Research has shown that consuming fruits and vegetables is linked to a lower risk of all-cause mortality (Wang et al., 2014). However, the recommended daily intake of 400 g of fruit and vegetables is not met by many people (World Health Organization, 2019). The food environment during childhood plays a role in shaping eating preferences (Birch & Fisher, 1998). In Europe, a significant percentage of children do not eat enough fruit and vegetables on a daily basis (Williams et al., 2020). Schools have been identified as an important setting to influence children’s health and eating habits, leading to various nutrition interventions (Jourdan et al., 2021; Sawyer et al., 2021). These interventions have had mixed success in improving children’s eating practices (O’Brien et al., 2021; van Cauwenberghe et al., 2010; Verdonschot et al., 2023). Review findings show school-based nutrition interventions have a positive effect on fruit and vegetable consumption in children aged 6–18, although the quality of the studies supporting this finding is questioned (O’Brien et al., 2021). Multicomponent interventions have a strong effect on fruit and vegetable intake in children aged 6–12 but have limited effects on adolescents aged 13–18. (van Cauwenberghe et al., 2010). Two recent systematic reviews specifically looked at providing fruit and vegetables to children in schools (Micha et al., 2018; Verdonschot et al., 2023). Verdonschot et al., found that fruit and vegetables provision has the greatest effect on fruit intake (Verdonschot et al., 2023). Micha et al., report an increase of 0.27 servings of fruit per day and 0.04 servings of vegetables per day as a result of policies, programs and interventions providing fruit and vegetables to children in schools.

The European Union (EU) School Fruit and Vegetables Scheme (‘the Scheme’) is a supranational school-based nutrition policy program in Europe. Launched in 2009, it is available to all EU member states and consists of three main components: distributing fruit and vegetables to children, providing educational measures, and conducting information campaigns, monitoring, and evaluation (European Commission, 2022a). Several studies have examined the impact of the Scheme in individual member states (Brunello et al., 2014; Haß et al., 2018; Methner et al., 2017; Roccaldo et al., 2017). In Italy, the Scheme was found to decrease the consumption of unhealthy food among children in affluent areas (Brunello et al., 2014) and improve adherence to the Mediterranean diet through nutrition education training for teachers (Roccaldo et al., 2017). In Germany, the delivery of fruit and vegetables to schools two to three times per week led to a significant increase in consumption on delivery days compared to other school days (Haß et al., 2018). In the same region one year after the introduction of the Scheme, there was a significant increase in children’s fruit and vegetables consumption without a decrease in consumption at home (Methner et al., 2017).

Although initial results on the impact of the Scheme may be encouraging, looking at impact is only one part of the story, with implementation offering further clarity as research indicates that the implementation process directly affects the outcomes of a program (Durlak & DuPre, 2008). There are systematic reviews with an international scope on the implementation of nutrition-related policies (McIsaac et al., 2019) and direct provision interventions of fruit and vegetables (Meshkovska et al., 2022). These reviews emphasize the significance of macro-level support, financial considerations, alignment with school priorities, shared purpose and responsibility among stakeholders, and characteristics of schools and communities (McIsaac et al., 2019). Additionally, the involvement of parents in supporting implementation and as targets of the intervention, along with positive perceptions of the intervention by children, are considered important factors (Meshkovska et al., 2022). However, there has been limited research on the implementation of the Scheme specifically, with only one descriptive study conducted in Spain (Soares et al., 2019).

An additional reason which makes the Scheme a relevant implementation case to study, is its origin from the Common Agricultural Policy (CAP) and thus, its characterization as a nutrition policy-based program, rather than an intervention (Commission of the European Communities, 2008; Elles et al., 2012; European Commission, 2022a; European Parliament, 2013a, 2013b). Implementation science and policy implementation are related fields of study (Nilsen & Cairney, 2020; Nilsen et al., 2013). Whereas implementation research is defined as ‘the scientific study of methods to promote the systematic uptake of research findings and other evidence-based practices into routine practice, and, hence, to improve the quality and effectiveness of health services and care’ (Eccles & Mittman, 2006); policy implementation research focuses on how government bodies put policies into practice (Nilsen et al., 2013). Most recent work in policy research has made some attempt to situate policy implementation, within the wider framework of policy-making (Howlett, 2019). However, policy program implementation remains a black box (Howlett, 2019). In the context of this study, we define public policies as ‘decisions, plans and actions that are enforced by [supranational] national or regional governments which may directly or indirectly achieve specific health goals within a society’ (Lakerveld et al., 2020). In continuation, nutrition policy programs are understood as specific opportunities for enhancing nutrition-related behavior, attitude and/or knowledge in a target population, for a specific time period (Woods et al., 2021).

The goal of our work is to understand how to improve implementation of school-based fruit and vegetables provision programs, as well as to better understand the policy implementation. From the perspective of the country-level government bodies in charge of the Scheme and across EU member states, we explore the barriers and facilitators influencing Scheme implementation. We use the Consolidated Framework for Implementation Research (CFIR) to guide our analysis of national-level implementation factors. (Damschroder, 2020; Damschroder et al., 2009; Nilsen, 2015).

