First, mental health lessons are offered by mental healthcare professionals in schools to improve mental health literacy. Second, schoolteachers can undergo GKT, through which they learn to identify adolescents who show signs of suicidal behaviours and how to respond to those students. Third, a screening of students’ depression and suicide risk is conducted by the Public Health Service (PHS, in Dutch: GGD) using the Childhood Depression Inventory 2 (35) and the Questionnaire Assessing Suicide and Self Injury (36). Students identified as at risk for suicide are seen within 48 h by Child and Youth Health (CYH) professionals from the PHS for further assessment and referral, if necessary. Students with elevated depressive symptoms based on the Child Depression Inventory 2 are offered the indicated depression prevention intervention called OVK, which is based on cognitive behavioural group therapy (23). This intervention is usually provided by a duo of a care professionals within the school and a care professional in the youth care domain.

An integral part of STORM is collaboration within the network of care and education for adolescents (22, 23, 37). There are four main partners in this network: secondary and vocational schools, municipalities, PHS, and mental health professionals. Secondary and vocational schools are the settings for all interventions that are part of STORM (23). Within a region, these schools collaborate with municipalities in educational partnerships in supporting and caring for youth (38). Also, municipalities financially facilitate the implementation of STORM in practice (23). While regions can apply for a start-up budget through the Dutch National Agenda Suicide Prevention 2021–2025 subsidised by the Ministry of Health, Welfare, and Sports (28), municipalities still have to be involved for sustained financing after 2025. A team of mental health professionals provides consultation, training and personnel for carrying out the interventions (23).

An overview of the regional STORM programme structure, including the tasks of each partner, is provided in the second and third columns of Figure 1. Stakeholders from education, the PHS, mental health services, and municipalities collaborate in each part of the programme structure. For the current study, we defined four stakeholder categories: regional management, regional programme leaders, policymakers, and service providers. These are also indicated in the first column of Figure 1.

As part of the Dutch National Agenda Suicide Prevention 2021–2025 (28), STORM is being scaled-up to new regions in the Netherlands between 2021 and 2025. Interested regions could apply for a start-up implementation budget. The first two regions to receive this budget were selected in December 2022 and started implementing STORM in the academic year of 2023–2024 (39). To apply for the budget, the regions had to prepare for implementation and had thus already initiated several implementation strategies before the start of the current study. Furthermore, existing STORM regions have already introduced several implementation strategies in recent years. These strategies have already been formulated and provided to the new regions. An overview of the existing implementation strategies can be found in Supplementary File 2. The current study seeked to identify additional strategies from the literature that can help to overcome implementation barriers.

The tasks of implementation mapping (IM) described by Fernandez et al. (32) offer a systematic approach to developing an implementation plan by combining theory and co-creation with stakeholders in practice. IM has previously helped to identify implementation strategies for various preventive interventions and programmes (40–46). The five tasks of IM are the following: (1) conduct an implementation needs assessment to identify barriers and facilitators for implementation, (2) identify adoption and implementation outcomes, performance objectives, and change objectives, (3) select theoretical methods and design implementation strategies, (4) produce implementation protocols and materials, and (5) evaluate implementation outcomes (32).

We used the Consolidated Framework for Implementation Research (CFIR) from Damschroder et al. (47) to identify barriers to and facilitators for implementation in Task 1. The CFIR describes constructs in five domains to consider as potential barriers or facilitators. First, the innovation domain, which includes constructs related to the innovation being implemented. Second is the inner setting into which the innovation is implemented. Third is the outer setting within which the inner setting exists. The fourth domain concerns individuals and pertains to the roles and characteristics of individuals involved in the innovation being implemented. The last domain implementation process consists of constructs related to the activities and strategies used to implement the innovation (47).

For the selection of theoretical implementation strategies in Task 3, Powell, Waltz (48) compiled a list of 73 implementation strategies based on the results of the Expert Recommendations for Implementing Change (ERIC) study. Recently, this compilation has been adapted to improve its utility in educational settings in the School Implementation Strategies, Translating ERIC Resources (SISTER) Project (49). This project resulted in a list of 75 school-adapted implementation strategies. Both the ERIC and SISTER compilations guided the selection of strategies in this study.

