Dissemination and implementation of suicide prevention training in one Scottish region

Background As part of a national co-ordinated and multifaceted response to the excess suicide rate, the Choose Life initiative, the Highland Choose Life Group launched an ambitious programme of training for National Health Service (NHS), Council and voluntary organisation staff. In this study of the dissemination and implementation of STORM (Skills-based Training On Risk Management), we set out to explore not only the outcomes of training, but key factors involved in the processes of diffusion, dissemination and implementation of the educational intervention. Methods Participants attending STORM training in Highland Region provided by 12 trained facilitators during the period March 2004 to February 2005 were recruited. Quantitative data collection from participants took place at three time points; immediately before training, immediately post-training and six months after training. Semi-structured telephone interviews were carried out with the training facilitators and with a sample of course participants 6 months after they had been trained. We have utilized the conceptual model described by Greenhalgh and colleagues in a Framework analysis of the data, for considering the determinants of diffusion, dissemination and implementation of interventions in health service delivery and organization. Results Some 203 individuals completed a series of questionnaire measures immediately pre (time 1) and immediately post (time 2) training and there were significant improvements in attitudes and confidence of participants. Key factors in the diffusion, dissemination and implementation process were the presence of a champion or local opinion leader who supported and directed the intervention, local adaptation of the materials, commissioning of a group of facilitators who were provided with financial and administrative support, dedicated time to provide the training and regular peer-support. Conclusion Features that contributed to the success of STORM were related to both the context (the multi-dimensional support provided from the host organisation and the favourable policy environment) and the intervention (openness to local adaptation, clinical relevance and utility), and the dynamic interaction between context and the intervention.


Background
Across Scotland during the period 1989-2004 male suicide rates increased by 22 percent and female suicide rates by 6 percent [1]. Over this period, Scotland has experienced a greater number of deaths from suicide than other countries in the UK [2] and the increase in deaths in men has led to a change in its rank from 12 th to 11 th highest rate within Europe [3].
Highland Region has one of the highest rates of suicide in Scotland [4,5]. Deaths by drowning, firearms and by others gases and vapours (mainly due to car exhausts) are overrepresented in the Highlands in comparison with the rest of Scotland. The reasons for these higher rates in the Highlands are unknown, but a number of factors could be involved. Many people who attempt to harm themselves are under the influence of alcohol and misuse of alcohol is a common problem in Scotland [6]. NHS Highland's catchment area comprises the largest and most sparsely populated part of the UK http://www.nhshigh land.scot.nhs.uk and there are barriers to accessing mental health care in rural areas including distance from services and a fear of being identified as accessing services [7]. Additionally, the excess suicide rate in farmers is well recognised [8]. Farmers have access to lethal means including firearms [9] and may differ in their help-seeking behaviour, with more presentation of physical symptoms of emotional distress [10].
As part of a national co-ordinated and multifaceted response to the excess suicide rate, the Choose Life initiative [11], the Highland Choose Life Group launched an ambitious programme of training for National Health Service (NHS), Council and voluntary organisation staff. Training for health workers has been advocated as a key route to suicide prevention in a number of countries [12]. However, reports of evaluation of staff training interventions are relatively scarce in the literature [13][14][15][16][17].
STORM (Skills-based Training On Risk Management) is a package originally developed by the University of Manchester. The content of the intervention reflects established assessment and management methods for patients with suicidal ideation and/or feelings of hopelessness [18][19][20][21]. STORM uses a handbook to provide background knowledge, and to reinforce and remind participants of the skills that are the main focus of the training.
The format of the intervention derives from adult learning theory [22], Bandura's Social Learning Theory [23] and published literature on changing the behaviour of health care workers [24]. STORM is primarily concerned with developing complex clinical communication skills and so utilizes role-play and video-feedback on performance. Attitudes also need to be addressed, which requires inter-active self reflection and reflection on the experiences of peers and case material demonstrated on videotape. If participants' current beliefs are challenged and changed in an interactive learning setting, then, with practise, they may also change their practice to reflect the change in belief (otherwise they will encounter cognitive dissonance, usually experienced as anxiety or frustration).
We have carried out three previous evaluations of STORM training and have demonstrated that the intervention brings about a change in skills and attitudes of participants [25][26][27]. We have also demonstrated that the training can be delivered to a multidisciplinary, multiprofessional audience across a single health district in England [26], but this intervention did not result in a change in the suicide rate [28]. Therefore, we hypothesized that training was likely to be a necessary, but not sufficient component of a comprehensive multi-faceted initiative. In a further study with mental health professionals, we examined the process and outcome of implementation of suicide prevention training in specialist mental health trusts in England and raised concerns about the lack of an educational culture within these organisations that might foster acquisition of new skills [27]. In this study of the diffusion, dissemination and implementation of STORM training in the Highland Region of Scotland, we aimed not only to replicate the key features of our original English study of community delivery of training [26] but to explore in detail the processes involved in the uptake of the educational intervention; Working definitions in this paper are summarised by Greenhalgh and colleagues [29]: diffusion is 'a passive phenomenon of social influence' whilst dissemination is defined as 'active and planned efforts to persuade target groups to adopt an innovation'. Adoption of an intervention, either by individuals or an organization, is underpinned by both of these processes. Implementation is the 'active and planned efforts to mainstream an innovation within an organisation'. Implementation incorporates the concept of 'sustainability' (known also as routinization and institutionalisation). It could be argued that implementation is the embodiment of all these processes but for simplicity we have here dealt with each separately.

