Co-Design Practices in Diet and Nutrition Research: An Integrative Review

Co-design, the method of involving users, stakeholders, and practitioners in the process of design, may assist to improve the translation of health evidence into tangible and acceptable intervention prototypes. The primary objective of this review was to identify and describe co-design techniques used in nutrition research. The secondary objective was to identify associations between co-design techniques and intervention effectiveness. An integrative review was performed using the databases Emcare, MEDLINE, PsycINFO and Google Scholar. Eligible studies included those that: (1) utilised participatory research or co-design techniques, (2) described development and/or evaluation of interventions aimed at improving dietary behaviours or nutrition, and (3) targeted community-dwelling adults aged ≥18 years. We identified 2587 studies in the initial search and included 22 eligible studies. There were 15 studies that utilised co-design techniques, with a strong focus on engagement of multiple stakeholder types and use of participatory research techniques. No study implemented a complete co-design process. Most studies (14/15) reporting outcomes reported positive health (maximum p < 0.001) or health behaviour outcomes attributed to the intervention; hence, associations between co-design techniques and effectiveness could not be determined. Currently published intervention studies have used participatory research approaches rather than co-design methods. Future research is required to explore the effectiveness of co-design nutrition interventions.


Introduction
Over the past half-century, dietary intakes have changed dramatically, with increased consumption of processed foods containing added sodium, unhealthy fats, and refined carbohydrates/sugars [1]. Men and women across all age groups consume high amounts of discretionary (unhealthy) food with underconsumption of fruits and vegetables relative to health guidelines. These dietary factors are major drivers for common chronic conditions including cancer, heart disease and Type 2 diabetes [2], which are leading contributors to early death, illness, and disability [3]. Improving dietary behaviour is a cornerstone in the prevention and treatment of chronic diseases [4], but remains a significant challenge [4]. To effectively achieve dietary behaviour change, interventions must be embedded in best practice, associated with effectiveness, and be relevant and appealing to target populations to facilitate successful translation into practice. Hence, nutrition interventions are increasingly focused on patient-centred models [4].
Person-or community-centred care is the foundation of dietetic practice [5]. This refers to healthcare providers building relationships with people and their communities to manage health conditions in a personalised approach that provides equal sharing of power [6]. In this manner, public health (nutrition) research interventions should consider a similar approach, including "bottom-up" participatory research or participatory action research (PAR) designs. The benefits of PAR are widely acknowledged and include the development of research outputs closely aligned to community needs, while helping to build community capacity and promoting research equity [7]. Notably, PAR defies traditional "top-down" research methods to disassemble traditional power imbalances between participants and researchers.
Co-design, also known as co-creation, co-production, or participatory design, in a healthcare setting, refers to the integration of design thinking, stakeholder experiences, scientific evidence and participatory principles in the collaborative design of local solutions to local problems [8][9][10]. Co-design is considered to produce solutions based on understanding of the local context to meet the needs of all stakeholders [11], offers insights into the lived experience of the public and helps to answer the why questions as opposed to sciencebased research, which predominantly looks at what is happening. Therefore, co-design may have greater acceptance by providers and target users [9], and offer a more sustainable and effective translation approach into clinical practice. Furthermore, controlled trials provide rigorous evidence of the inherent value of community inclusion in public health research processes, particularly for increasing the effectiveness of interventions, achieving local customisation and strong community engagement [12], and improving the quality and appropriateness of study design [13].
Co-design research methods can also be effective in helping to overcome barriers to translation and improving the uptake and effectiveness of nutrition interventions, as it is unclear to what extent co-design has been incorporated into nutrition research, nor what the benefits are. However, studies evaluating the effectiveness of co-design appear to be scarce [8,14], and the effectiveness of co-design methodological techniques used in different research disciplines remains unfixed. Co-design and PAR approaches have also been the subject of intense debate with key criticisms including poor reporting practices [15] and tokenism: "small-scale, poorly funded and with limited incentives" in co-design activities [16]. Furthermore, there remains limited published studies, systematically reviewing the participatory design of nutrition/diet-based interventions. Hence, the purpose of this study was to conduct an integrative, systematic review to identify and describe participatory and co-designed methodological techniques previously used in nutrition research and to identify any associations between the use of participatory or co-design techniques and intervention effectiveness. This will assist to guide future nutrition research that deploys co-design or participatory research methods.