Below, we specify our study procedures in terms of sample and recruitment, data collection, and data analysis conducted for each of the five IM tasks. An overview of our procedures for each task is described in Figure 2.

The needs assessment helps to identify important actors and potential barriers to and facilitators for implementation (32). For this, we conducted semi-structured interviews of approximately half an hour in February and March 2023. Study participants were selected from the region that has already implemented STORM (region 1) and two regions that were planning to implement STORM (regions 2 and 3) using purposive snowball sampling. First, we invited the national STORM team and regional programme leaders (n = 6) for interviews. Next, they helped to identify and contact other relevant stakeholders within the regions. We aimed to represent all stakeholder groups, and reached out to management and intervention providers from schools (n = 9), the PHS (n = 5), mental health organisations (n = 5), and municipalities (n = 4). Additionally, we invited mentors from secondary schools that had already implemented STORM (n = 2) for interviews, because mentors from secondary schools in the new regions had not yet been informed about the STORM approach. Participants were included until data saturation was reached, which meant that two researchers (KJ and CG) agreed that the last two interviews did not lead to new information. Researchers met with the national coordinators and regional programme leaders prior to the interviews to discuss which other stakeholders to include in the study. Researchers did not meet with any of the other participants prior to the interviews. The researchers’ characteristics did not influence the research questions, approach, methods, results, or transferability.

The interviews followed an interview guide based on the updated CIFR (47). The topics included STORM characteristics (example question: what is your perspective on STORM?) and barriers and facilitators (example question: what are things you think could complicate the implementation of STORM?), as well as the sub-topics Outer Setting, Inner Setting, Individuals, and Implementation Process. The full topic list is added in Supplementary File 3. Interviews were conducted by two researchers (KJ & CG) and audio-recorded. KJ was a PhD student at the time of the study with previous experience in conducting and analysing qualitative research. CG was a bachelor student and intern at the time of the study with no previous experience in qualitative research. The recordings were transcribed verbatim. After the interviews, a short summary was sent to the participants for verification.

To draw up a codebook, three researchers (KJ, CG, and FN) analysed six of the 20 interviews using open coding in Atlas.ti. FN (PhD) has previous experience in conducting and analysing qualitative research. Two researchers (KJ & FN) ordered the codes under the five major domains of the CFIR framework (47) and then combined them into overarching codes using axial coding. The complete codebook can be found in Supplementary File 4. Next, two researchers (KJ & CG) separately coded three transcripts, after which the coding was compared and variations in coding were discussed until both researchers agreed. Subsequently, all interview transcripts were analysed using deductive coding. Finally, the researchers analysed the coded data to identify barriers to and facilitators for the implementation of STORM in new regions.

Two researchers (KJ and FN) presented the identified barriers to four programme leaders from regions 2 and 3 and two stakeholders who had been involved in the implementation of STORM in region 1 during a brainstorming session in May 2023. We asked these participants to indicate, on a scale from 1 to 5 per barrier, whether they thought a barrier required immediate action or not using Mentimeter. The results of this brainstorming are available in Supplementary File 5 (in Dutch). Barriers that were scored higher than 3.5 were selected, while barriers that were scored lower than 2.5 were not. For barriers that were scored between 2.5 and 3.5, a group discussion determined whether the barrier was selected.

We formulated performance objectives based on the most important barriers identified in Task 1. For each performance objective, we formulated change objectives across five determinants based on the example of Kang and Foster (46): knowledge, awareness, skills, outcome expectancy, and self-efficacy. We chose this example, because it was the most complete objectives matrix we found.

Next, we selected theoretical implementation strategies to achieve the change objectives. First, for determinants that match the first-version CFIR constructs (47), strategies were identified using the CFIR-ERIC tool (48, 50). These strategies were then compared to the adapted compilation of ERIC implementation strategies for school-based implementation, SISTER (49), to identify strategies that are suitable for school-based implementation projects. Then, for determinants that did not fit the first-version CFIR constructs, suitable strategies were selected from the SISTER strategies (49).