The Scheme was introduced in the school year 2009/2010, at the initiative of the European Commission and as part of its Common Agricultural Policy (Commission of the European Communities, 2008; Elles et al., 2012; European Commission, 2022a; European Parliament, 2013a, 2013b). It is based on EU Regulation 1308/2013 (European Parliament, 2013b), Articles 22 to 25 and Annex V; and EU Regulation 1307/2013 (European Parliament, 2013a), Article 5 and Annex I; and is accompanied by two implementing regulations 2017/39 (European Parliament, 2017a) and 2017/40 (European Parliament, 2017b). Regulations are legal acts that apply automatically and uniformly to all EU countries and do not need transposition into national law.

The aim of the Scheme is to address low consumption levels of fruit and vegetables and resulting obesity by targeting children (European Commission, 2022a). EU regulations state the Scheme is intended to improve distribution of agricultural products and eating habits of children (European Parliament, 2013b). Member states are also asked by the EU to formulate country-specific needs and objectives (European Parliament, 2016). Predominating ones are: increase fruit and vegetables consumption, create healthy eating habits for children, teach children about nutrition and agricultural production and finally, help suppliers and producers through creating new sales markets (European Commission, 2022b). Monitoring reports are required for each school year, while evaluation reports should cover five school years (European Commission, 2022a; European Parliament, 2017a).

In order to implement the Scheme, EU member states are granted an overall budget of up to 150 million Euros per year (European Commission, 2022a). This total of 150 million is distributed among countries based on the number of children and level of regional economic development (European Commission, 2022a). Countries are required to draw up a six year country strategy and are given flexibility in regard to both (1) design of the Scheme (frequency of distribution of the fruit and vegetables, duration, type of accompanying educational measures) and (2) implementation (the beneficiaries of the Scheme at local level can be municipalities, suppliers, schools) (European Commission, 2022a, b). The primary responsibility for design and implementation of the Scheme at country-level belongs to the respective ministries of agriculture, often in cooperation and coordination with the ministries of health and education (European Commission, 2022b). In its current form, the Scheme is a unique case of a nutrition-related, policy-derived program, introduced by a supranational body, adopted and implemented by the national-level, and implemented further down to the regional, municipal and school level in 26 of the 27 EU member states (Sweden has not adopted the Scheme).

Methods

This work is a qualitative, exploratory study into the determinants of implementation of the Scheme, using the CFIR to guide data collection and analysis. The CFIR (Damschroder et al., 2009; CFIR Guide, 2022) is a descriptive framework that lists 26 key determinants (constructs) of implementation grouped into five broad domains: (1) intervention characteristics (e.g., design quality and packaging); (2) the outer setting (e.g., external policy and incentives); (3) the inner setting (e.g., implementation climate); (4) characteristics of individuals (e.g., knowledge and beliefs); (5) characteristics of the implementation process (e.g., planning). The CFIR remains one of the most comprehensive and widely used determinant frameworks in implementation science offering a cross-disciplinary comparison of findings (Birken et al., 2017; Kirk et al., 2016).

The CFIR was used to develop the semi-structured interview protocols (CFIR Guide, 2022). The process of formulating the research questions linked to each domain/construct/sub-construct, into open-ended interview questions is available from the first author, upon request. As there is most often one main person per government of an EU country participating in the Scheme, who is responsible for the implementation at country-level, we conducted semi-structured, open-ended interviews rather than focus groups. The interview protocol used with the EU official followed the one used with country officials, with slight adaptations to fit the supranational level. For example, question about the decision to adopt the Scheme at country-level, was reformulated to ask about the introduction of the Scheme as a policy program at EU level. After the first several interviews we made slight modifications (such as change of order of questions, and a re-formulation of questions to improve clarity). The final, full list of the questions asked is available as Online Resource 1: Interview protocol.

Participants and Recruitment

Recruitment and selection of respondents was purposive, with the use of key informants, to recruit additional persons to interview (Bryman, 2016). We began by establishing contact with a key official at the European Union Directorate General for Agriculture and Rural Development who is involved in the program. As personal contact information for the key individuals at the country-level government institutions is not publicly available, the EU contact assisted the recruitment. The EU contact sent out an email to key persons responsible for the Scheme at country-level with the lead author in copy. The email contained an invitation to participate in the study and informed about the origin and aim of the study. Those interested were asked to contact the first author and arrange for an online interview. In cases where the team of authors had professional contact points in particular countries (most often from academia, working on the topic of nutrition), these were used to reach out to the country-level persons from the ministries of agriculture responsible for the Scheme. The latter was in addition to, and as a follow-up to the recruitment process through the EU contact. Overall purpose was to ensure as many countries would participate in the study as possible.

