Advancing Food Is Medicine Research: An EPIS-Informed Qualitative Study of Stakeholder Views

Abstract

Introduction

Diet quality in the United States remains poor and contributes to chronic disease. Food is medicine (FIM) interventions integrate healthy foods into health care to manage or treat diet-related illness. Although evidence supports the nutrition–health link, less is known about how to implement FIM programs effectively in real-world settings or how to design FIM research across complex delivery systems. Guided by the Exploration, Preparation, Implementation, and Sustainment (EPIS) framework, this qualitative study examines early implementation experiences from the American Heart Association’s Health Care by Food™ (HCXF) initiative to inform future FIM research.

Methods

We conducted five virtual focus groups with HCXF-funded FIM researchers and community-based organization (CBO) partners and interviewed one food vendor. Discussions explored lessons learned, evidence gaps, methodological improvements, implementation barriers, motivations for participation, and future research priorities. We performed thematic analysis and mapped themes to the EPIS framework.

Results

Themes spanned all EPIS phases. In Exploration, participants described a fragmented evidence base, measurement gaps, and uncertainty about payer requirements amid shifting policy. In Preparation, they emphasized earlier engagement of CBOs and vendors, limited partner research capacity, and tensions between standardization and cultural tailoring. In Implementation, unclear roles, logistical complexity, and variable capacity affected feasibility; recruitment and retention relied on trust-based engagement, and technology was helpful but constrained by access and use. In Sustainment, participants highlighted outcomes beyond clinical measures, payer-specific communication and champions, and the need for stronger national coordination to support scale and coverage.

Conclusions

Early HCXF implementation experiences suggest that advancing the FIM evidence base will require larger, more coordinated studies and clearer approaches to partnership governance, measurement, cultural adaptation, and trust-based engagement. Aligning implementation science–grounded study designs with payer needs, capturing outcomes beyond clinical endpoints, and strengthening national coordination may accelerate translation into scalable models and sustainable coverage.

Keywords

food is medicine nutrition community health focus groups qualitative methods

Background

Diet quality in the United States remains poor. Most individuals do not meet national recommendations for fruits, vegetables, and whole grains and exceed recommended limits for sodium, saturated fat, and added sugars (per the 2020–2025 Dietary Guidelines for Americans¹), contributing to chronic disease burden.^2,3 Food is medicine (FIM) programs seek to address diet-related chronic disease by providing healthy foods as part of health care aligned interventions, including medically tailored meals, medically tailored groceries, and produce prescriptions.⁴ This approach reflects a shift from nutrition counseling alone toward incorporating food provision into clinical care and aligns with evidence linking poor diet to cardiovascular outcomes and premature mortality,⁵ as well as policy attention to the costs of diet-related conditions.⁶

A growing evidence base suggests that some FIM models can improve intermediate outcomes such as diet quality and nutrition security, and in some studies cardiometabolic risk markers.⁵ Randomized trials in the United States remain limited in number and many important questions about intervention dose, duration, populations, and outcomes remain to be answered more systematically.

Even with this emerging evidence, the field lacks consistent guidance on how to deliver and study these interventions in routine care and community settings, where multi-partner coordination, workflow integration, staffing, and data capture shape what is feasible and how results should be interpreted for scale and coverage. FIM research has begun to examine how programs are designed and delivered in real-world settings, including FIM programs implemented through citywide systems,⁷ food bank and health care partnerships,⁸ Federally Qualified Health Center (FQHC) collaborations,^9-11 and U.S. Department of Veterans Affairs^12,13 and hospital workflows.^14-17 Qualitative studies in this area describe recurring implementation issues that are directly relevant to conducting FIM research, including partner coordination,^8,9,18 workflow integration,^{8,12,13,15,19} staffing capacity,^8,9,15,18 referral and navigation processes,^12,13,15 and limits in data capture^8,20 needed to assess engagement and outcomes. Implementation science guidance for FIM research in health care settings is still developing, leaving feasibility, scalability, and sustainment incompletely characterized across routine practice contexts.²¹

Many FIM interventions rely on coordination across health care organizations, community-based organizations (CBOs), and food vendors to deliver services and support patients, yet there is less empirical guidance on how to integrate these multi-partner models into clinical and CBO workflows, particularly in resource-constrained settings. Implementation science frameworks can support this work by specifying how contextual factors shape program delivery across phases of adoption and operation. The Exploration, Preparation, Implementation, and Sustainment (EPIS) framework offers an organizing structure to examine multilevel determinants of FIM research and implementation and to interpret variation in delivery and outcomes beyond efficacy alone.²¹

In 2022, the American Heart Association launched the Health Care by Food™ (HCXF) initiative to strengthen the FIM evidence base and inform coverage decisions. As of 2026, HCXF has supported over 40 FIM research projects and planning grants, including ones focused on enrollment, engagement, and short-term behavior change. This study synthesizes early implementation experiences across the HCXF portfolio to inform the design and execution of the next phase of research as HCXF moves toward larger evaluations of clinical outcomes and cost effectiveness.⁶ Because implementation conditions in multi-partner delivery systems shape who is reached, what is delivered, and how consistently outcomes can be measured across sites, identifying these determinants is critical for producing valid, interpretable, and coverage relevant evidence.

