Open Dialogue services around the world: a scoping survey exploring organizational characteristics in the implementation of the Open Dialogue approach in mental health services
www.frontiersin.org/articles/10.3389/fpsyg.2023.1241936/fullObjective: This cross-sectional study investigates the characteristics and practices of mental health care services implementing Open Dialogue (OD) globally.
Methods: A structured questionnaire including a self-assessment scale to measure teams’ adherence to Open Dialogue principles was developed. Data were collected from OD teams in various countries. Confirmatory Composite Analysis was employed to assess the validity and reliability of the OD self-assessment measurement. Partial Least Square multiple regression analysis was used to explore characteristics and practices which represent facilitating and hindering factors in OD implementation.
Results: The survey revealed steady growth in the number of OD services worldwide, with 142 teams across 24 countries by 2022, primarily located in Europe. Referrals predominantly came from general practitioners, hospitals, and self-referrals. A wide range of diagnostic profiles was treated with OD, with psychotic disorders being the most common. OD teams comprised professionals from diverse backgrounds with varying levels of OD training. Factors positively associated with OD self-assessment included a high percentage of staff with OD training, periodic supervisions, research capacity, multi-professional teams, self-referrals, outpatient services, younger client groups, and the involvement of experts by experience in periodic supervision.
Conclusion: The findings provide valuable insights into the characteristics and practices of OD teams globally, highlighting the need for increased training opportunities, supervision, and research engagement. Future research should follow the development of OD implementation over time, complement self-assessment with rigorous observations and external evaluations, focus on involving different stakeholders in the OD-self-assessment and investigate the long-term outcomes of OD in different contexts.
1. Introduction
Finding its roots in Need-Adapted Treatment (Alanen et al., 1991; Alanen, 1997), OD emerged as an innovative approach within the Finnish Western Lapland mental health services during the 1980s and 1990s. Seven principles became evident during the first research programs and psychotherapy training: (1) immediate help, (2) a social network perspective, (3) flexibility and mobility, (4) responsibility, (5) psychological continuity, (6) tolerance of uncertainty, and (7) dialogism (Seikkula et al., 2001). The first five principles regard the organizational logistics in which mental health services are provided, while the last two refer to the dialogic practice in which mental health professionals engage during network meetings with clients (Seikkula et al., 2003).
Since the 1990s, positive outcomes associated with OD have been documented in Western Lapland (Seikkula et al., 2006). Researchers observed that 82% of patients experiencing acute psychosis following the OD treatment showed no symptoms at the 5-years follow-up. Moreover, 86% of the patients had returned to a full-time job or studies, whereas only 14% were on disability allowance. Encouraging results were also observed during the following decade. A follow-up study confirmed that more than 80% of patients treated with the OD approach were fully employed or engaged in their studies after 2 years (Seikkula et al., 2011). Moreover, the study highlighted a cultural change in the use of the mental health service that led to earlier initiation of treatment, with a shorter duration of untreated psychosis and patients’ first contact happening at a lower age. Findings from a nineteen–year outcomes study indicated that many positive outcomes documented in previous studies are sustained over a long period (Bergström et al., 2018, 2022).
By 2011, OD was “well-established” in Western Lapland but still “little-known elsewhere” (Thomas, 2011). However, in the following decade, the approach started to be applied globally in different contexts and with disparate results. A review which focused on OD implementation in Scandinavia outside of Finland highlighted a significant variety of OD applications that, according to the authors, could be related to the intentional lack of operationalization of the OD principles (Buus et al., 2017). Other authors suggested that the different integrations of the OD approach into clinical practice may depend on the double challenge of introducing a transformation at the individual and the service level (Freeman et al., 2019).
Notwithstanding the heterogeneous panorama of OD applications, the approach has been investigated mainly using a naturalistic research design. The first randomized controlled trial on OD, evaluating the approach’s clinical and cost-effectiveness, was launched in the UK in 2017. The trial is part of the ODDESSI (Open Dialogue: Development and Evaluation of a Social Network Intervention for Severe Mental Illness) research program and compares OD against standard treatment in six mental health services in the UK. Results are expected in 2024 (Pilling et al., 2022).
Overall, the gradual implementation of OD into mental health services has not been described in detail, not even in Finland, despite the breadth of studies reporting on the origin of the approach (Buus et al., 2021). Research focusing on the implementation obstacles has been very scarce for many years, with one study describing organizational challenges observed among the nursing staff in Finland (Haarakangas et al., 2007) and a case study reporting the difficulties of an outreach team practising OD in Denmark (Søndergaard, 2009). More recent research (Gordon et al., 2016; Heumann et al., 2023; Skourteli et al., 2023) highlighted organizational and ideological barriers such as lack of time and resources, rigid professional hierarchy and the burden of working across two different models at the same time (Dawson et al., 2021; von Peter et al., 2023). Although these qualitative studies suggest some adaptation strategies, more global and quantitative research on the implementation of the OD approach is still needed.