A second round of brainstorming sessions was hosted in May 2023, one session with two national coordinators and one with four regional programme leaders. In both sessions, the main author (KJ) presented identified strategies for the selected barriers. Participants first discussed which of the identified strategies overlapped with existing implementation strategies. For the remaining strategies, participants discussed the extent to which they were realistic and relevant for practice. The implementation strategies that were considered both realistic and relevant for practice by national coordinators and regional programme leaders were selected for the implementation plan. If participants did not reach consensus about how realistic or relevant a strategy would be, the main author (KJ) made the final decision to include or exclude the strategy. A detailed description of our selection process can be found in Supplementary File 2. Next, the selected strategies were translated into practical applications in collaboration with the programme leaders from regions 2 and 3. The applications were reported following the recommendations for reporting implementation strategies by Proctor et al. (51). The implementation strategies found in the current study were added to the STORM implementation guide developed by the national STORM team. Besides implementation strategies, this document contains a detailed description of the programme components and programme structure of STORM. All future STORM regions will be offered this guide to aid their implementation efforts.

Based on the implementation plan, KJ and FN developed a plan to evaluate the implementation of STORM, as well as the application of the implementation plan in practice.

Twenty stakeholders were interviewed. Stakeholders were included until data saturation was reached. Their characteristics are presented in Table 1. Most participants had a coordinating role in the project, followed by providers, management, and policymakers. The number of participants per organisation type were spread evenly, except for municipalities, which were represented by only two participants.

We identified 21 barriers to and 13 facilitators for the implementation of STORM in new regions. An overview of all identified barriers and facilitators can be found in Supplementary File 6. These determinants include CFIR constructs and barriers that did not match the CFIR constructs. In the first brainstorming session with programme leaders, five barriers were selected for which implementation strategies should be identified. An overview of these barriers is presented in Table 2.

Multiple participants mentioned that schools are seen as the ideal setting for prevention, not only regarding mental health, but also for prevention of obesity or smoking. This leads to a high demand for schools. For example, one of the regional program leaders mentioned that “many societal developments are occurring and often it’s the schools who have to solve it.” Related to this, most participants recognised high work pressure in both education and healthcare as an important barrier for implementing a new approach. A school principal indicated that “it is an important theme, but honestly I do not have the people, the time, and the money to properly implement STORM.”

Another important issue raised by participants was that, at the time of the interviews, they, nor their colleagues, sufficiently understood the content of the STORM approach and what it means in practice. A school therapist mentioned, for example, that they “still need to receive a lot of information.” Moreover, participants thought that not all stakeholders saw added value in STORM compared to other interventions. A national coordinator indicated that “[organizations] struggle to de-implement [what they were already doing] to implement of STORM.”

It was also noted by some participants that network collaboration required improvement, especially between schools and mental healthcare services. A manager in mental healthcare mentioned that “education sometimes complains: ‘[mental healthcare organizations] do some test, but they never refer back to us’,” while a school therapist mentions that “collaboration [with mental healthcare organizations] does not exist in our school.”

Additional to the interview results, lack of attention to the sustainability of STORM in the current implementation efforts was identified as a barrier during the brainstorming sessions. Programme leaders felt that long-term sustainability was not receiving enough attention yet from stakeholders involved.

For each barrier selected in Task 1, we formulated performance and change objectives. The performance and change objectives are listed per stakeholder category in Table 3. For example, a performance objective for programme leaders related to the barrier ‘insufficient network collaboration’ and was that they should stimulate the development of sustainable partnerships between involved organisations. Change objectives for this performance objective were formulated under skills (i.e., able to connect organisations within the STORM network) and self-efficacy (i.e., are confident that they are able to connect organisations within the STORM network).

