“To be means to communicate dialogically.” – Mikhail Bakhtin, Problems of Dostoevsky’s Poetics (1984, p. 252)
Open Dialogue has attracted considerable interest in mental health services, and it is not difficult to understand why.
Developed in Western Lapland, Finland, initially in services treating psychosis, Open Dialogue is not simply another psychological therapy. It is both a therapeutic practice and a way of organising mental health care: responding rapidly in crisis, maintaining continuity of clinicians, involving family and social networks where wanted and appropriate, tolerating uncertainty rather than rushing towards decisions, and making decisions as collaboratively as possible.
These principles have considerable intuitive appeal, particularly in systems where a person in crisis may encounter multiple teams, repeated assessments and fragmented relationships. However, enthusiasm has historically run ahead of the quality of the evidence. Early observational studies, mainly involving people with psychosis, reported encouraging outcomes, including less hospital treatment and better longer-term functioning (Seikkula et al., 2006; Bergström et al., 2018). Systematic reviews have nevertheless highlighted substantial risks of bias, heterogeneous implementation and, until ODDESSI, an absence of randomised evidence (Freeman et al., 2019).
ODDESSI therefore matters. It is the first large randomised evaluation of Open Dialogue and attempted something unusually difficult: testing a whole-service model in routine NHS mental health care.
So, did Open Dialogue work?
Perhaps the most useful answer is that it depends on what we think better mental health crisis care should achieve.
Methods
ODDESSI was a pragmatic, multicentre, cluster-randomised trial conducted across five NHS mental health services in England. GP-practice clusters were allocated to Open Dialogue or treatment as usual, and adults presenting to participating services in a mental health crisis were recruited according to the cluster in which their GP practice sat.
Cluster randomisation made sense because Open Dialogue aims to reorganise how whole teams work; asking clinicians to deliver Open Dialogue to one person and conventional care to the next would risk substantial contamination. Participants allocated to Open Dialogue were intended to receive crisis and continuing care from an integrated team, including network meetings, peer support and greater continuity. Treatment as usual comprised existing NHS crisis and community mental health services (Pilling et al., 2022).
The primary outcome was time to first relapse after recovery from the initial crisis over two years. Recovery and relapse were reconstructed from electronic clinical records using structured summaries rated by assessors who were blinded to allocation. Recovery meant an absence of significant symptoms with adequate social functioning; relapse meant the return of significant symptoms alongside deterioration in social functioning.
This is important because it is separate from the trial’s self-rated recovery measure. The Questionnaire about the Process of Recovery (QPR-15) was a participant-reported secondary outcome. Other secondary outcomes included time to initial recovery, days in recovery, hospital admission and inpatient days, re-referral to crisis or secondary services, general health, satisfaction with care and social-network measures (Pilling et al., 2022).
A total of 494 people participated: 271 allocated to Open Dialogue and 223 to treatment as usual.
Results
On its prespecified primary outcome, ODDESSI did not show that Open Dialogue significantly prolonged time to relapse. Among 389 participants contributing to the primary analysis, the marginal hazard ratio was 0.95 (95% CI 0.67 to 1.32). At 24 months following recovery, an estimated 54% of participants receiving Open Dialogue and 49% receiving treatment as usual remained relapse-free. There was also no statistically significant difference in time to initial recovery or total days spent in recovery.
The secondary outcomes were more favourable to Open Dialogue.
In the raw data, 21% of participants in Open Dialogue had a psychiatric inpatient admission compared with 44% receiving treatment as usual. Following adjustment, participants allocated to Open Dialogue had more than three times the odds of avoiding admission altogether (marginal OR 3.30, 95% CI 2.18 to 5.22).
Re-referral to crisis or other secondary mental health services occurred in 40% of the Open Dialogue group and 60% of treatment as usual (marginal OR 0.44, 95% CI 0.30 to 0.64). Among people who were admitted, there was no clear difference in inpatient days.
This pattern is broadly consistent with earlier observational research from Denmark, which found lower use of emergency psychiatric services among young people receiving Open Dialogue, although findings across different forms of service use were mixed (Buus et al., 2019).