Methods

We used a qualitative design with focus groups and interviews involving three stakeholder groups engaged in FIM research: researchers, CBO representatives, and a food vendor. We used purposive sampling to recruit participants. Eligible participants were directly involved in implementing HCXF-funded projects, along with one externally funded project included at the recommendation of HCXF leadership because it was similar in scope and implementation timeframe to the HCXF portfolio. The HCXF research team distributed email invitations using program contact lists. For each HCXF project, we invited the principal investigator and key implementation partners, including the CBO partner and food vendor involved in delivery.

Data Collection

In July 2025, we conducted five virtual focus groups (90 minutes each) and one individual interview (60 minutes) via Microsoft Teams. Three focus groups included HCXF-funded researchers, two included CBO representatives, and the interview included a food vendor. Sessions were organized by stakeholder role to facilitate discussion among participants with similar responsibilities.

The discussion guide covered lessons learned, evidence gaps, opportunities to strengthen study methods, implementation barriers, and motivations for research participation. A facilitator from the HCXF team led each session with support from two co-facilitators. The co-facilitators had prior involvement in HCXF and were known to many participants, whereas the lead facilitator had limited prior interaction with them.

Participants received a $75 gift card. All participants provided informed consent. The Heartland Institutional Review Board approved the study on May 20, 2025 (project no. 052025-1205).

Data Analysis

We coded de-identified transcripts inductively to develop themes grounded in participants’ accounts. The analytic team included four members with training in qualitative research and implementation science; several had prior involvement in HCXF, which informed contextual understanding but also necessitated attention to reflexivity. To address this, team members engaged in ongoing memoing and discussion to identify potential assumptions during analysis.

An initial codebook was developed through close reading of a subset of transcripts, using an inductive coding approach to identify preliminary codes. All four team members then independently coded each transcript in Dedoose, applying the shared codebook while remaining open to the addition of new codes as they emerged from the data. The coding framework was iteratively refined through repeated review of the data, with codes added, collapsed, or clarified based on team discussion and comparison across transcripts.

The team met regularly to review coding decisions and resolve differences through discussion and consensus. Rather than calculating interrater reliability, we emphasized analytic consistency through iterative consensus building, in line with interpretative qualitative methodologies. We maintained an audit trail and analytic memos documenting coding decisions, codebook iteration (including addition, refinement, and consolidation of codes), and summaries of team discussions regarding theme development.

Following coding, we grouped related codes into broader categories and examined patterns across transcripts to develop themes that captured recurring concepts and variations in participants’ experiences. Themes were reviewed and refined through comparison with the full dataset, and attention was given to identifying both convergent and divergent perspectives.

After inductive theme development, we organized themes within the EPIS framework to support interpretation within an implementation science context. This mapping was conducted after themes were generated inductively, allowing the framework to guide interpretation rather than constrain initial analysis. Themes were grouped according to the four EPIS phases: Exploration (identifying needs and assessing fit), Preparation (building capacity, partnerships, and readiness), Implementation (addressing delivery and operational challenges), and Sustainment (maintaining, adapting, and scaling interventions). Themes were also aligned with the appropriate EPIS context layer and construct, including outer context factors (e.g., sociopolitical context, funding, leadership, and patient/client characteristics), inner context factors related to organizational and individual characteristics, bridging factors that connect systems across settings, and innovation-related characteristics influencing implementation. Any discrepancies in framework alignment were resolved through team discussions and consensus.

We assessed thematic sufficiency by monitoring the emergence of new codes and themes across transcripts; as analysis progressed, few novel concepts emerged, suggesting that additional data collection was unlikely to yield substantially new insights. Unless otherwise noted, themes were consistent across stakeholder roles, including researchers, CBO representatives, and the food vendor.

Results

Participant Characteristics

We invited 38 researchers and 18 participated (47%), 13 CBO representatives and seven participated (54%), and three food vendor representatives and one participated (33%). Overall, 26 of 54 invited individuals participated (48%). All participants who enrolled attended their scheduled sessions (Table 1). For each research study, a maximum of one individual per stakeholder group participated.

Table 1.

Study Participants

	Researchers	CBOs	Food vendors
Interview			1
Focus group 1	5
Focus group 2		3
Focus group 3	6
Focus group 4
Focus group 5	7	4
TOTAL	18	7	1

Exploration

Participants generally viewed the diet and health relationship as well established, but they described the FIM evidence base as dispersed across small pilot studies and narrow intervention models, which complicated decisions about adoption and scale. This fragmentation was associated with persistent uncertainty about core intervention parameters, including dose, duration, and tapering, limiting the extent to which findings could inform broader implementation or policy decisions. Across stakeholder groups, a need for larger, coordinated studies to address these unresolved intervention design questions was described.

Measurement and data infrastructure limitations were also described, including difficulty assessing changes in diet quality, tracking adherence over time, and using electronic medical record data. These constraints reduced opportunities for implementation learning and limited the evidence available to inform sustainment.