Those primarily responsible for the implementation of the Scheme were most often the first interviewed. These were from the ministries of agriculture, or bodies working under the umbrella of this ministry, tasked to administer the Scheme. Following principles of snowball sampling (Bryman, 2016; Green, 2018), the first contact point from each country was then asked by the first author (BM), to recommend other persons for further interviews. These were most often from partner institutions such as ministries of health and education. Otherwise, we interviewed persons involved in the Scheme that work at institutes, agencies or directorates under the umbrella of the aforementioned ministries. Country 8 was the only case where the interviewee was from a university, as the university was hired by the agriculture ministry to design and implement the educational measures. Additionally, from Country 10 we only interviewed a person from the institute of health working under the health ministry, as we could not reach the relevant person from ministry of agriculture.

Ultimately a total of 10 countries with 28 representatives took part in this study. They came from the ministries of agriculture, health and education (or subordinated authorities or institutions officially mandated to implement the Scheme). In addition, we also interviewed the EU contact person. We refrain from writing the full name of the ministry/agency/mandated institution but rather use a more generic reference ministry/agency/institute. The purpose is to preserve anonymity. For the same, we do not report on the names of the 10 countries that have taken part in the study. When labeling the interviews, we use the following system: ‘C-number’ to reflect the country, followed by ‘I-number’ to reflect the number of the interview from the particular country, followed by a generic reference to the institution, for instance: ‘HealthMinistry’—example: C1I4HealthMinistry, to signify the fourth interview, from Country 1, ministry of health.

Data Collection

We held interviews online from June 2020 to June 2021, most often on Zoom, lasting 32 min to 2 h. Although most interviews were one on one, there were four with more than one representative from the relevant body. Thus, we interviewed a total of 29 persons during 23 interviews. We conducted four interviews with two co-authors present to ensure that all questions and topics were covered and as a training for those less experienced in interviewing. We gave all respondents information about the study via email and asked them to sign consent forms. All respondents gave consent to record the interviews. The first author conducted the interview with the EU representative and with representatives from eight out of 10 countries. Most (20 out of 23) interviews were conducted in English. Due to language barriers, co-authors conducted the interviews of representatives from two countries in the national language. We translated interview questions into the required language and where necessary adjusted phrasing and terminology to the local language. Five co-authors were involved in conducting the interviews, two of whom had previous experience with qualitative methods. All interviewers received a training session by the first author on conducting interviews and qualitative data collection and with the purpose of standardization. The first author is a qualitative researcher with experience in interviewing vulnerable groups and experts. Interviewers engaged in ongoing discussions to address any questions related to the interviewing procedures.

Ethics

The study methods and procedures, as well as data handling linked to the project have been reviewed and approved by the Norwegian Center for Research Data (reference: 288181). Funding for the study has been provided by the Research Council of Norway (grant number: 297894/H10). The funding source had no role in the design of this study, its execution, analyses, and interpretation of the data or decision to submit results.

Data Analysis

All interviews were transcribed verbatim, after which they were imported to NVivo 12 Pro for analysis. Co-authors who are native speakers translated three interview transcripts into English. Using the CFIR we explored implementation across 10 EU countries with the country-level representing CFIR's inner setting domain. To ensure that no barriers and facilitators are missed and the CFIR is appropriate to use at this level of analysis, we decided to code the data inductively and thereafter, use the framework to map the inductive themes to the CFIR constructs. For this purpose, we followed principles of inductive thematic analysis and framework analysis at different stages of the process. The inductive thematic analysis was based on Braun and Clarke (1) dataset familiarization, (2) data coding, (3) initial theme generation, (4) theme development and review and (5) theme refining, defining and naming (Braun & Clarke, 2006, 2019, 2022). The framework analysis was based on (Gale et al., 2013), steps five and six in particular (1) transcription, (2) familiarization with the interview, (3) coding, (4) developing a working analytical framework, (5) applying analytical framework and (6) charting data into the framework matrix. Our team of authors bring together expertise in nutrition, epidemiology, implementation science, political science, psychology and health-related qualitative research. In addition, authors are of diverse ethnic backgrounds that contributed toward a better understanding and interpretation of the findings from various EU countries. We drew on the particular skills of each author throughout the analysis process. Each step and decisions along the way were discussed in regular meetings between the first author (BM) and principal investigator (NL) and scheduled meetings with the co-authors (JW, DAS., SF). The coding, thematic analysis and mapping of themes onto the CFIR was done by the first author (BM). For detailed steps of the analysis process, please refer to Online Resource 3: Data analysis.

The results of the analysis were discussed during a 2 hours co-author workshop (n = 7 researchers). During the workshop, two main goals were accomplished. Firstly, the co-authors agreed on final categorization of text and assigned codes/themes under specific domain/construct/sub-construct of the CFIR. For example, content related to perceived goals of the Scheme by respondents was shifted from ‘goals and feedback’ (inner setting) to ‘knowledge and beliefs about the intervention’ (characteristics of individuals). Secondly, authors agreed on exemplary constructs from each domain for the results section of the manuscript.