Data and study sample
Participants attending STORM training in Highland Region provided by 12 trained facilitators during the period March 2004 to February 2005 were recruited. The trainers came from mental health services, and the majority of these were nurses, however others trained included psychologists, social workers, managers and a service user. Data collection from participants took place at three time points; immediately before training, immediately posttraining and six months after training (see table 1). Writ-ten consent was obtained from all participants. The immediately pre and immediately post training questionnaires were handed out by facilitators within the training, but the six month questionnaires were sent out by post; if no response was received within a month, a second copy of the questionnaire was sent out and telephone contact was attempted to remind participants about the questionnaire. Unlike in our original study of training, we did not seek to collect any pre-and post-training videotaped recordings of participants role-play, as this would have been logistically difficult given the distance between Manchester and the Highland region, with no on-site researcher.
We asked participants about age, gender, ethnicity, number of years in profession and previous suicide prevention training. Our main quantitative outcome measures were change in attitude to suicide prevention and confidence in the management of suicidal patients/clients. Attitudes were rated using the 'Attitudes to Suicide Prevention (ASP) Scale', where lower scores indicate more positive attitudes [30]. Confidence in the assessment and management of suicidal patients/clients was measured using a 100 mm visual analogue scale developed for the first STORM study. A minimum score of '0' is rated as 'not at all confident', while a maximum score of '100' is rated as 'very confident'. Satisfaction with participation in training was assessed using a questionnaire developed for our previous studies, focusing on satisfaction with specific aspects of the training package. Impact on clinical practice was assessed by an open-ended written questionnaire in which we asked participants to provide comments about how each of the training modules had been of use in their everyday work, All of the above areas were also addressed in the semi-structured telephone interviews with a sample of course participants and with a sample of course participants 6 months after they had been trained. A purposive sample of 12 participants was recruited on the basis of confidence and attitude scores (high or low initial scores and no change, high or low change scores), gender, age, previous training and profession to obtain a sample with maximum variation but that was also representative of the group as a whole.
In order to explore the processes of diffusion, dissemination and implementation in depth we carried out semistructured telephone interviews with the training facilitators (10 of the 12 facilitators were interviewed between 6 and 24 months post training with 5 of the facilitators being interviewed on two separate occasions at least 7 months apart) see table 2 for interview schedules and  table 3 for details of interviewees). We also interviewed the Consultant in Public Health (Dr Cameron Stark) and the Choose Life Coordinator for Highland (Michael Perera). All interviews were recorded and transcribed.