FIM research was described as a pathway to policy change, but researchers reported limited clarity on the evidence payers require. Participants indicated that study questions do not always align with coverage decision points and that shifting federal and state policy environments add uncertainty. Cost effectiveness was viewed as relevant, although many questioned whether it is sufficient on its own to support broader system change.

Preparation

CBOs and the vendor described a preference for full partnership roles rather than late-stage implementation support. They emphasized early involvement in study design, budgeting, and outcome selection and formal recognition through mechanisms such as co-investigator roles or subawards. When engaged after protocols were set, partners reported frustration, difficulty navigating academic requirements, and unclear expectations within research teams. Across stakeholder groups, collaborative problem solving was described as a source of practical, low-cost solutions that leveraged existing resources within CBOs, vendors, and other organizations.

Constraints in CBO staffing, data systems, and logistics were viewed as limiting capacity to participate in research and deliver complex interventions. Researchers and CBOs emphasized the need for investment in research readiness, particularly for smaller CBOs that were viewed as trusted community partners but had limited resources for large scale implementation. In some settings, basic infrastructure gaps such as limited access to clean water, refrigeration, or childcare shaped what could realistically be delivered. Real-world barriers sometimes limited delivery even when they were not a part of the intervention itself.

Across stakeholder groups, a persistent tension between standardization and flexibility in FIM interventions was described. Standardized protocols were viewed as supporting scalability, evaluability, and payer confidence, whereas flexibility was described as necessary to reflect cultural practices and community preferences. Cultural tailoring was considered central to intervention fit but difficult to operationalize given variation in dietary preferences and traditional food practices.

Implementation

Role clarity was described as a key implementation requirement in multi-organization FIM studies. When responsibilities and communication channels were not specified, coordination was less efficient and clients or patients expressed uncertainty about whom to contact for questions or support.

Researchers and practitioners reported strong staff engagement in FIM programs, and staff often described providing food as meaningful work. Practitioners viewed FIM as expanding their ability to address patient needs that previously lacked a programmatic response. Participants indicated that this engagement supported morale and organizational readiness and may facilitate program uptake.

Implementation barriers were identified related to food sourcing and distribution, long travel distances in rural settings, and the coordination demands of multi-organization delivery. Some questioned the long-term scalability of models such as medically tailored meals. Variation in vendor capacity and limited provider engagement were viewed as additional constraints on consistency and quality.

Recruitment and retention were characterized as difficult, and participants contrasted FIM clients with the convenience samples often used in research; one participant characterized it as an “inconvenient sample,” but one that was meaningful. Engagement was attributed primarily to trust and relationships rather than program features or technology. Advisory boards, incorporation of lived experience, and practical incentives, including cooking tools, were framed as strategies that may support participation and continued involvement.

Participants viewed technology as a potential support for efficiency and access, but described constraints related to limited digital access, low app use, and high mobility among populations facing social and economic constraints. Many indicated that in the populations they serve that direct human contact was more effective than digital approaches for establishing trust and sustaining engagement.

Sustainment

Researchers and CBOs described outcomes that extend beyond individual clinical measures, including cultural connection, social relationships, local economic activity, and environmental considerations. They also noted that benefits may accrue at the household level, yet current evaluations were viewed as rarely capturing these broader effects.

Messaging, champions, and advocacy were described as influencing coverage decisions for FIM. Several noted that evidence alone may not be persuasive to policymakers or insurers and that narrative communication can carry weight in these settings. They also noted that advocacy approaches differ by payer type and that the field still lacks a coordinated strategy for engaging insurers.

The FIM field was described as expanding but not well coordinated. Many called for stronger national leadership to set shared priorities, align research efforts, and develop tools that connect local studies to broader policy change.

Table 2 maps the implementation themes identified in this study to the EPIS framework by phase, context layer, construct, and summarizes the corresponding key insights with illustrative quotes.

Table 2.