Not all barriers were relevant for each stakeholder category. ‘Partnerships and connections’ and ‘Sustainability’ were relevant for all categories, because all stakeholders are part of the STORM network, and the sustainable implementation of STORM should be achieved for all categories. ‘Relative advantage’ was not relevant for the programme leaders, because this is not a barrier for this stakeholder category based on the interviews. Since only the schools, the PHS, and mental healthcare organisations are involved in implementing STORM, the barriers ‘High demand for schools’ and ‘High work pressure’ were only relevant to these stakeholder categories. Finally, ‘Insufficient understanding of programme content’ was regarded as a relevant barrier for management only, because management has the final decision to participate in STORM, and thus needs to be well informed of the content.

To address the performance and change objectives formulated in Task 2, 14 implementation strategies were selected using the CFIR-ERIC tool and SISTER that matched the barriers identified (49, 50). Using the second round of brainstorming sessions with national coordinators and regional programme leaders, five implementation strategies were deemed relevant and realistic for practice: ‘pruning competing initiatives’, ‘change/alter environment’, ‘tailor strategies’, ‘identify and prepare champions’, and ‘promote network weaving’ (49). We translated these to practical applications and report on the strategies in Table 4 following the recommendations of Proctor et al. (51). The first strategy described is ‘collaborate with similar initiatives’, in which the idea is that regional program leaders actively identify other mental health school-based initiatives that are (being) implemented in their region, and look for ways to collaborate in the implementation process. The goal is to relieve the pressure on schools and minimise extra workload for service providers. Second is ‘free up time for STORM tasks’, in which organisation management allocate time to for implementing and executing STORM, while program leaders reserve budget to support organisations in doing so. Third is ‘tailor strategies’, meaning programme leaders adapt their communication style and message about STORM to the specific needs of various stakeholders, with the aim of improving adoption. Fourth is ‘identify and prepare STORM champions’, which entails both identifying and supporting individuals within involved organisations that are enthusiastic about the approach. The goal is to promote sustainment within those organisations through these individuals. Last is ‘regional network weaving’ through developing a social map of the organisations and individuals involved in STORM, and organising a joint kick-off session for those organisations and individuals.

In the final task, we developed a plan to evaluate the implementation of STORM in the new regions, as well as the application of the implementation plan over the course of two academic years. We identified outcomes from the implementation outcomes defined by Proctor et al. (52) and process evaluation guidelines from Moore at al. (53) and Saunders et al. (54). In Table 5, we summarise the outcomes for the implementation of STORM, including definitions, and how and when the outcomes will be measured. Providers, regional management, and policymakers will be involved in the evaluation of the implementation process through a survey and interviews at multiple time points. The measurement instruments to be used in the survey comprise a shortened version of the Acceptability of Intervention Measure, Intervention Appropriateness Measure, and Feasibility of Intervention Measure from Weiner et al. (55), and the Normalisation Measure Development Questionnaire (56). In Table 6, we summarise the outcomes for the application of the implementation plan, including definitions of the outcomes, and how and when the outcomes will be measured. Programme leaders and programme groups (see Figure 1) will be involved in the evaluation through a checklist of implementation strategies and focus group sessions. Additionally, we will analyse the administrative data for both evaluations.

A strength of the current study is that we systematically developed an implementation plan by following the tasks of IM (32). We did this in close collaboration with stakeholders who will implement STORM in practice, ensuring that the implementation plan matches the needs in practice. Additionally, we enhanced the credibility and transferability of our results through member checks, data and investigator triangulation, and sampling until we reached data saturation.

We recognise some limitations to our study as well. To begin, we mainly identified determinants related to the adoption and implementation of the intervention, and not to sustaining STORM over time. This is mostly likely because sustainability was not an explicit topic in our interviews and interviewees were in an early stage of pre-implementation. We discussed the lack of determinants with programme leaders and accordingly added a general determinant for sustainability. Furthermore, we reached out to multiple stakeholders and interviewed those who responded. Possibly, this led to selection bias if only participants with strong opinions about STORM, be these negative or positive, responded to our invitation. However, we asked participants to reflect on the perceptions of others in their field to minimise this bias.