Methods
Ethics approval was not obtained for this study since human participants were not involved. An integrative review using a systematic review search approach was undertaken following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A protocol for this review was written, agreed upon by all co-authors and registered with Open Science Framework (osf.io/s8cv7). All stages of literature searching and screening were conducted by the first author (B.S.J.T.) with assistance from the coauthors as specified.
An integrative review approach was used as it enables systematic and rigorous review of studies that contain diverse methodologies, for example, experimental, nonexperimental, quantitative, and qualitative work [17]. Integrative reviews share common search strategies for promoting rigor associated with systematic reviews, but diverge at the point of data analysis, with integrative reviews drawing upon inductive techniques such as the identification of noting patterns and themes, seeing plausibility, clustering, counting, making contrasts and comparisons, discerning common and unusual patterns, subsuming particulars into general, noting relations between variability, finding intervening factors, and building a logical chain of evidence [17].

Eligibility Criteria
Eligibility criteria are outlined below, according to the PICOS structure.

Population
Target population was restricted to community-dwelling adults who were 18 years old or older. Studies including children and adolescents were excluded due to vast differences in characteristics, learning and behavioural issues compared with adults. Studies conducted in any geographical locations including metropolitan, rural, and remote areas and online settings were included. Gender and health status of the study population was not limited, i.e., healthy populations or individuals with a specific health condition were included. Study populations excluded were ex vivo, in vivo, and in vitro studies and studies where the participants were not humans, or animal models.

Intervention
Eligible studies reported an intervention that aimed to improve dietary behaviours or any aspect of nutrition-for example, interventions that aim to increase vegetable consumption, examine the effects of a specific food, ingredient, or compound on a healthrelated outcome, or to encourage compliance with an entire diet. Mixed interventions (interventions targeting dietary and other health components risk factors such as physical activity) were also included. Interventions could be delivered via any format including digital, face-to-face, or mixed. To be included, co-design or participatory research methods must have been used to develop the intervention. The definition of co-design varies with different authors; however, the general rule is that the research methods should include active collaboration between participants, researchers and other relevant stakeholders in the process of intervention design [18]. To exclude interventions with limited participant involvement in their development, only studies classified as having involved "collegiate", "collaborative", or "consultative" participation were included. Classification was based on the seminal definitions of the four modes of participation described by Cornwall and Jewkes [19], built upon by Biggs [19,20]. Contractual participation is considered to reflect shallow participation, while consultative research approaches genuine participation, and collaborative and collegiate meet standards for genuine participation [19].

Control or Comparator
Studies were not limited based on control group or comparator.

Outcomes
Studies that described either the design, development or evaluation of a co-designed dietary behaviour intervention were eligible. Therefore, eligible outcomes include: Description of characteristics of an intervention: Studies that describe the participatory, co-design, or stakeholder engagement techniques that they applied in the development of a diet or nutrition intervention were included. Hence, papers describing study protocols were eligible.
Outcomes: Studies were included if they reported dietary behavioural outcomes (e.g., increased fruit and vegetable intake) and/or health outcomes (e.g., weight loss). Study outcomes could also be the development of a co-designed intervention, a new tool, or specific intervention components. Qualitative studies undertaken to directly inform the design of a specific intervention were included, while qualitative or consultative research for general knowledge purposes was ineligible (e.g., identifying barriers and facilitators to dietary behaviours).

Study Design
As this is an integrative review, any type of primary study (qualitative or quantitative) was eligible. Studies of any sample size, protocols for planned studies and studies that reported descriptions of the co-design process or methods without outcome evaluations were included. Studies that reported process evaluations of the intervention were also eligible. Other publication types such as review articles, opinions or editorials and conference abstracts of less than 1000 words were excluded.
To be included, study methodologies had to meet collegiate, collaborative, or consultative levels of participation [18]. Studies that met just the contractual level were excluded (see Table 1). Table 1. Definitions, explanations, and eligibility of different levels of participation as described by Cornwall and Jewkes [19] and Biggs [20].

Definition Further Explanation Eligible for Review
Collegiate (deepest form of participation) Researchers and local people work together as colleagues with different skills to offer, in a process of mutual learning where local people have control over the process.
Deepest level of participation. Researcher's role shifts from director to facilitator and catalyst.

Collaborative
Researchers and local people work together on projects designed, initiated, and managed by researchers.
Collegiate techniques are applied but are influenced by institutional agendas. Genuine participation occurs within the confines of a larger, pre-designed research process.