Mapping FIM Implementation Themes to the EPIS Framework

EPIS phase & context layer: Construct	Implementation theme	Key insight	Quotes
Exploration
Innovation factors: Innovation characteristics	Fragmented evidence base	Fragmented and small-scale FIM studies limit shared learning and leave key questions about intervention design and measurement unresolved.	“How do we move things to be more broadly adopted across the healthcare system? …right now we’ve got these pockets where we’re producing evidence but it’s very patchworked and small scale. So how do we go bigger?” (R)
Outer context: Funding	Research to policy gap	Uncertainty about how well current FIM research aligns with payer evidence needs and policy decision-making.	“And so really what I would like to know is with great specificity, ‘What is the price point you’re going to need to see? What is the ROI you’re going to need to see? What are all of these elements that you will need to see?’ And then sign on the dotted line that if we can deliver that, that you’ll cover it. Because I think the lack of clarity is often what’s holding us back and we’re going on wild goose chases.” (R)
Outer context: Sociopolitical	Policy instability	FIM research is seen as a route to policy change, but the shifting federal/state environment creates uncertainty.	“We are trying to advance policy change. We have a very robust Section 1115 waiver here that we’ve advocated for some time. And this research is really the lifeblood of these policy changes. And we plan to continue to work on research because it’s just so important to continuing to advance policy change.” (C)
Outer context: Sociopolitical	Policy instability		“We’re anticipating a lot of questions around what will happen with changes to the Medicaid program.” (C)
Preparation
Bridging factors: Community-academic partnerships	Partnership model	CBOs and vendors are pushing for full research partner status rather than being late-stage service providers.	“We can sometimes be seen as the meal provider or the intervention provider. But there’s more of an opportunity and there’s sort of a disparity across organizations as to the level of agency and equity in partnership with our academic partners and how much we’re able to inform anything from the budget, whether it’s appropriate in covering, you know, the cost, what does the intervention design look like? We’ve been doing this work for years and years and so how do we get to inform the intervention alongside our academic partners?” (C)
Bridging factors: Community-academic partnerships	Partnership model		“We were brought in after the study had already been designed… and so we think about how we know the limitations of our products… and [then we] had to stretch a little bit of [our product’s] functionality.” (V)
Inner context: Organizational characteristics	Capacity of CBOs	Limited staffing, data systems, and basic infrastructure constrain CBO capacity to participate fully in FIM research and implement complex interventions.	“It’s important to keep in mind that a lot of those places, they’ve never done research before. They’re very passionate about the idea. They really want to do this, and they’ve supported us in every way that they can, but they’ve never done research before. So there’s a huge learning curve.” (R)
Innovation factors: Innovation characteristics	Standardization vs personalization of interventions	Tension exists between the need for standardized protocols to support scale and payer confidence and the need for flexibility to accommodate cultural practices and community preferences.	“It does take more money and time to really curate meals, but it’s so important…” (R)
Implementation
Bridging factors: Community-academic partnerships	Roles and communication	Effective collaboration depends on establishing clear communication channels and defined operational roles to ensure role clarity and coordinated action between partnering organizations.	“There are a lot of moving parts within these research studies and I think really understanding the role that everyone plays [is important]. We had [X] involved and sometimes it was hard to know what next step was and who is owning that next step and really moving the needle forward.” (V)
Inner context: Individual characteristics	Staff engagement	Staff enthusiasm for FIM work was described as a key facilitator of morale, readiness, and successful program uptake.	“An extremely gratifying part of the project has been the feeling that we’re all kind of on the same team here and that everybody’s as excited about this as the evaluation team is and that there’s just enthusiastic support from all parties.” (R)
Inner context: Organizational characteristics	Logistical barriers	Operational barriers related to logistics, coordination, and variable partner capacity complicated implementation and raised concerns about scalability.	“I cannot even reiterate how often people’s addresses and phone numbers change. And so how keeping all of that, coordinating all that, is really a lot.” (C)
Outer context: Patient/client characteristics	Real-world recruitment	Including hard-to-reach populations ensures the evidence base incorporates diverse perspectives often excluded from clinical research.	“I worked in academic medicine and typically what we enroll is a sample of convenience. So the fact that we are enrolling a sample that is very inconvenient really warms my heart because those perspectives don’t necessarily make it into a lot of the big population studies.” (C)
Inner context: Individual characteristics	The human factor	Effective recruitment and retention depend heavily on trust-based, relationship-driven engagement	“A lot of the interviewees talked about the kind of human side and human element of like building trusting relationships with the participants and how important that was for redemption and engagement.” (R)
Innovation factors: Innovation characteristics	Dual-edged role of technology	Technology was seen as useful for efficiency but only if people use it	“I think the caveat of technology is that by only focusing on technology, we’re forgetting people who might need a paper form. We should not forget about everyone else who doesn’t have access.” (C)
Sustainment
Innovation factors: Innovation characteristics	Broader outcomes	Importance of capturing social, cultural, household, and community-level benefits that extend beyond individual clinical health outcomes.	“And all the ripple effects that come with that… If farmers and ranchers locally are now able to get their produce to market and sustain themselves, then they can actually expand and grow. And then there’s more availability of produce for everyone in the community… So trying to think about the benefits, just beyond the patient level… are very interesting questions.” (R)
Outer context: Patient/client advocacy	Messaging and advocacy	Messaging, advocacy, and trusted champions were viewed as essential complements to evidence in influencing coverage and policy decisions.	“We finally are getting some traction at a policy level… but it’s because we found a new, what I call a nutrition champion, who is working with us and meeting with us every two or three weeks and drafting things and saying tell me more, send me more. I didn’t know it exists.” (R)
Outer context: Leadership	National leadership	The FIM field is seen as growing but fragmented, with a need for stronger national leadership and coordination.	“Maybe it's happening and we don’t know about it, but I would just love to see more coordination among these leaders in terms of the next phases in research and the road map that is envisioned… so that we can get to that place of policy change that I think a lot of us want to see.” (C)

Participant quotes: R is researcher. C is CBO representative. V is vendor representative.

EPIS Framework

Figure 1 displays the study themes positioned within the EPIS phases and corresponding context layers, illustrating how determinants spanned outer context, inner context, bridging factors, and innovation characteristics.