Throughout the process of data collection and analysis, the research team members (BM, SF, DAS, JW, NL) discussed and reached consensus that the country-level, reflected through the perceptions of ministries of agriculture, health and education will represent the ‘inner setting’ domain of the CFIR. The inner setting characterized through the ‘structural, cultural and political environments’ (Damschroder et al., 2009; Pettigrew et al., 2001), is the primary setting in which implementation occurs. Accordingly, data linked to the design and content of the Scheme as a program is considered to be part of the ‘intervention characteristics domain’. Data exploring relationship with the EU is considered to be part of the ‘outer setting’ domain. Data linked to knowledge and beliefs of the respondents is part of the ‘characteristics of individuals’ domain. Finally, data relating to the administrative process of implementation, from the supranational level, down to the school setting is part of the ‘process’ domain of the CFIR. This research was completed in accordance with the COREQ checklist (Tong et al., 2007) (See Online Resource 4: COREQ checklist).

Results

We conducted 23 semi-structured interviews with 29 respondents from 10 EU countries and the EU. Interviews were with representatives of ministries of agriculture or related (13), ministries of health or related (5), ministries of education or related (4), and with the contact for the Scheme at EU level (1). The countries taking part in the study were from various EU regions, including the northern (1), western (4), central (1), southern (3) and southeastern (1). For the full country descriptions, see Online Resource 2: Sample description.

As this is an exploratory qualitative study, our goal is to give an overview of the constructs/sub-constructs of the CFIR we found relevant, and give an overview of the range of different barriers and facilitators to implementation. The goal is not to evaluate the strength of each. Results are based on perceptions of respondents. Figure 1 highlights the main findings, where of the 26 constructs, we identified content in regard to 19. The results section is organized by the CFIR domains, where under each section we give an overview of an exemplary construct. We chose exemplary constructs based on a determination by the co-authors that content pertaining to those constructs is relevant to other CFIR constructs as well. For example, external policy and incentives (exemplary under outer setting) has to do with the financial cost of the Scheme which is covered by the EU. As such, the content from this construct is linked to the construct cost (Scheme characteristics domain) as well as readiness for implementation (inner setting domain) which both touch on the financial resources necessary for implementation. A summary overview of each barrier and facilitator reported is available as Online Resource 5: The CFIR content overview. Full content for every constructs/sub-construct identified is available as Online Resource 6: Full results barriers facilitators.

Fig. 1
figure 1

Overview of Main Findings of Barriers and Facilitators to Implementation. Columns represent the 5 domains of CFIR. Blue-gray boxes indicate constructs in regard to which content was reported. White boxes indicate that no content was reported in regard to the particular constructs. (structure adapted from (CFIR Guide, 2022; Damschroder et al., 2009))

Scheme Characteristics Domain

Within the ‘scheme [intervention] characteristics’ domain we identified the following CFIR determinants: Scheme source (both barrier and facilitator), evidence strength and quality (present with unclear influence), relative advantage (facilitator), trialability (facilitator), adaptability (both barrier and facilitator), design quality and packaging (barrier, facilitator and with unclear influence) and finally, cost (both barrier and facilitator). In the following section we highlight adaptability as an example construct from this domain.

According to the EU respondent, one key facilitator of the Scheme is its flexible design and implementation approach, allowing countries to adapt the program to their specific contexts. Countries have the autonomy to determine who the recipients of the aid or beneficiaries of the Scheme are, which can include schools, municipalities, or fruit and vegetable suppliers. Additionally, countries have the authority to decide the duration of the program, the frequency of delivery, portion sizes, types of fruits and vegetables offered, accompanying measures, and the age range of the targeted children. Some countries, like Country 3, particularly highlighted the flexibility given to schools in terms of selecting suppliers and implementing accompanying measures.

The flexibility of the Scheme has allowed countries to make adjustments based on ongoing evaluations and feedback. These changes mainly focused on the type of produce, aiming to increase the variety of fruits and vegetables offered. When evaluations suggested targeting vulnerable children based on their socio-economic background, countries made the Scheme free. Additionally, countries experimented with different administrative models for implementation, particularly regarding the selection of suppliers and reimbursement calculations. For example, some countries transitioned from complex calculations based on the quantity of produce to a simpler approach based on the number of enrolled children benefiting from the Scheme per school.

Although flexibility in some areas is a strength, lack of flexibility in other aspects was a barrier. For instance, respondents from Countries 1, 3, 5 and 6 pointed to the inability to provide the fruit and vegetables free to the teachers, as role models for the children. Respondent from Country 7 noted that the limitation on funding provided for the accompanying measures is a barrier:

We do think, you know, other countries are spending the minimum amount on accompanying measures and are doing so just to tick a box really and, you know, just deliver a few lesson plans in schools and I think we’d like to get more recognition for that because we do strongly believe that those strong accompany measures are key to the success of the Scheme and, you know, we would like, you know, maybe if there is a bit more flexibility at EU level where they could say, you know, you can claim up to a bit more depending on the country (C7I1AgricultureDepartment)

It should be stressed that the inability to adapt the Scheme on the raised points (free provision for teachers, more funding for accompanying measures), is a barrier to implementation due to the regulatory framework at the EU level.