Consultative
People are asked for their opinions and consulted by researchers before interventions are made.
People are involved as informants for the purposes of verifying and amending research findings. People are contracted into the projects of researchers to take part in their enquiries or experiments.
People are involved to fulfil a data collection role and they have no control or input into projects that are scientist-led, designed, and managed. X

Data Sources and Search Strategy
The search strategy centred upon two concepts: (1) co-design and (2) dietary intervention, and was developed with input from an academic librarian at Flinders University (Adelaide, South Australia). After conducting experimental searches, the proximity searching technique was used for the "dietary intervention" concept to improve search precision and reduce the number of results produced. In Google Scholar where proximity searching is not applicable, "dietary intervention" was collapsed into two distinct concepts, i.e., "diet" and "intervention". Hence, the search strategy used in Google Scholar was the combination of synonyms of "co-design", "participatory action research", "diet" and "intervention".
The systematic search was conducted on 11 August 2020 using four electronic databases (Emcare, MEDLINE, PsycINFO and Google Scholar). Consistent with evidence regarding optimal coverage for health and medical topics, Emcare, MEDLINE and PsycINFO databases were searched with the following filters applied: (1) humans, (2) full text available, (3) published date 2010-2020 and (4) English language [21]. Search terms included synonyms of the search concepts "co-design" and "dietary intervention" (see Table 2) and were searched in all fields. screening was undertaken in duplicate, independently, by the same individuals and any disagreements were resolved through discussion with an independent adjudicator. Table 2. Search concepts and synonyms included in searches.

Concept 1: Co-Design
Concept 2: Dietary Intervention co-design* OR codesign* OR co-creat* OR cocreat* OR "participatory design" OR "design research" OR "collective creativity" OR "user-centred design" OR design* OR "consumer participation" OR pre-design* OR participatory OR "participatory action research" OR "action research" OR "community-based participatory research" OR "co-production" OR "user-centred" OR "human-centred" OR "human-centred design" OR "design thinking" OR "experience based design" OR "experience-based design" OR "experience based co-design" OR "experience-based co-design" OR "experience based codesign" OR "experience-based codesign" diet* OR nutrition* OR eat OR eating OR food* OR meal* OR "meal plan*" OR menu* adj1 intervention* OR activit* OR strateg* OR program* OR service* OR plan* OR advice OR regime* OR therap* OR provision AND → → The arrow is assumed to be understood as an indicator that synonyms under concept 1 AND synonyms under concept 2 were searched.

Classification of Studies Based on Modes of Participation
Due to variation in participation across different studies further screening was conducted to classify the extent to which participatory techniques were utilised in each study according to the four modes of participation described by Cornwall and Jewkes [19]. Since contractual research involves only minor and superficial consultation with participants, articles were only included in this review if the intervention design reached collegiate, collaborative, and consultative participatory standards. Two authors (B.S.J.T. and J.C.R.) classified the studies in independent duplicate and any conflicts were discussed and resolved.

Data Extraction and Management
Data were extracted into a purpose-developed data extraction table. Information regarding the studies' characteristics (aim, participants, inclusion of other stakeholders, setting, intervention, main outcome or finding, PAR standard reached) and co-design methods (theoretical framework, co-design approach, data collection/analysis techniques, research stage at which participant feedback was sought, and extent of engagement) were included.

Sufficiency of Reporting
An assessment of sufficiency of reporting was undertaken using an adapted version of the eight-item checklist for reporting non-pharmacological interventions, originally adapted from the Consolidated Standards of Reporting Trials (CONSORT) checklist [14,22,23]. Studies were scored against items relating to the (1) setting, (2) stakeholders, (3) facilitators, (4) co-design methods, (5) materials, (6) length of design and sessions, (7) interval and frequency of sessions and (8) description of the overall co-design process.

Results
Identification and selection of studies is summarised in Figure 1. After full-text screening, 36 studies were eligible. Following further screening to exclude contractual modes of participation, 22 studies were included in this review.

Results
Identification and selection of studies is summarised in Figure 1. After full-text screening, 36 studies were eligible. Following further screening to exclude contractual modes of participation, 22 studies were included in this review.

Characteristics of Included Studies
Study characteristics are depicted in Table 3.

Theoretical Frameworks
Twelve studies reported using a theory-based framework in developing the intervention. Social Cognitive Theory [52] was the most common theory used (n = 3), followed by PAR (n = 2) and Intervention Mapping (n = 2).

Recruitment Methods
Recruitment methods varied. Three studies [36,44,48] did not report methods of participant recruitment and only two studies [33,47] reported providing reimbursement to participants for their contribution.