Figure 1.

Study-informed EPIS model for FIM implementation

Discussion

This qualitative study applied the EPIS framework to examine implementation determinants shaping FIM research in routine delivery settings. Participants expressed optimism about FIM alongside unresolved questions about intervention parameters, measurement strategies, and the evidentiary pathway to coverage. These results support the view that implementation science can inform FIM study design by clarifying feasibility constraints, supporting partner alignment, and identifying factors likely to affect sustainment. As the body of implementation-focused studies grows, they reinforce the value of explicitly measuring feasibility, acceptability, and delivery constraints alongside clinical endpoints.^13,21,22

Across EPIS phases, participants identified both opportunities and tensions shaping the conduct and future direction of FIM research (Table 3).

Table 3.

Opportunities and Tensions Shaping FIM Research Across EPIS Phases

EPIS phase & insights	Opportunities/STRENGTHS	Tensions/Challenges
Exploration
Research evidence	Diet-health link is well established	Evidence dispersed across small pilots; unclear dose, duration, tapering; dilution from too many small questions
Policy and research alignment	Cost-effectiveness is valued; research viewed as pathway to policy	Payer needs unclear; evidence alone insufficient; shifting policy contexts creates instability
Preparation
Partnerships	Co-design among researchers, CBOs, and vendors yields practical solutions	Vendors and CBOs treated as contractors and brought in too late; unclear roles and responsibilities
CBO capacity	CBOs bring community trust and reach in underserved communities	Limited staff, data systems, and logistics; unmet social needs (e.g., housing, utilities, childcare) extend beyond research scope
Standardization vs personalization	Standardization supports scalability, payer confidence, and evaluation clarity	Personalization supports cultural fit and individual preferences but increases complexity and resource demands
Implementation
Staff engagement	Providing food described as rewarding and meaningful; boosts morale and readiness	Enthusiasm does not always translate into referrals, fidelity, or consistent program delivery
Operations	Food provision and distribution models spur innovation; local vendors strengthen community ties	Rural distances, uneven vendor capacity, high coordination burden, provider disengagement
Participant and community engagement	Trust, advisory boards, and lived experience improve recruitment and retention	Recruitment difficult; especially challenging in control groups; incentives inconsistently applied
Technology	Platforms expand reach, integrate SNAP/EBT, and reduce transportation barriers	Digital divide, transient populations, low app engagement, preference for human contact
Sustainment
Broader outcomes	Programs foster cultural pride, social connection, reduced stress, environmental benefits	Outcomes rarely prioritized by payers or systematically measured in evaluations
Messaging and advocacy	Storytelling and champions resonate with payers and policymakers	No clear playbook for insurer advocacy; strategies must vary by payer type
Field leadership and evolution	National organizations can set priorities, provide tools, and link micro-level studies to macro change	Field remains fragmented; varied stakeholder priorities; need for greater coordination

Supporting Cumulative Learning as the Evidence Base Expands

Participants’ descriptions of an evolving and heterogenous evidence base are consistent with rapid diversification in FIM research across intervention types, settings, and delivery partners. Heterogeneity in study designs supports innovation and may be necessary while optimal intervention parameters (including dose and duration) remain uncertain. However, cross-study synthesis is limited when similar intervention intensity is described or measured inconsistently across studies, even when variation in intensity itself is expected as the field tests different models. Reviews of FIM programs in specific populations similarly conclude that variation in program components and inconsistent reporting make it difficult to compare studies, reinforcing the need for shared definitions and common metrics.^20,23 These findings point to value in pairing local adaptation with greater coordination across projects, including shared definitions of core intervention components and use of a validated set of common measures, such as those developed by HCXF,²⁴ to support comparison and aggregation while allowing sites to tailor delivery models to context.

Stakeholders’ accounts also indicate that measurement and data infrastructure remain constraints on implementation learning. Limited capacity to assess diet quality, track adherence over time, and obtain complete follow up data from electronic records can reduce interpretability of outcome variation and obscure which implementation strategies are associated with better performance. Implementation research in food bank–health care partnerships and FQHC settings identifies persistent challenges in tracking and sharing program data and in staffing capacity, emphasizing why measurement infrastructure is foundational to implementation learning.^8,10,25

Aligning Research Questions With Policy and Coverage Decisions

Participants linked FIM research to policy change and coverage expansion but acknowledged uncertainty in how evidence is weighted in coverage decisions. Policy analyses of medically tailored meals suggest that even when economic models are favorable, decision-relevant evidence often depends on whether evaluations incorporate real-world variation in program design, dose, duration, and ancillary services.²⁶ Future studies may benefit from specifying the decision context at the design stage, since evidence needs may differ across Medicaid programs, commercial insurers, employers, and health systems. Participants also characterized cost effectiveness as relevant but insufficient on its own, emphasizing feasibility, operational requirements, and the role of credible messengers in communicating value. Implementation studies in Medicaid pilots further indicate that reach and enrollment can be constrained by contact instability, staff time, and workflow complexity; these are practical factors that shape real-world impact irrespective of modeled cost-effectiveness.²⁷ Study designs that align endpoints and analyses with a defined policy use case may produce findings that are more interpretable for the intended decision maker.