Outer Setting

Within the ‘outer setting’ domain, we identified the following CFIR determinants: school needs and resources (both barrier and facilitator), cosmopolitanism (barrier), peer pressure (facilitator) and external policy and incentives (both barrier and facilitator). In the following section we highlight external policy and incentives as an examplary construct from this domain.

The construct of external policy and incentives has to do with perceptions of the sustainability of the Scheme. When asked about the future of the Scheme, respondents said that as long as EU funding is available, implementation of the Scheme continues. In addition, in many cases, it is EU level changes that precipitate changes at country-level. Consequently, in some countries, such reliance on the EU level is a barrier. Unless changes are asked from the EU, improvements will not come at initiative of the national-level: “Now that we have implemented it [the Scheme], it kind of goes by itself, if it works very well or not very well [regardless of how well it works] but it works anyway [it continues anyway]” (C1I2AgricultureMinistry).

Continuing with external policy and incentives acting as a barrier, a respondent from the health ministry in Country 2 reported discussing the possibility of making improvements to the Scheme with the counterpart from agriculture. As both see the same challenges, they agree on the needed changes. However, the respondent from health stated in regard to the colleague from agriculture “why should he do extra work, it comes from Brussels and its regulated, you have the program this is how you do it, and then you do it”(C2I2HealthMinistry).

Inner Setting

Within the ‘inner setting’ domain, we identified the following CFIR determinants: structural characteristics (both barrier and facilitator), networks and communications (both barrier and facilitator), implementation climate (barrier, facilitator and with unclear influence) and readiness for implementation (both barrier and facilitator). In the following section we highlight networks and communications as an example construct from this domain.

A potential facilitator related to the construct networks and communications was the relationship between the three ministries primarily involved in the Scheme at country-level—agriculture, health and education. According to several respondents, the EU requires that the ministries work together in the design and coordination of the Scheme. Respondents from three different countries (Countries 2, 3, 6) stated that the current relationship between agriculture and health is due to the Scheme. Further, this is a relationship that should be utilized in other programs: “I think also the cooperation between the parties agriculture, education, health only that, is already worth now to have this program.”(C3I2HealthRegional). Initial doubts subsided about the cooperation between health and agriculture in Country 2, as the primary persons in each institution began and continued to work together:

[I] didn’t have the feeling with these Agri people, I will work nicely together, I had the feeling this will be fights fights fights [four years later] I have to admit [we work] much better than sometimes with my colleagues in my own ministry. It's really a trustful relationship now and I think without the school milk school fruit Scheme this would have never happened (C2I2HealthMinistry)

Nonetheless, we identified barriers to maintaining a balanced three-way relationship. Several countries (Countries 1, 10, 2, 8) reported minimal to no participation of their counterparts from education, regardless of the efforts by the other two to involve them. From Country 3, one respondent stated that whereas agriculture and health were ‘key players’, colleagues from education were important at the ‘political level’. A respondent from Country 5 reported some tensions between the partnering ministries in regard to the choice of products:

It’s not like we, at the ministry of agriculture, do not know anything about it, actually, we might even be more experienced than them [ministry of health] on the topic of fruit and vegetable, but you always need an opinion from the authorities in charge, and these multidisciplinary skills sometimes create problems. (C5I1AgricultureMinistry)

Finally, a respondent from Country 6 identified the disagreement between agriculture on the one hand, and health and education on the other, as to who the beneficiaries of the Scheme should be, one of the primary barriers to implementation of the Scheme in the country. In this regard, agriculture has been outvoted two to one, and municipalities or schools rather than suppliers are the main beneficiaries. This has created significant challenges for agriculture when it comes to implementation, as they lack the contacts with municipalities that they otherwise have with suppliers.

Characteristics of Individuals

Within the ‘characteristics of individuals’ domain, we identified only one CFIR determinant– knowledge and beliefs about the Scheme (both barrier and facilitator). Overall, respondent attitudes toward the Scheme were positive. However big or small, its continued implementation, despite limited reach was a success in itself.

Knowledge and beliefs construct links with constructs from other domains in the CFIR. As reported under implementation climate (tension for change) (see Online Resource 6), nearly all respondents saw a need for the Scheme. Respondents emphasized the significant role of the Scheme for children from lower socio-economic backgrounds. However, concerns were raised about its effectiveness and inclusivity for all children. These concerns stemmed from specific design aspects, including its restriction to primary school children, short duration within the school year, low frequency of sessions per week, and limited participation by schools. The respondents also highlighted a perceived lack of awareness about the Scheme among parents, teachers, and schools in general, which contributes to the low participation of schools. This issue was touched upon under design quality and packaging (information/communication measures) and engaging (innovation participants, key stakeholders) (see Online Resource 6). Further, respondents expressed concern that due to low engagement with parents as a group, some parents who are aware of the Scheme, stop giving fruit and vegetables to their children at home. This is because parents believe what children get in school is enough to satisfy the nutritional recommendations. Respondents also shared the perception that suppliers may not always see a benefit from the Scheme, and consider the profits made insufficient considering the administrative burden of the Scheme. This is under engaging (key stakeholders) (see Online Resource 6). Administrative burden was one of the reasons why schools and municipalities do not participate.