Moreover, the fact that the OD approach has not gone through the process of manualisation – that is, the development of a procedure that can be replicated with sufficient uniformity (Waters et al., 2021) poses additional challenges, especially in assessing OD-fidelity. A measure called COMFIDE (Alvarez Monjaras, 2019; Alvarez-Monjaras et al., 2023) was developed as part of the ODDESSI trial to evaluate a good standard of care for community mental health services providing OD and standard crisis and community care. Although more research on OD-fidelity is needed to identify specific and measurable elements (Waters et al., 2021), items and topics from the COMFIDE scale may currently be used for fidelity assessments at a global level.
Different approaches to implement Peer supported Open Dialogue (POD), connecting social and professional networks, have also been described in the last years (Razzaque and Stockmann, 2016; Kemp et al., 2020; Lorenz-Artz et al., 2023). Bellingham et al. (2018) reported that several models of POD had been embedded into clinical practice. In some cases, peer supporters may have a role very similar to that of professional therapists, whereas, in others, they have more limited space. For example, persons with lived experience may not participate in network meetings but be involved as supporters of the community. In other models, they may participate in the network meetings but not attend the reflection spaces addressed only to the clinicians (Bellingham et al., 2018). Due to the heterogeneity of models and scarcity of research on peer workers, a more comprehensive investigation is needed in this area (Kemp et al., 2020).
Pivotal elements in the development of OD services are training, supervision and intervision which need to be “carefully planned” and considered an integral part of the approach (Buus et al., 2017) – intervision is hereby a form of colleague-based supervision practised in Peer-Supported Open Dialogue (see Razzaque, 2019). In Western Lapland, the training of the staff members was one of the three central components of the community psychiatric system (Alakare and Seikkula, 2021), together with the “Family and Team-centeredness” and the research project (Seikkula et al., 2011). Training activities cover theory, supervision, and seminars in which participants are required to analyze their background and family of origin. Experiences of training from different countries, including Norway, the US, the UK, Australia and Italy, have been reported in the literature (Hopfenbeck, 2015; Aderhold and Borst, 2016; Buus et al., 2017; Cubellis, 2020; Florence et al., 2020; Hopper et al., 2020; Jacobsen et al., 2021; Schubert et al., 2021; Pocobello, 2021b). Intervision, intended as a form of colleague-based supervision, and training, including “intentional peer support,” are also part of the activities for peer workers (Hopfenbeck, 2015; Razzaque and Stockmann, 2016; Razzaque, 2019; Hopper et al., 2020). As far as we know, there has been no global investigation on the extent of training and supervision practices in OD services worldwide. Quantitative data on how many people involved in OD services have completed or are completing the training are unavailable. Moreover, the frequency and type of supervision have not been explored so far.
Overall, the requirements for and barriers to the implementation of OD on both the level of organizational structures and staff competencies need to be addressed in research and require a deeper investigation (Mosse et al., 2023).
The present scoping survey was designed to map and explore the existing evidence about the implementation of OD-services globally (Pocobello, 2021a) and to investigate the impact of factors such as OD-training, supervision, research, the involvement of experts by experience and organizational characteristics on services’ OD-self-assessment (OD-SA). In this context, the term “expert by experience” refers to an individual who has/had personal, lived experience with mental health challenges or the mental health care system. This term acknowledges that individuals who have gone through these experiences possess a unique and valuable perspective that can contribute significantly to the improvement of mental health services, policies, and practices (Gupta et al., 2023).
The objectives of the global scoping survey can be summarized as follows:
a. To describe services practising Open Dialogue around the globe;
b. To pilot testing and validating an Open Dialogue Service Survey Scale including an OD- self-assessment (OD-SA) scale;
c. To construct an exploratory model of the organizational predictors of OD self-assessment;
d. To provide a measure of teams’ degree of self-assessed adherence to the seven OD principles and
e. To identify services ready for outcome evaluation studies.
The study is part of the project HOPEnDialogue,1 financed by the Open Excellence Foundation, which aims at investigating the implementation and effectiveness of Open Dialogue in different mental health care contexts around the world.
2. Methods
The study is reported according to the CROSS Checklist for Reporting Survey Studies (Sharma et al., 2021) to ensure rigor and credibility.
2.1. Study design
We used a cross-sectional study design to collect data from multiple teams providing OD services in mental health care across different countries. The study design involved (1) the development and validation of a OD-self-assessment scale and (2) a quantitatively structured questionnaires to gather information on various aspects of OD services, including their structural characteristics, personnel OD-trainings, as well as practices regarding supervisions, involvement of experts by experience, and research activities.