Treating Partnership Design as a Research Strength

CBOs and the vendor emphasized early involvement and formal recognition as research partners. Their accounts describe co-design as improving study fit and feasibility by incorporating delivery partner expertise in community engagement, logistics, and cultural fit. Hospital-based pilots that used participatory design and iterative testing demonstrate how early partner input can improve workflow integration (e.g., referral platforms) while also clarifying delivery costs and operational requirements relevant to scalability.¹⁵ Our findings support partnership structures that specify roles, decision authority, and communication channels at study initiation to reduce coordination burden and support consistent implementation and evaluation.

Participants also described capacity constraints among some CBOs, particularly smaller and local organizations that are often central to trust-based outreach. Across multi-partner models identified in the literature, recurring barriers include staffing capacity, partnership maintenance, and data-sharing infrastructure, suggesting that ‘research readiness’ investments are often prerequisites for stable implementation.^8,9 Our findings support a role for deliberate investment in CBO research readiness, including staffing, training, and data infrastructure, to support meaningful participation and sustain multi-partner collaboration.

Advancing the Field’s Approach to Standardization and Cultural Fit

Stakeholders described tension between standardization to support scalability and payer confidence and flexibility to reflect cultural practices and community preferences. FIM studies may benefit from specifying which elements must remain consistent across sites and which elements can be adapted. Perinatal FIM implementation research similarly argues for defining core program functions while allowing structured adaptation of form, with the aim of improving interpretability across sites.^20,23

Defining core intervention functions can support interpretation, while structured adaptation of form can improve cultural fit. For example, a multi-site produce prescription study could hold constant eligibility criteria, minimum dose, enrollment duration, the targeted behavior change goal, and primary outcomes with common follow-up time points. Sites could adapt delivery modalities and food options to local context and document what was changed and why, allowing subsequent assessment of whether specific adaptations are associated with engagement, retention, or diet quality across settings.

Designing for Real World Feasibility and Sustained Engagement

Participants described a gap between pilot level evidence and the real-world implementation conditions needed for long-term operation in community settings. Many of the factors that later affected implementation were identified early, during Exploration and Preparation, and participants reported that these determinants carried forward into later phases. Consistent with this, other multi-clinic FIM initiatives show that structured pre-implementation planning can substantially refine target populations and intervention specifications before delivery begins.¹¹ As study designs progress to embed in clinical workflows, lessons learned from involving vendors have substantial applicability to partnerships with clinical teams. Involving their perspectives early and having them as substantive partners in the research design will be critical in bridging the gaps between research and clinical implementation.

As the evidence base expands, studies should more directly address the operational demands of multi-partner FIM program delivery to support comparability and coverage decisions. Participants highlighted challenges including rural travel distances, variable vendor capacity, and coordination workload, consistent with prior studies emphasizing the importance of partner networks and infrastructure for reliable delivery and scale.^7,28 Future studies may benefit from assessing feasibility, partner workload, and operational requirements alongside effectiveness. Recruitment and retention were also described as dependent on trust-based engagement, with advisory boards, lived experience input, and practical supports functioning as core implementation strategies; consistent with prior work, engagement was facilitated by ongoing communication and navigation supports, while logistical and financial barriers persisted.^18,29 Participants also viewed technology as potentially improving efficiency and access, though feasibility was limited by digital access, low uptake, and population mobility, suggesting that digital tools are most effective when offered as optional supports alongside low-technology alternatives and direct human contact, particularly for participants with unstable contact information or limited device access, reinforcing the need for flexible, mixed-mode engagement approaches.^27,30

Broadening the Definition of Value to Support Sustainment

Stakeholders described outcomes beyond individual clinical measures, including household level benefits, social connection, cultural meaning, and community economic effects. Long-running health system programs have demonstrated the feasibility of embedding evaluation measures beyond clinical outcomes, including patient experience, program satisfaction, and local economic effects.¹⁴ These accounts are consistent with the view that FIM may produce impacts that are salient to participants and delivery partners even when they are not emphasized in payer decision making. Future studies may benefit from maintaining clinical and economic endpoints while also measuring broader impacts, particularly when these outcomes affect participation, partner commitment, and sustainment.

Participants also described the influence of champions, messaging, and advocacy on coverage decisions and noted that engagement approaches may differ by payer type. Research programs may benefit from specifying dissemination and translation strategies early, including identification of credible champions and tailoring messages to distinct decision maker audiences.

Limitations

This study has several limitations. Participants were drawn from HCXF-funded projects, and perspectives may not reflect experiences across the broader FIM field. Participation rates were moderate and varied by stakeholder group, which may have introduced nonresponse bias if those who participated differed systematically from those who did not. Only one food vendor participated in the study, which may limit the breadth and representativeness of perspectives captured from the vendor viewpoint. Although we used strategies to support analytic rigor, qualitative interpretation may be influenced by the research team’s perspectives. All facilitators had prior involvement with HCXF, which may have affected participant responses or discussion dynamics. Because participants were funded through HCXF, responses may have been influenced by social desirability bias, with participants potentially emphasizing favorable experiences or framing findings in ways they perceived as aligned with funder expectations. Findings are intended to provide in-depth, context specific insights rather than generalization. The study also did not include clinical partners; future work should examine how partnerships with clinical organizations shape FIM implementation and research processes.