Ethical approval
This study was carried out as part of a multi-site evaluation of STORM training, which was approved by the North West Multi-Centre Research Ethics Committee MREC reference number: 04MRE08/8.

Analysis
Quantitative data were analysed using SPSS version 13.0 (SPSS Inc., Chicago, IL, USA) with paired two-tailed t-tests for data that were normally distributed and Wilcoxon Signed Ranks tests for ordinal data. Statistical significance was set at the 5% level.
Qualitative data was subjected to a Framework analysis [31] within MAXQDA software. The conceptual model of the diffusion, dissemination and implementation of interventions in health service delivery, described by Greenhalgh and colleagues [29], as utilized in the development of a framework/coding schedule. A template coding schedule was developed collaboratively by all three authors, from an initial discussion of a sub-set of ten transcripts. Additionally, all authors participated in coding the transcripts and ongoing discussions, which led to revisions of and improvements to the template. The theory of

Results and discussion
Attitudes, confidence and overall satisfaction Some 203 individuals completed a series of questionnaire measures immediately pre (time 1) and immediately post (time 2) training during the period March 2004 to February 2005. Sixty (31%) individuals also returned follow-up questionnaires sent out six months post training (time 3); although disappointingly low, this figure is in the same range as our previous experience of following up health professionals [26].
Of the total sample, the mean age was 43, 73% were female, 60% were born in Scotland (27% in England) and 98% specified 'White' as their ethnic group. 64% had received no previous training on suicide risk assessment or prevention and for the majority of those who had received previous training (68%) the duration had been of eight hours or less.
Nurse (38%) and social worker (20%) were the most common professional groups to receive/take part in training with 50% of those trained working in adult mental health services. However, a wide range of health and social care professionals participated including support workers (19), doctors (16 including 3 psychiatrists), health visitors (5), occupational therapists (4), a housing officer, nursery nurses (2) and a police officer. They managed a full range of client groups including children and young adults.
As in our previous studies [25][26][27], there were significant improvements in attitude and confidence (see tables 4 to 7). Confidence scores increased significantly from time 1 to time 2 (table 6) and from time 1 to time 3 (table 7), for each of the four items; therefore, confidence levels improved during/with training and remained improved over a period of 6 months. For each of the fourteen attitude items, scores decreased from time 1 to time 2 (table  4); therefore attitudes became more positive with/during training. All but two items (3 and 7) also showed a decrease in attitude score from time 1 to time 3 (table 5); therefore attitudes remained more positive up to six months post training.
Analysis of non-responders revealed that a higher proportion of males than females returned the time 3 questionnaire measures (42% compared with 26%) and a chisquared test showed this to be a significant difference at the 5% level (p < 0.05). There were no differences in terms of age, previous training or profession but those who responded at time 3 had more years experience in their profession on average than those who did not respond (mean years 16 compared with mean years 13). Again, a chi-squared test showed this difference was significant at the 5% level (p < 0.05). There is no significant difference between those individuals who responded at time 3 and those who did not in terms of both total confidence scores and total attitude scores at both time 1 and time 2, although those who responded had slightly higher confidence levels ( At time 2 people were asked to rate their satisfaction with the course across seven different elements (questionnaire available from authors). For each item the majority of people were entirely satisfied (60-92%); finding the course enjoyable, useful and relevant, with the right amount of detail. The lowest satisfaction scores related to people watching themselves on video in the 'therapist' role (item 5); however, only 6% of people reported this as being 'not at all' useful. In the qualitative interviews, participants also emphasised the positive value of both networking with colleagues and mutual learning: 'I just think that fact that it was a varied group of people, you know they all, they all contributed something either through case histories or just, you know relating their own experiences of management risk, of assessing and managing risk.' (Trainee H110)