2.2. Ethical clearance
All respondents to the survey have completed an informed consent form embedded in the first page of the questionnaire. A skip-logic survey method was in place in the online form to ensure no collecting of information from respondents who disagreed with the informed consent question. Respondents were informed about the possibility of withdrawing from the survey at any time. Respondents could leave questions not answered.
The survey was not anonymous, since the address of the service and personal contact information of the professional completing the survey on behalf of the OD team was used to check for accuracy and prevent multiple participation. Confidentiality was guaranteed by limiting access to this information to the research team of the ISTC-CNR and saving electronic data on password-protected computers.
Ethical clearance with authorization value was not necessary for this study.
2.3. Respondents
Team members of OD-services with leadership responsibility were invited to complete the survey on behalf of the entire facility or OD-team. Individual OD practitioners were excluded.
As the survey is part of the project HOPEnDialogue, it was advertised and primarily distributed through its website Members of the HOPEnDialogue advisory board helped disseminate the survey in their different countries and networks through social media and mailing lists. We have contacted professionals from countries not represented on the board to ask for their support in spreading the survey at a national or local level. The first round of data was collected online using the Survey Monkey platform from January to September 2020. In total, 136 questionnaires were filled out online. The data were exported into Excel. The second round of data collection happened from January 2021 to February 2022 and involved six teams just concluding their foundation training. The questionnaires were filled and sent as PDFs to RP and FC, who added them to the Excel data set. The reason for this late recruitment was related to our intention to include all the services contacting us to have as much as possible comprehensive view of OD implementation globally. In total, 142 services participated in the survey.
2.4. Data diagnostics
Data was checked and controlled for consistency. Where available and possible, missing data were completed by checking back with survey responders via email. Of the 142 questionnaires received during the data collection period, the data of 24 OD-services had to be excluded due to incomplete datasets, mainly from the 6th item (clients’ characteristics served in the center) onwards. Often, the unavailability of informants made it impossible to assist in completing the missing questionnaire sections. We undertook a missing data diagnosis on the data from the remaining 118 centers and did not detect systematic patterns (checking summary statistics for variables, counting the number of missing and non-missing values for each variable, correlations to examine if the missingness in one variable is associated with another variable).
2.5. Data analysis strategy
To evaluate the statistical validity and reliability of the measurement model of the OD-self-assessment (OD-SA) scale, non-parametric Confirmatory Composite Analysis (CCA; Dijkstra and Henseler, 2011; Schuberth et al., 2018) was calculated with SmartPLS 4® (Ringle et al., 2022). We followed the procedural steps for CCA outlined by Hair et al. (2020). The reliability of the variables was tested using Cronbach’s Alpha and Composite Reliability (ρA).
Descriptive data of the survey have been checked for consistency in Excel spreadsheets and transferred to SPSS® 27.0 (IBM Corp, 2020) for the descriptive and explorative Cluster analysis.
For the descriptive analysis Continuous variables were described using means (M) and standard deviation (SD); for discrete count variables, proportions were reported. The Shapiro–Wilk test was used to assess the normal distribution of continuous variables. As a non-parametric test for differences in group value of ps Kruskal-Wallis’ test was used. Association between structural aspects of OD-services was assessed using Loglinear modeling when it concerned the frequency of categorical data (see structural characteristics). The significance level was determined as p < 0.05 for all analyses.
For the explorative data analysis bivariate non-parametric correlations were computed between the services OD-SA score and the descriptor variables to identify significant associations.
• To explore structural characteristics of the MHS in which OD-teams emerged and operated, an unsaturated model was chosen using SPSS Statistics’ hierarchical loglinear model selection process with a backwards elimination stepwise procedure;
• To explore professional taxonomies in OD-services hierarchical cluster analysis was used; Provided the sample size of n = 118 teams, the number of clusters was estimated to range between n/30 = 4 and n/60 = 2. To identify equally sized clusters, hierarchical cluster analysis with Ward’s method was used. Count values per variable of the eight professional profiles was standardized to correct for important differences in the counts of personnel in teams. A chi-squared measure of distance was used as a similarity measure;
• A Kruskal-Wallis’ test was calculated to test for significant differences between the OD-teams belonging to different professional clusters. Visual inspection of boxplots was used to assess the similarity of the distributions of OD-SA scores (OD-SA 15) of groups/clusters. Pairwise comparisons were performed using Dunn’s (1961) procedure with a Bonferroni correction for multiple comparisons. Adjusted value of ps are presented.
Finally, partial least squares (PLS) regression analysis was conducted to explain the variance of OD-Teams self-assessment scores based on teams- and their services’ characteristics. PLS regression, is a statistical method used in the presence of many predictor variables which may be highly correlated. It is especially useful when the number of predictor variables is larger than the number of observations, a situation where traditional regression methods like ordinary least squares (OLS) struggle (Hair et al., 2018).