Extent of Participation
The extent to which participants had input was classified according to the six intervention development phases identified by Eyles et al. [14]. All, except two studies, assessed user needs to inform intervention focus. Six studies had end-user input in pilot/real-world testing, whereas four studies included end-users in prototype testing. Four studies assessed background knowledge and evidence, two studies assessed user needs to inform technology, and two studies involved participants in developing intervention content.

Intervention Effectiveness
Fourteen out of 22 studies reported outcomes, with 13 studies reporting statistically significant changes in diet-or nutrition-related outcomes or behaviour attributed to the intervention [29,30,33,35,40,41,43,[45][46][47][48][49][50]. No studies empirically evaluated the effect of participant engagement on the results of the study or effectiveness of the intervention. Since all studies that reported outcomes reported positive outcomes, the relationship between stage of end-user consultation and intervention effectiveness was explored. The highest percentage (75%) of studies that showed positive outcomes were studies that involved endusers in prototype testing. This was followed by studies that assessed user needs to inform intervention focus (67%) and those that involved end-users in pilot testing (67%). None of the studies that assessed user needs to inform the technology used for the intervention found positive outcomes. Six (55%) out of 11 studies which involved multiple phases reported positive outcomes [29,32,33,35,44,45].

Discussion
This integrative review identified 22 original research studies or protocols for nutrition/diet intervention studies that featured participant engagement in the design or development and were published in the last decade. Within the participatory design methods and processes used, there was no evidence of the explicit use of co-design; however, some studies utilised co-design techniques and 11 studies engaged participants to a collegiate or collaborative level, indicating genuine partnership and meeting PAR requirements. No studies empirically evaluated the specific impact of participant engagement on intervention effectiveness.
To our knowledge, just one published study has reviewed the use of co-design practices within digital health research [14], while the current study is the first to review co-design practices within nutrition research specifically. Similar to Eyles et al., the participatory techniques reviewed in the current study were varied, ranging from conventional methods including focus groups and surveys, to less conventional methods such as photovoice. Different methods were used specific to various research contexts. For example, the photo-voice method, a visual technique of capturing participants' concerns which may be sensitive, is pertinent to the needs of the Indigenous population [53]. The interventions included in the current review varied and substantially focused on community-based programs seeking to provide tools and resources to help people, families, or workplaces to adopt healthier dietary behaviours.
The research designs used were also varied. Qualitative research was common for studies that were at the earlier stages of design (i.e., to inform intervention development), while pre-post and randomised controlled trials were used to evaluate co-designed in-terventions. Finally, the research populations and samples included a range of end-user stakeholders from specific communities (e.g., ethnic or cultural groups), typically consisting of intervention end-users, although it was not uncommon to include other stakeholders such as health practitioners or other professionals. Promisingly, a substantial number of studies reached a sufficient level of participation whereby power over decision making is shared, suggesting genuine inclusion in the research process [54]. Overall, the body of research demonstrates a heterogenous application of participatory and co-design research techniques that were adapted to the unique needs and characteristics of the health problem or population at hand. Hence, future participatory research could adopt methods suited to similar contexts and evaluate their suitability where necessary.

Methodological Considerations of the Included Research
A strength of the included studies was that details of the participatory design methods and stakeholders involved in intervention development were reported sufficiently in many key areas. However, it was challenging to determine timeframes of the intervention development process including the total number and time interval between sessions. Although materials used in the design process were named, most were not adequately described. Insufficiently detailed reporting of methodological considerations is a limitation of PAR research previously identified [14,55,56], highlighting the need for more detailed methodological reporting in this field. Notably, studies which involved collegiate participation (the highest form) did not necessarily translate into sufficient reporting. Similarly, consultative participation (the lowest) did not necessarily equate to insufficient reporting.
In this review, only four studies involved a minority population group (African Americans; Bangladeshi migrants) and only two studies involved Indigenous population groups. These findings are contrary to the general acceptance that participatory design is common and best practice in research involving under-served and/or Indigenous populations [57]; however, absence of reporting in the peer-reviewed literature could be a factor and future reviews should include grey literature. Encouragingly, variation in population groups in the current review suggests participatory approaches are applied broadly across populations.