Impact on clinical practice
At time 3 respondents were asked in the written questionnaire about what elements of the STORM training modules had been of use in their everyday work. The responses to this question were categorised in two ways. Initially the responses were coded as to whether or not the participant had used, found useful or made a positive comment about a particular module, then the positive responses were further broken down into emergent themes. Of those who provided a response to these items the majority made a positive comment for both the risk assessment (90%) and the crisis management (84%) modules. Both modules 'provided reassurance' with 12% of responders commenting on this in relation to the risk assessment module and 16% of responders commenting on this in relation to the crisis management module. The responders spontaneous comments also indicated that the assessment module 'improved confidence' (15%) while the management module 'improved knowledge' (18%). The highest number of comments for any one theme related to 'asking more, and more direct questions' during 'risk assessment' (25%). The latter theme was highlighted in one of the qualitative interviews with a participant:  However, there was also a viewpoint from some experienced workers that 'a lot of it was what we were doing anyway, but just possibly to a higher standard' (Trainee H172).

Diffusion and adoption of the STORM intervention
Utilising the conceptual framework described by Greenhalgh and colleagues [29] we were able to identify the key stages and processes in diffusion and implementation of the intervention from the interview data. The STORM training intervention had not previously been used in Scotland. The Consultant in Public Health who was the   And dissemination within the system began, with organisation of the necessary resources and commissioning of the training of a group of facilitators to provide training across Highland Region.

The dissemination process
If dissemination is 'active and planned efforts to persuade target groups to adopt an innovation' [29], the most important part of this stage was engaging and training the facilitators who were to implement the training within the organisation. Their engagement was absolutely essential to the task. Identified facilitators were provided with the training necessary to run local STORM courses.  8). Each module follows the same structure of brief lecture covering the salient points, demonstration with video excerpts, rehearsal using role-play and videotaped role-play with group feedback and discussion. STORM Facilitator training in suicide prevention is a four-day course. All four modules of suicide prevention are covered in the first two days (familiarisation with the model as participants) followed by a further two days of consolidation, practice delivering the course and discussion (training in facilitation of the STORM modules). The course can be delivered in two formats, either four consecutive days or split into 2 two-day sessions over two-weeks. The first group opted for the split format of training, spread over a three-week period to accommodate previously booked holidays; a practice not endorsed ordinarily but agreed upon due to the urgency of the training. The following two groups received training in the four-day training format. Each Facilitator was provided with a package of STORM materials needed to deliver courses Educational methods used in each training session: 1. Brief lectures on background knowledge and the skills to be acquired and rehearsed 2. Focused group discussion 3. Video demonstration of skills by health care professionals 4. Role-play (rehearsal of skills) in trios (professional-client-observer) and pairs (professionalclient) using pre-prepared role-play scripts to facilitate the practice of specific microskills 5. Video-feedback in small group setting of recorded role-played interviews carried out by course participants 6. Discussion to consolidate learning; specifically the translation of skills learned into practice The material can be modified in content for: • Primary care teams • Mental health care staff • Accident and emergency staff across Highland Region. Evaluation of outcomes (see above) was built into the training.
From the interviews with the facilitators a number of key factors in dissemination emerged. Financial support was available from the Choose Life initiative to purchase equipment required for training. One of the facilitators was funded to be a central coordinator and trainers were also funded for dedicated training time out of their usual jobs in order to implement the programme. This dedicated time and resources proved to be invaluable:

Implementation of the training
Some of the factors that have already been mentioned in association with diffusion, adoption and dissemination (for example, leadership and administrative support) were important in supporting the implementation process. However the motivation, capacity and competence of individual workers was also of key importance [29]. A devolved, flexible and supportive managerial and administrative structure, provided both co-ordination and administrative support functions for the implementation process. Furthermore, this structure supported the roll-out of the training by the facilitators with regular peer support and training meetings:

Routinization
Greenhalgh and her colleagues [29] do not distinguish implementation from the process of 'routinization', thus giving the impression that it is a part of the same process. However, little research has been undertaken to fully understand how an innovation becomes standard practice [29,34]. What evidence does exist suggests that successful implementation does not necessarily account for routinization [35][36][37]. According to Yin [32] 'routinization', which we understand to mean how an intervention becomes 'standard practice': 'may follow a series of stages: an improvisation stage, an expansion stage, and a disappearance stage (disappearance means that the innovative practice continues, but is no longer regarded as new)(p22).
When we followed up the facilitators there was some evidence of the intervention becoming no longer regarded as new, but incorporated into standard practice:

Conclusion
Utilizing primarily the model described by Greenhalgh and her colleagues [29] we have described in detail the key stages of diffusion, dissemination and implementation of a training intervention for suicide prevention, the STORM programme, as part of the multifaceted Choose Life initiative [11] in the Highland Region of Scotland. Key factors in the process were the presence of a champion or local opinion leader ('The Adopter') who supported and directed the intervention, local adaptation of the materials ('Linkage'), commissioning of a group of facilitators who were provided with financial and administrative support, dedicated time to provide the training and regular peer-support ('The implementation process' and 'System readiness for innovation'). We have also described the impact of training on participants (excluding acquisition of clinical skills which we did not assess on this occasion), which was comparable with our previous studies.
We can conclude that the features that contributed to the success of STORM in this context were both related to the context (the multi-dimensional support provided from the host organisation and the favourable policy environment) and the intervention (openness to local adaptation, clinical relevance and utility), and these appeared to interact in a dynamic and recursive relationship. The initial drive from the champion and supported facilitators (context) being followed by increase in demand for training if it was perceived as relevant and useful (intervention); necessitating organisational support to meet demand and implement policy changes (context) and resulting in further change to and improvement of the training (intervention) and so on. Applicability and acceptance of the innovation was evidenced by way of a change in staff attitudes, confidence and reported impact on clinical skills. Networking across organizational structures, as a result of multidisciplinary training sessions, helped to formalise the experience and communicate to other practitioners the innovation's usefulness.
We acknowledge a number of limitations of this study. We relied largely on quantitative data collected immediately after the training, by those who had indeed delivered it. Those who responded at six months may well have been biased towards individuals with a more positive view of the intervention. We did not capture the views of those who did not participate in training at all, and only have the facilitators' views of the challenges in recruiting staff for training. Those whom we were able to contact for interview may also have been favourably biased towards the STORM intervention. We also acknowledge that, given that the dissemination was led by a healthcare agency, we have not considered the wider implementation of STORM across and within other agencies.
From the perspective of this evaluation, the Greenhalgh et al. model [29] has a number of strengths. It provided us with a pragmatic model and framework for the analysis of our data that is evidence based and makes logical sense. However it also has some shortcomings. It does not adequately attend to the process of routinization which we believe is an important and distinct process. It is also more concerned with understanding what elements are necessary within diffusion, dissemination and implementation processes rather than how those elements interact to be of importance to the process. The model does not focus on the work that, for example, a champion or local opinion leader actually has to do in order to facilitate this recursive process. For this we consider that Normalisation Theory as described by May [35], which specifically considers the 'workability' of interventions in everyday practice, would provide a superior framework for investigating how interventions can be incorporated into health and social care systems; unlike the model described by Greenhalgh and her colleagues [29], the Normalisation Process Model specifically addresses incorporation into routine practice, but describes a more complex process than the stepwise 'routinization' described by Yin, addressing the work that has to be done to achieve 'normalisation' [32]. This may be problematic and meet particular barriers late into the process of implementation when the initial enthusiasm and support, particularly at a managerial and policy level, has waned or specific project related funding has ceased. What is unclear therefore is whether 'routinization' inevitably occurs if the process we have described above is effectively managed, and what is required, from whom, to make that management 'effective', in other words what are the mediating practices of these recursive relationships? Exploration of this process requires a lengthy commitment on the part of both organisations and researchers to continue in the collection of data through different stages of this process. This will be investigated in a study of the wider implementation of STORM training throughout Scotland.