Across countries, there was agreement regarding the overall goals of the Scheme, which is a facilitator. The perceived overall goals of the Scheme were the following (not in order of importance): (1) increasing consumption of fruit and vegetables, and thereby supporting agricultural producers (2) educating children in regard to healthy eating (3) creating healthy eating habits that would persist in the long-term. Respondents from five of the 10 countries included (Country 2, 3, 4, 7, 10) emphasized the significance of the Scheme for achieving a healthy diet for children of lower socio-economic background. Some respondents expressed that the goals of the Scheme have evolved through time from primarily aiming to support agricultural producers, to now fighting child overweight and obesity and in one case addressing child malnutrition (Country 10). For others, it is a fortunate coincidence that the Scheme is agreeable to all: “We are so lucky that this goal [to raise consumption of fruit and vegetables] is in the same direction as the health one, and so to make pupils more willing to taste and eat fruits and vegetables.” (C5I2AgricultureResearchOrganization).

However, there was disagreement on the primary goal of the Scheme: is the Scheme a food provision program, or an educational one aiming toward the creation of healthier food habits. Such lack of consistency in views is a potential barrier and may have implications for the design of the Scheme, as well as implementation. For example, a food provision program may have a long duration and high frequency, while a program aiming toward habit formation can be short but intensive. In Country 5, the primary goal of the Scheme as identified by one respondent was educational. Thus, the same person expressed the view that what is needed is a shorter, more intense program for higher grades as well as for the younger groups, which would contribute toward the goal of habit formation.

Process

Within the ‘process’ domain, we identified the following CFIR determinants: engaging (barrier, facilitator and with unclear influence), executing (both barrier and facilitator) and reflecting and evaluating (both barrier and facilitator). In the following section we highlight executing as an example construct from this domain.

When exploring the execution of the Scheme the main barrier raised across several countries (Country 1, 2, 4, 6, 9) was the yearly variability of the country budget allocated as well as the timing of the budget allocation. These factors pose challenges when it comes to planning and execution from the perspective of suppliers, municipalities and schools. Timing of the budget allocation as a barrier from the EU level, was particularly pronounced in countries with a regional governing structure, such as Country 4, 6 and 9. In these countries, the process of budget distribution from the national-level to the respective regions implementing the Scheme is particularly time-consuming. This causes significant disruptions to the implementation of the Scheme at the school level.

For instance, in Country 4, for an exemplary school year 2020/2021, the national-level informs the EU of the budget requested in January 2020. The national-level is informed by the EU of the budget allocated in March 2020, and must inform the regions. The regions then have to further organize the disbursement of funding over the spring and summer period, to ensure delivery of the fruit and vegetables for a school year beginning in September 2020. In Country 6, which follows the same timeline in regard to disbursement of funds at national-level (country is informed about the granted budget by the EU in March of a given year), tendering procedures for the selection of suppliers must be set up during the summer period. Suppliers are selected by September, when the new school year already begins. Any administrative delays, which do happen, postpone the starting date of delivery of the fruit and vegetables, as well as the accompanying measures in the country. Thus, schools cannot plan in advance, and have to contend with a different starting date of the Scheme each year. Schools share a similar problem of not being able to plan for the year in Country 9, where suppliers are chosen during the summer period for the current school year. As one respondent from Country 4 noted, although the EU is very responsive to any inquiries concerning the Scheme, the timing described above, is not optimal:

The deadlines are unfavorable. It’s one of those things that regularly goes nowhere. It’s a lot of work with a lot of coordination, with a lot of queries, with a lot of hope, for little output, I would say. That's less helpful. (C4I1AgricultureMinistryNational)

In contrast, we identified a facilitator based on the administrative setup in two countries having to do with the computerization of the application process for schools. Namely, in both countries, there are online platforms accessed by all schools that wish to apply to the Scheme. In order to apply, a school only needs to log in and report the number of children that would be receiving the fruit and vegetables. The digital platform aims to make applications easier for the schools.

Discussion

In this research, we found content pertaining to barriers and facilitators to implementation of the Scheme for 19 of the 26 constructs of the CFIR (Fig. 1). Built-in flexibility of the Scheme is a facilitator of country-level implementation as is good cooperation between implementing ministries. Timing of the budget distribution is a barrier and taking EU funding for granted is a potential disincentive to improvement of design and implementation. However, funding facilitates sustainability. Despite agreement on what the overall goals of the Scheme are, there is some ambiguity as to the primary goal, which may influence design as well as implementation at country and school level.