Conclusion

Stakeholders’ accounts suggest that as FIM research expands, the central challenge is not interest or momentum, but how to generate comparable, decision-relevant evidence across diverse delivery models. Use of the EPIS framework helps identify actionable opportunities to strengthen future research, including greater cross-study coordination, clearer articulation of evidence needs and pathways to coverage, investment in partnership research infrastructure, and intervention specifications that balance evaluability with routine delivery conditions. These approaches may support an evidence base that is more consistent and decision relevant as the field moves toward larger studies of clinical outcomes and cost effectiveness.

Footnotes

Acknowledgements

The authors would like to thank the HCXF researchers, community-based organizations, and food vendor who generously shared their perspectives. We also appreciate the constructive comments from reviewers, which helped refine the framing and interpretation of our findings.

ORCID iD

Dara O’Neil

Ethical Considerations

The study was conducted in accordance with applicable ethical standards and was approved by the Heartland Institutional Review Board on May 20, 2025 (project no. 052025-1205).

Consent to Participate

Verbal informed consent was obtained from all participants prior to participation in the study. The consent process was audio-recorded.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: All authors work as part of the Health Care by Food™ initiative at the American Heart Association. The Health Care by Food™ Initiative is supported by anchor funding from the Rockefeller Foundation.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The qualitative data generated and analyzed during the current study are not publicly available due to the need to protect participant confidentiality.*

References

Snetselaar

de Jesus

DeSilva

Stoody

. Dietary Guidelines for Americans, 2020-2025: Understanding the Scientific Process, Guidelines, and Key Recommendations. Nutr Today. 2021;56(6):287-295. doi:10.1097/NT.0000000000000512.

Liu

Micha

Mozaffarian

. Trends in Food Sources and Diet Quality Among US Children and Adults, 2003-2018. JAMA Netw Open. 2021;4(4):e215262. doi:10.1001/jamanetworkopen.2021.5262. Published 2021 Apr 1.

GBD 2021 Stroke Risk Factor Collaborators . Global, regional, and national burden of stroke and its risk factors, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Neurol. 2024;23(10):973-1003. doi:10.1016/S1474-4422(24)00369-7.

Schwartz

Wohrman

Holubowich

Sanders

Volpp

. What Is 'Food Is Medicine,’ Really? Policy Considerations On The Road To Health Care Coverage. Health Aff (Millwood). 2025;44(4):406-412. doi:10.1377/hlthaff.2024.01343.

Seligman

Angell

Berkowitz

, et al. A Systematic Review of “Food Is Medicine” Randomized Controlled Trials for Noncommunicable Disease in the United States: A Scientific Statement From the American Heart Association. Circulation. 2025;152(4):e32-e46. doi:10.1161/CIR.0000000000001343.

Volpp

Berkowitz

Sharma

, et al. Food Is Medicine: A Presidential Advisory From the American Heart Association. Circulation. 2023;148(18):1417-1439. doi:10.1161/CIR.0000000000001182.

Hildebrand

Levi

Marpadga

Perez-Velazco

Seligman

. RE-AIM evaluation of the first 5 years of a citywide produce prescription program. Transl Behav Med. 2025;15(1):ibaf057. doi:10.1093/tbm/ibaf057.

Houghtaling

Short

Long

, et al. Barriers and facilitators to implementing Food is Medicine programs: Evidence from 21 food bank-healthcare partnerships. Transl Behav Med. 2025;15(1):ibaf013. doi:10.1093/tbm/ibaf013.

Kim

Thang

Imai

, et al. A Qualitative Study of Collaborative Food Programs: Insights from a FQHC-University Partnership During COVID-19. Nutrients. 2025;17(11):1856. doi:10.3390/nu17111856. Published 2025 May 29.

10.

Raber

Brozovich

, et al. Staff Perspectives and Organizational Climate Factors Related to Implementation of Food is Medicine Programming in a Federally Qualified Health Center. J Health Care Poor Underserved. 2024;35(3):763-776.

11.

Wetherill

Bridges

Talavera

Harvey

Skidmore

Burger

. Planting Seeds for Food Is Medicine: Pre-Implementation Planning Methods and Formative Evaluation Findings From a Multi-Clinic Initiative in the Midwest. J Prim Care Community Health. 2024;15:21501319241241465. doi:10.1177/21501319241241465.

12.

Butler

Kamdar

Brostow

, et al. Veterans Affairs FreshConnectProduceRx: a study protocol for a pragmatic quasi-experimental study assessing health, healthcare costs, and implementation processes of a produce prescription program in VA medical centers. BMC Public Health. 2025;25(1):2353. doi:10.1186/s12889-025-23355-2. Published 2025 Jul 3.