The Breusch-Pagan test was used to assess Heteroskedasticity; the PLS algorithm was set to heteroskedasticity consistent standard errors (HC3) to handle the distribution in case of a positive Breusch-Pagan test. HC3 correction calculates robust standard errors that take into consideration the potential heteroskedasticity in the data. It provides more accurate standard errors that are less affected by the presence of heteroskedasticity. This, in turn, ensures that hypothesis tests and confidence intervals derived from the regression analysis are more reliable and valid, even when heteroskedasticity is present (Kaufman, 2013). To deal with missing data the algorithm was set to mean replacement (no weighting vector was used).
2.6. Instruments: the Open Dialogue teams survey scale development
RP and TeS developed a first draft of the questionnaire after reviewing the current literature on OD implementation. All authors revised the first draft, and RP further refined the revisions until a consensus was reached. At the end of the development process, 65 questions were finalized for this survey. The full questionnaire is attached as Supplementary material to the article. The items related to OD-team’s transparency, self-disclosure, intervision, intended as a form of colleague-based supervision (Razzaque, 2019) and training were adapted from the OD addendum of the COMFIDE-Questionnaire (Alvarez Monjaras, 2019). The questionnaire was then pilot tested with one OD team, but no changes to the survey content were necessary.
The survey was structured in six sections, each dedicated to an independent dimension of mental health services. In the general part (1) the year the OD-service first started, (2) the presence of other therapeutic models integrated in the mental health service; (3) the age range of patients the OD-service was dedicated to; (4) what diagnostic groups of patients the OD-service works were inquired. Furthermore, three characteristics of the structural domain of mental health services were inquired: (a) the sector to which the MHS belongs [public/other (private, third sector); since the distinction between the private and third sectors was not always clear to respondents, we collapsed these two categories into one category (‘non-public sector’)]; (b) whether the MHS operates as an inpatient or outpatient service, or both; (c) if the MHS is stand-alone- or integrated with other services or other. We further asked about estimating the number of professionals (nurses/occupational therapists/peer-support workers/psychiatrists/psychologists/psychotherapists/social workers /support workers/others) constituting the OD-team.
2.6.1. OD-self-assessment scale: development and validation
For the teams’ OD-self-assessment (OD-SA), we developed 17 items by reviewing the literature on good practice in Open Dialogue. The starting point for the development of the items were the seven principles of OD (Seikkula et al., 2003) with the aim of formulating a minimum of two items for each principle as affirmative statements. Respondents were asked to indicate for each statement the extent to which it reflected the clinical practice in their services over the past 3 months on a five-point Likert scale from 1 = “never,” 2 = “rarely,” 3 = “sometimes,” 4 = “frequently” to 5 = “almost always.” Consequently, higher scores reflected better OD-self-assessment (OD-SA) than lower scores.
2.6.2. Scale validation: confirmatory composite analysis
The content validity of the 17 items composing the OD-self-assessment scale is based on the conceptual review of the OD-Principles formulated by Seikkula et al. (2003).
Discriminant validity refers to the extent to which model constructs may be distinguished from each other. Different to the first five organizational principles, principles 6 (Tolerating Uncertainty) and 7 (Dialogicallity) relate to the way of being and engaging with clients during the network meetings. Due to a low discriminant validity of the two scales – Heterotrait-Monotrait ratio (HTMT) of 0.917 was above the recommended 0.900 threshold (Henseler et al., 2015) – they were merged into one four-item scale of ‘OD-Adherence’ (OD-ADH). For the resulting scales the values of average variance extracted (AVE) exceeded the Fornell and Larcker (1981) criterion (a minimum of 0.5) and HTMT ratio was significantly below 0.90 indicating a good discriminant validity.
Assessing first Cronbach’s alpha reliability of the constructs, it turned out to be ‘good’ for P1 (r = 806) and P2 (r = 0.806), acceptable for ‘ADH’ (r = 0.767) however ‘doubtful’ for P3 (r = 0.683), P5 (r = 0.632), and ‘not acceptable’ for P4 (r = 0.332). Reviewing all factor loadings, we eliminated two critically low loading items (I26: λ = 0.52 -> P3; I30: λ = 0.63 -> P4) from each of the two scales, turning the P4 scale into a single-item construct consisting of I29 only and the P3 scale into a three-item scale with close to ‘acceptable’ reliability (r = 0.698); the internal consistency of P5 (r = 0.623) remained low according to the generally applied Cronbach’s Alpha criterion (r = 0.705).
New research suggests that the use of a single criterion for established instruments as well as newly explored and developed studies – as the one at hand – may be too conservative for scales developed within the context of the latter (see Hair et al., 2019, p. 9; Hair et al., 2021, p. 119). Composite reliability is therefore recommended for the reliability assessment of newly developed scales (Hair et al., 2018) and the values evidence the scales acceptable level of reliability according to the standards for exploratory studies (see ρA in Table 1).
rest in link