Nutrition-Related Policies and Interventions

Overall, our findings are consistent with recent systematic reviews focusing on implementation of nutrition-related policies/interventions in kindergarten/school settings (Ismail et al., 2021; Lobczowska et al., 2022; Meshkovska et al., 2022). Both Lobczowska et al. (2022) and Meshkovska et al. (2022) utilized the CFIR to identify relevant constructs for the implementation of diet-related policies and direct provision of fruit and vegetable interventions, respectively. Consistent with the present study’s findings, both systematic reviews identified similar constructs as important, including cost, cosmopolitanism, implementation climate, readiness for implementation, and knowledge and beliefs about the intervention (Lobczowska et al., 2022; Meshkovska et al., 2022). Additionally, Meshkovska et al. (2022) highlighted the significance of design quality and packaging, adaptability, structural characteristics, networks and communications, engaging, and executing. In relation to design quality and packaging, Gosliner (2014) found that higher quality fruit and a variety of vegetables were associated with a substantial increase in fruit and vegetable consumption (Gosliner, 2014).

Although Ismail et al. (2021) did not use the CFIR they identified adaptability as a facilitator and lack of funding as a barrier to implementation. In contrast, the Scheme benefits from the availability of funding at the EU level, which contributes to its long-term sustainability. The importance of adaptability in policies and interventions implemented in school settings aligns with existing research in the field of nutrition (Aarestrup et al., 2014; Randby et al., 2021). Funding availability is both a potential barrier and facilitator to implementation in previous studies (Carbone et al., 2016; Lee et al., 2019; McIsaac et al., 2019). Challenges identified in the present work, such as engagement of teachers and suppliers, were also barriers to implementation in Ismail et al. (2021). Specifically, insufficient school participation or the perception that the program is too demanding were factors mentioned (Ismail et al., 2021). Similarly, the administrative burden was identified as a disincentive to participation by suppliers, which is consistent with research conducted in North-Rhine Westphalia (Germeten & Hartmann, 2016, 2017).

Implementation Science and Policy Implementation

We used a determinant framework from implementation science (Damschroder et al., 2009) to explore barriers and facilitators to the implementation of a nutrition policy program. However, the findings of our work are best understood, if contextualized both within implementation science and policy implementation. The EU Scheme is a public policy-derived program, which requires government involvement in its implementation. This is not always the case with school-based interventions. In the context of implementation science our approach may be considered a top-down view of implementation (country-level focus for a program ultimately targeting the school setting). In policy implementation, we are closer to applying a bottom-up view of implementation (reflecting the perceptions of non-political, ‘street level’ bureaucrats) (Hill & Hupe, 2014; Matland, 1995). Depending on level and perspective a specific construct may either be perceived as a barrier or as a facilitator.

Firstly, the public policy origins of the Scheme bring us to the issue of evidence use in policy-making. Policy theory has emphasized that public policies are evidence ‘informed’ rather than evidence ‘based’ (Boaz et al., 2019). Namely, the latter is an idealistic view on how public policy should be designed, rather than on how it is designed (Boaz et al., 2019). The concept of ‘bounded rationality’ is often used to describe the behavior of policy actors, where they are selectively using information to portray policy problems and solutions in a certain way (Boaz et al., 2019). The lack of complete agreement as to what the primary goal of the Scheme is may be a reflection of the bounded rationality thinking within each of the ministries involved. Consistent with this, evidence strength and quality did not figure prominently in our work, with a few exceptions. On the contrary, respondents expressed doubts as to the impact of the Scheme, and concern at the lack of a standardized approach to impact evaluation.

We can also explore our findings when contextualized in more explanatory public policy implementation models. In an attempt to overcome the clash of the top-down, bottom-up study approaches to policy implementation, Matland introduced the Ambiguity-Conflict Model of Policy Implementation (Matland, 1995). The model gives guidance about factors that influence implementation. Factors can come from (1) the ambiguity of the goals and means (actions) of a policy and (2) level of conflict that may exist among implementing bodies (Matland, 1995). Matland outlines four cases of policy implementation processes based on the ambiguity-conflict level. These are (1) administrative implementation (low conflict, low ambiguity); (2) political implementation (low ambiguity, high conflict); (3) experimental implementation (high ambiguity, low conflict); (4) symbolic implementation (high ambiguity, high conflict) (Matland, 1995).

With the Scheme, we found ambiguity in regard to the goals (lack of agreement on priority goals reflected through the construct knowledge and beliefs) as well as means (high flexibility in program and administrative model design reflected through the construct of adaptability). However, we found low conflict among implementing bodies—as they all agree on the need for the Scheme (reflected through the construct implementation climate). Thus, the Scheme may be a fitting example of experimental implementation. Although an in-depth discussion in this direction is beyond the scope of this research, the Ambiguity-Conflict Model may provide some explanatory power to the descriptive findings based on the CFIR.