13.

Brostow

Kamdar

Napa

, et al. Incorporating a Produce Prescription Intervention Into Clinical Practice: Lessons Learned From a Food Is Medicine Intervention. Health Promot Pract. 2025;16:15248399251365947. doi:10.1177/15248399251365947. Published online September.

14.

Thomas

Gago

Egan

Sheward

Burt

. Evolving food as medicine programs to advance health equity: insights from two decades of practice. Curr Opin Pediatr. 2026;38(1):31-39. doi:10.1097/MOP.0000000000001531.

15.

Hashimoto

Primo

, et al. HealthyLink-Integrating Food Is Medicine (FIM) Into Inpatient Discharge Process. J Public Health Manag Pract. 2026;32(1):E12-E21. doi:10.1097/PHH.0000000000002241.

16.

Haddad

Miles

Alejandro-Rodriguez

, et al. Feasibility of self-investment in a medically tailored meals program by a large health enterprise: Cleveland Clinic experience. Nutr Health. 2025;31(4):1503-1513. doi:10.1177/02601060241307980.

17.

Thomas

Liang

Getzoff

Silver

Peters

Fuentes

. Enhancing cultural humility in food is medicine services: A mixed-methods investigation of patient food and dietary preferences. Healthc (Amst). 2024;12(3):100749. doi:10.1016/j.hjdsi.2024.100749.

18.

Calloway

Houghtaling

Mitchell

, et al. Participant Redemption and Engagement in Produce Prescription Programs: A Qualitative Analysis of Implementer Perspectives. Curr Dev Nutr. 2025;9(9):107530. doi:10.1016/j.cdnut.2025.107530. Published 2025 Aug 23.

19.

Adeleye

McDaniel

Iribe

, et al. Applications of Human-Centered Design to Food Is Medicine Interventions: The THRIVE Pilot Trial. J Am Heart Assoc. 2025;14(15):e041846. doi:10.1161/JAHA.125.041846.

20.

Palmer

Byker Shanks

Balis

Shaw

Lima Dos Santos

Yaroch

. Food is medicine programs for pregnant women in the United States: a systematic review. Transl Behav Med. 2025;15(1):ibaf060. doi:10.1093/tbm/ibaf060.

21.

Houghtaling

Short

Shanks

, et al. Implementation of Food is Medicine Programs in Healthcare Settings: A Narrative Review. J Gen Intern Med. 2024;39(14):2797-2805. doi:10.1007/s11606-024-08768-w.

22.

Mathur

Yeragi

Prabhu

, et al. Implementation and Effectiveness Outcomes of a Pilot Comparative Effectiveness Randomized Controlled Trial Evaluating a Food Is Medicine Program Among At-Risk Pediatric Populations. Child Obes. 2025. doi:10.1177/21532176251403284. Published online December 19.

23.

Balis

Yaroch

Palmer

Shaw

Lima Dos Santos

Byker Shanks

. Implementation and Impact of Perinatal Food Is Medicine Programs: A Qualitative Research Study. J Acad Nutr Diet. 2024;124(10):1255-1265.e21. doi:10.1016/j.jand.2024.02.007.

24.

Health Care by Food . Common measures. Health Care by Food. https://www.healthcarexfood.org/en/science-of-hcxf/common-measures. Accessed 26 February 2026.

25.

Ridberg

Yaroch

Nugent

Byker Shanks

Seligman

. A Case for Using Electronic Health Record Data in the Evaluation of Produce Prescription Programs. J Prim Care Community Health. 2022;13:21501319221101849. doi:10.1177/21501319221101849.

26.

Palar

Cox

. Modeling The Value Of 'Food Is Medicine': Challenges And Opportunities For Scaling Up Medically Tailored Meals. Health Aff (Millwood). 2025;44(4):443-448. doi:10.1377/hlthaff.2025.00161.

27.

Folta

Burch

Alcusky

, et al. Facilitators and barriers to reach and enrollment into a medically tailored meals program within a section 1115 Medicaid pilot: clinic staff perspectives. Front Public Health. 2025;13:1526564. doi:10.3389/fpubh.2025.1526564. Published 2025 Feb 5.

28.

Radtke

Xiao

Chen

, et al. Real-world implementation of a clinic-community Food is Medicine intervention for patients with type 2 diabetes. Diabetes Res Clin Pract. 2025;226:112376. doi:10.1016/j.diabres.2025.112376.

29.

Zimmer

Strahley

Shenberger

, et al. Fresh Food Rx: Evaluating the Impact of a Produce Prescription Program on Engagement and Well-being Using the RE-AIM Framework. J Gen Intern Med. 2025;41:997-1009. doi:10.1007/s11606-025-09687-0. Published online July 14.

30.

Vilme

Zhang

O'Tierney-Ginn

, et al. Gaining stakeholder perspectives to shape a produce prescription program to improve maternal and birth outcomes: a qualitative study. Front Public Health. 2025;12:1462908. doi:10.3389/fpubh.2024.1462908. Published 2025 Jan 15.