The Effectiveness of Co-Design in Nutrition Research
A secondary objective of this study was to understand the effectiveness of co-design in nutrition/diet-based interventions. However, no studies identified empirically evaluated the effect of participant engagement on these outcomes. Of the 14 studies that assessed intervention outcomes, all but one reported a statistically significant effect in the desired direction, which may have been a result of publication bias towards successful studies [58]. Nonetheless, this review found that a higher percentage of studies reported positive health or health behaviour outcomes if they involved end-users in prototype testing, pilot testing or assessed their needs to inform the intervention's focus. For example, Adams et al. utilised photo-voice techniques to guide their intervention design and made sure that it was appropriate for the social contexts and met the cultural and practical needs of local Australian Aboriginal people [24]. In future, research may benefit from greater inclusion of end-users at early stages of research design to preliminarily identify the optimal direction.
There is a dearth of evidence assessing the association between different modes of participation and nutrition intervention effectiveness. Future studies should include controlled trials of treatments that vary levels of participation and co-design or no participation. Future studies should improve reporting, including deficiencies identified in the current review (see below) and to facilitate future best practices. Additional reporting should seek to cost participation in design and estimate return on investment through long-term follow-up. To date, to our knowledge, no study has examined whether co-design is more effective than traditional approaches to intervention development with evaluations tending to be descriptive in nature and not experimental [8,14]. This highlights the need for robust, empirical evaluations that can evaluate these effects. While a randomised controlled trial approach would help to establish the efficacy of a co-designed intervention, it is likely that a RE-AIM approach that considers translational outcomes as well as efficacy outcomes [59] is more appropriate for assessing the effect of co-design.
A lack of robust evaluations of the impact of co-design have been noted in related fields. Similar to our findings, a recent rapid review found 11 studies reported on the use of co-design in research within acute healthcare settings and that while many studies provided qualitative and descriptive data regarding the perceived value of co-design, robust evaluations were limited [60].

Strengths and Weaknesses of the Review
A strength of this review was that systematic approaches were used across three different scientific databases with studies independently reviewed in duplicate by two co-authors. The integrative review approach also enabled inclusion of both quantitative and qualitative research including studies at different stages of the research process (e.g., protocols, intervention development papers). The addition of Google Scholar to the search also identified a further study that was included. Despite this, by limiting the search to the last 10 years and to articles published in English only, it remains possible that eligible studies were missed. For example, our search strategy did not capture work reflecting a Kaupapa Māori approach to co-design of health or diet interventions [61], potentially because these approaches and cultural groups conceptualise dietary behaviours within a wholistic, whole-of-health model that our search strategy did not detect. Therefore, future reviews should consider cultural and local differences in language and conceptualisation of health to ensure coverage of different groups. The exclusion of grey literature where co-design work may be more commonly published is another limitation of this review. Additionally, although intervention effectiveness was examined, risk of bias could not be assessed due to variation in outcome measures reported in the included studies. However, reporting practices were analysed, which is a valuable outcome of this review.

Implications for Future Research
This review has several implications for future research. Reporting practices around participatory research have previously been reported to be poor, highlighting the need for researchers to use standard checklists for reporting interventions designed using participatory or co-design methods. Eyles et al. highlighted that checklists can be adapted from existing relevant and appropriate checklists [14], such as those described by Hoffman et al. [62] or Borek et al. [63]. Sufficiency of reporting can provide clearer guidance for future studies to employ methods that are replicable and consistent. Furthermore, it is important to note that co-design techniques and tools are often adapted to the specific research questions, contexts, and populations at hand. Future research would benefit from more open and detailed descriptions of these adaptation processes and the rationales that underpin co-design decision making. To take this even further, it would be ultimately beneficial if researchers begin to openly publish their co-design techniques, similar to how datasets and survey instruments (for example) are increasingly published in opensource libraries, where they can be accessed by researchers and amended for other research purposes.
Other than that, it is suggested that researchers give a higher level of consideration to the time and resources required to design interventions within participatory research. It is important to think about the range of multilevel stakeholders' representatives that researchers plan to invite to a co-design activity and consider carefully what their drivers and motivations to participate might be. Co-design does not have to be undertaken independently from other research methods; in fact, it works well with other quantitative methods as part of a mixed methods model. Lastly, to determine whether co-design is more effective than traditional approaches to intervention development, high-quality process evaluations and randomised controlled trials should be conducted to assess intervention effectiveness compared to non-co-designed comparator interventions or waitlist control groups.

Conclusions
Reviews summarising the methods and processes used in participatory and co-design of dietary interventions remain limited. The 22 studies included in this review used participatory research, but not co-design, methods. More studies reported positive health or health behaviour outcomes if they involved participants in prototype testing, pilot testing or needs assessment to inform the intervention focus. Most of the studies did not achieve an adequate level of reporting for their intervention development processes. Further research to explore co-designed nutrition/diet interventions and their effectiveness is warranted.