Finally, goal ambiguity in our work was a potential barrier to implementation at country-level, as it may have direct influence on the adequacy of Scheme design. Policy implementation research, in particular as it pertains to policy-making at EU level may give a valuable context to such a finding. Zahariadis (Zahariadis, 2008) uses multiple streams theory (Kingdon, 2014) to explore EU policy-making, and stresses that ambiguity is a characteristic of EU policy-making. When problems, politics and policy come together to create a potential window for change, ambiguity (goal and means), together with informed use of evidence, frames problems in such a way as to enable their passage through the policy process (Zahariadis, 2008). The Scheme fits within this discussion, as its framing as a nutrition policy program with the purpose of both supporting agricultural producers, and creating healthy habits in children can be a justification for its continued existence among politicians. Thus, as much as goal ambiguity may be a potential barrier at country-level and possibly further down at school level, it is a potential facilitator at supranational level and from the political perspective. In contrast, ambiguity regarding means for implementation, in the case of the Scheme that is flexibility in action design and administrative setup, consistently facilitates implementation at all levels—supranational, country and school.

Fit of the CFIR for Study of Policy Implementation

With the above initial reflections in mind, we can say that the CFIR was a good fit as long as findings are contextualized in wider research from different fields (implementation science, policy implementation). Although to ensure no newly arising themes from the data would be missed, we opted to follow a process of analysis where we initially coded our data inductively, and the resulting themes were mapped to the framework, we eventually found that the CFIR worked well for this level of data collection and analysis. In this regard, recent changes to the framework (CFIR Guide, 2022), such as the introduction of sub-constructs (key stakeholders, innovation participants as part of engaging) was particularly helpful. These assisted in distinguishing the different groups of actors relevant to implementation—a need emphasized in recent implementation research more generally (Albers et al., 2020, 2021; Leeman et al., 2017).

The constructs we did not find in our study also merit a brief discussion. These are complexity (Scheme characteristics), culture (inner setting), planning (process) and most of the constructs under the individual characteristics domain with the exception of knowledge and beliefs about the intervention. Lack of content in regard to complexity, which is contrary to existing research (Lobczowska et al., 2022) and planning may be due to way of categorization of the data and decision to avoid double coding when possible. Some aspects of executing (steps and timelines of implementation) and design quality and packaging (administrative design) could also be coded under complexity. On the other hand, lack of content under culture and the individual characteristics domain may be due to the overall approach where we took the country-level as the inner setting domain of the CFIR. Linked to this, we did not explicitly explore potential influence of country-specific cultural characteristics on implementation. Similarly, individual characteristics may not come to the fore as strongly as they would if the inner setting was the school.

Limitations and Strengths

Through our recruitment strategy, we made significant efforts to ensure that any EU country enrolled in the Scheme had the opportunity to take part in this study. However, country representatives of the Scheme decided by themselves whether they wanted to participate. This may mean that our sample of the 10 countries that took part does not include those that face significant implementation challenges, as they may be more reluctant to participate. Data collection for this work was conducted during the COVID-19 pandemic, thus, a particularly challenging period for any persons involved in the Scheme, and in particular for persons from ministries of health and education. This may have prevented their participation. We did not explore sociodemographic variables of the respondents (such as gender or age) and their potential influence on the implementation process (if any). Finally, only one person (BM) has coded the qualitative interview data.

Nonetheless, the focus on the country-level implementation of the Scheme offered a unique opportunity for studying Scheme implementation across multiple EU countries, while using the CFIR for data collection and analysis. The latter in particular, is a valuable contribution of our research.

Implications for Research and Practice

Method and Theory Specific

Links between different constructs and across domains of the CFIR exist. This is relevant not only to the analysis of data, but in designing interview protocols that are not too long. An informed selection (rather than all) of constructs to explore makes the framework easier to use, while still identifying relevant barriers and facilitators. Future research is recommended to further explore links and preferably move toward developing system-based models for the study of barriers and facilitators to implementation rather than static linear frameworks.

Exploring public policies/programs through the lens of implementation science may still provide the more detailed and comprehensive guide to addressing implementation related questions. However, policy implementation research and theory provides much needed contextualization and explanatory power of findings. Both fields should be consulted to best understand policy implementation processes.

Scheme Specific

Although the adaptability of the Scheme is greatly appreciated, there are certain restrictions which could be re-considered at the EU level (ex. allow provision of fruit and vegetables to teachers). The relationships established between representatives from agriculture, health and education ministries may be a unique and unexpected positive outcome of the Scheme. These relationships should be utilized in future nutrition-related programs at EU and country-levels. Timing of the budget allocation is a challenge for some countries, causing significant barriers to implementation of the Scheme at the school level. The EU should explore what can be done for this challenge to be addressed. Although goal ambiguity can be useful at the supranational level, enabling politically acceptable framing and continuation of the financing of the Scheme, this seems to translate into a potential barrier to design and implementation that best fit the needs at country and school level. Priority goals should be set, at least at the country-level.

Conclusion

When looking at policy implementation at different levels, from the supranational, to the national, regional and finally school level as is the case with the Scheme, barriers and facilitators may change. What may be a facilitator at one level, becomes a barrier further down the line. Integrating what we know from different fields such as policy implementation and implementation science; political science and public health, may help in identifying particular barriers and facilitators at different stages of the policy implementation process.