAID-CT-MH participants had a diagnosed behavioural health condition. Table 1 provides a complete overview of participant demographics.

Table 1 Demographic characteristics of BRAID-CT and CT-MH participants

Insight collected from post-conversation circle feedback forms illuminated same-day experiences with BRAID. The form asked participants to rate the extent to which they agreed with various statements regarding that day’s BRAID circle on a 5-point Likert scale, from strongly disagree (1) to strongly agree (5).

As presented in Table 2, there were 87 feedback forms submitted overall from 40 unique participants; thus, 40 participants submitted feedback on at least one BRAID circle, 31 submitted feedback on at least two circles, 13 did so for at least three circles and 3 did so for all four. Despite this attrition in completion rates, perceptions of BRAID remained consistently positive among respondents: the overall mean across all eight statements exceeded four (agree), with most values being even higher. This remains true even for participants who only provided feedback on one circle – oftentimes, the first one they attended – with the overall mean of six out of eight statements exceeding four (agree), and the remaining two coming very close.

Table 2 Same-day feedback on experiences with BRAID conversation circles

Furthermore, post-surveys completed at the conclusion of the conversation circle series concluded captured perspectives on BRAID overall. As presented in Table 3, 84.4% of respondents (N = 27/32) stated that BRAID strengthened their trust in science and research, with three additional respondents providing affirmative free-text answers such as “I understand more about the process”. Of note, both respondents who indicated that BRAID did not strengthen their trust in science and research were members of the BRAID-CT-MH cohort. Multiple unique features of the BRAID model were cited as factors that contributed to trust-building, including learning about the importance of diversity in research (88.2%, N = 30/34), having the ability to directly ask questions of researchers (82.4%, N = 28/34) and scientists (58.8, N = 20/34%), seeing data (67.6%, N = 23/34) and learning about health disparities affecting the community (64.7%, N = 22/34).

Table 3 Post-series perspectives on the BRAID model

Excitingly, most respondents indicated that they had already shared information acquired through BRAID with another person (91.7%, N = 33/36), including family, friends, neighbours and fellow community leaders. All respondents (100.0%, N = 36/36) stated that they would either definitely (83.3%, N = 30/36) or probably (16.7%, N = 6/36) encourage other people like them to participate in a BRAID circle.

Discussion

Evaluating the BRAID model

These findings support the notion that BRAID may be an effective, efficient, scalable and reproducible CBPR model. BRAID’s effectiveness can be considered in the context of its ability to achieve its primary objectives of building trust and enabling longitudinal engagement between the health system and the community. On the former point, most participants indeed affirmed that BRAID increased their trust in healthcare, science and research. This was not only true at the conclusion of the entire conversation circle series, as presented in Table 3, but was affirmed even after participants attended just one circle, as captured in Table 2. Participants agreed that BRAID’s unique features – in particular, learning information tailored specifically to their socio-cultural demographic and having a direct line to scientists and researchers – contributed to building trust.

The latter point regarding longitudinal engagement is most convincingly evidenced experientially: many former BRAID participants have remained actively engaged with the research team and the Montefiore Office of Community and Population Health to this day, including by helping to develop educational, outreach and evaluation materials; participating in manuscript development; and recruiting future BRAID cohorts [39]. Some have been connected to additional skill-building opportunities, such as fully funded enrolment in a community college course designed to help community members build research and communication skills to help address the cancer burden in New York City. These activities exemplify the win–win nature of the bidirectional community–health system relationships that BRAID can foster.

Moreover, survey data suggest that BRAID is efficient. The fact that nearly every participant had already shared information they learned downstream to members of their community, including other community leaders, demonstrates that BRAIDing enables rapid dissemination of accurate, impactful and trusted co-designed health messages. Efforts are already underway to develop additional capacity to evaluate the efficiency of each BRAIDer by using a proprietary website that tracks message dissemination through social media. In parallel, the research team is developing predictive models that can identify characteristics of a successful BRAIDer to further enhance recruitment.

Perhaps most strikingly, every community expert stated that they would encourage other people like them to participate in a BRAID conversation circle. This speaks to the model’s potential scalability. If considered as analogous to the widely used net promoter score, these data would indicate that BRAID exclusively has promoters, without any detractors [40]. Moreover, our promoters are inherently influential given their position as community experts. This makes them well positioned to recruit additional community experts for future BRAID cohorts, should they remain engaged in the model themselves.

Finally, experience has now demonstrated that BRAID can consistently receive positive feedback while being tested with different cohorts and focus areas, lending credence to the model’s reproducibility. This began during the pilot, in which participants reported feeling respected, heard and empowered during conversations surrounding the COVID-19 vaccine [18, 38], and remained true over two additional years of cohorts focused on the distinct topic of diversity in clinical trials.

Thus, data from two cohorts of 42 participants suggests that BRAID could be an effective, efficient, scalable and reproducible model for LHSs to adapt as they work to convert into LHCCs. Indeed, BRAID not only offers a mechanism for actively and continuously gathering community feedback that reflects what matters, but further offers healthcare organizations the opportunity to move their existing engagement strategies along the IAP2 Spectrum of Public Participation towards true community empowerment [26, 27, 34]. This is particularly important given the known shortcomings of the LHS; most notably, its inability to adequately address entrenched health equity issues [2].

Implementing the BRAID model

Multiple enabling factors can prime organizations to rapidly deploy and expand the BRAID model with fidelity. The first is strategic alignment. Our health system, like many, is guided by a multiyear strategic plan that includes values which closely align with the foundational principles of BRAID [41]. This strategic plan is not in name only; rather, it actionably guides our health system’s efforts, including the development of a robust community engagement strategy, under which BRAID sits.

The next is resource mobilization. Responsibility for executing Montefiore’s community engagement strategy resides with our existing Office of Community and Population Health, which serves as a bridge between the medical centre, the medical school and the community. In addition to housing BRAID, the office operates multiple community-based health equity programs and partnerships. Such an office can provide a foot in the door to leverageable community and academic relationships that can be used in recruitment of community and clinical and scientific experts, respectively. It can also provide the personnel needed to staff the model and the infrastructure needed to source additional grant funding.

The final piece is a champion. Within our team, D.G. is a trusted and well-connected leader within the medical centre, the medical school and the community. This enables her to generate excitement around BRAID’s potential to simultaneously address the priorities and needs of the health system and the community. Moreover, it positions her to identify potential future collaborators whose work may benefit BRAID implementation and secure the resources to support it. There are otherwise few core team member roles needed to stand up the model: a project manager is essential for operations, and community liaisons are vital to community relationship management. Beyond this, our team leverages the help of pre-health pipeline program and medical students interested in gaining research experience. Invited guests are not compensated, as many clinicians and scientists already view community engagement as their inherent professional responsibility. This makes the model highly nimble, augmenting its reproducibility.

Alignment with these features can promote ready adoption of the BRAID model. However, lacking them should not be considered prohibitive; rather, this obstacle simply presents another opportunity to apply BRAID to address systemic challenges in a community-engaged manner. Resources are available for healthcare organizations seeking to learn more about the BRAID model or implement it themselves. These educational, outreach and evaluation materials are included in the Appendix, and will soon be offered as an implementation toolkit.

Conclusions

Bridging Research, Accurate Information and Dialogue (BRAID) is a novel community-based participatory research model developed at the Albert Einstein College of Medicine/Montefiore Medicine. It was designed to ameliorate systemic health inequity by facilitating longitudinal trust-building dialogues between clinicians, researchers, and community experts. Data from multiple distinct cohorts supports that is effective, efficient, scalable and reproducible; thus, it can serve as foundational community engagement infrastructure for healthcare organizations interested in forming a LHCC.