Participant-reported outcomes also favoured Open Dialogue. At 24 months, participants reported better general health on the EQ-VAS and better self-rated recovery on the QPR-15. Satisfaction with services was higher at three months.
There was no statistically significant difference in social-network size, and the difference in social-network quality did not reach statistical significance. The planned shared decision-making measure could not be formally analysed because of missing data.
The authors reported no significant safety concerns; 376 of 386 serious adverse events were judged unrelated to the intervention.
Conclusions
The authors conclude that Open Dialogue did not reduce time to first relapse compared with treatment as usual. Its effects on inpatient use and participant-reported outcomes nevertheless justify further evaluation of its effectiveness and acceptability in crisis and continuing community care.
Strengths and limitations
The ODDESSI team deserve considerable credit for testing a whole-service intervention in routine practice. Whole-service models are harder to evaluate than medications or individual manualised therapies: teams had to be trained, care pathways altered, clusters randomised, people recruited during mental health crises, and follow-up maintained across changing NHS services. The trial also involved people with lived experience in its design, delivery, oversight and interpretation, assessed fidelity and adherence, and used blinded raters for the record-based recovery and relapse outcomes.
Importantly, Open Dialogue was compared with actual NHS care rather than an artificially standardised control condition. Treatment as usual inevitably varied, which also makes the comparison highly relevant to the practical question facing services: does reorganising existing care around Open Dialogue produce different outcomes from the system currently in place?
There are, however, important limitations.
First, the trial had one primary outcome alongside multiple prespecified secondary outcomes, with no adjustment for multiple comparisons. The admission, re-referral and participant-reported findings therefore cannot be treated as equivalent to a positive primary endpoint. At the same time, they should not simply be discarded: several outcomes pointed in a broadly favourable direction.
The appropriate next step is independent replication, ideally with prespecified primary outcomes that reflect what a whole-service intervention is intended to change. Second, cluster randomisation was appropriate for a service-level intervention. It nevertheless introduces potential selection and imbalance issues. Participants were recruited after clusters had been allocated, and neither clinicians nor participants could be blinded. The authors also note observed between-group differences in variables including primary diagnosis, drug and alcohol misuse, and educational attainment, which may have affected outcomes.
Third, implementation was complex. Overall fidelity and adherence were judged acceptable. The authors nevertheless note that fewer adherence recordings were available than planned and that network meetings with family members were not the majority of interactions in the Open Dialogue arm. This makes it harder to know which components participants actually received and which might account for the observed effects.
Fourth, the population was diagnostically heterogeneous. This is a strength if Open Dialogue is intended as a transdiagnostic crisis-care model; it also complicates interpretation of the primary outcome. Open Dialogue was developed with an initial focus on psychosis, much of the earlier evidence concerned psychosis, and the relapse measure used in ODDESSI originated in first-episode psychosis research. Yet depression was the most common primary diagnosis in ODDESSI (26%), while only 16% had a primary psychotic disorder. A measure developed in psychosis may not capture equally well meaningful deterioration and recovery across depression, trauma-related difficulties, personality disorder, anxiety and other crises. Relapse and recovery were also reconstructed from routine clinical records, which inevitably vary in completeness and emphasis.
Finally, 69% of participants were White British and eligibility required sufficient English to participate in research. For an intervention centred on dialogue, family and social networks, transferability across culturally and linguistically diverse communities should not be assumed. Future studies need deliberate attention to diversity and equitable implementation.
Implications for practice
ODDESSI does not establish that Open Dialogue prevents relapse. Claiming otherwise would go beyond the evidence. However, describing it simply as evidence that “Open Dialogue failed” would also miss important findings.
Participants receiving Open Dialogue were less likely to be admitted to hospital, less likely to return to crisis or secondary services, and reported better recovery, general health and satisfaction with care. These were secondary findings and require replication. Even so, they concern outcomes that matter to people using services and to systems attempting to provide more care in the community.
UK qualitative studies also suggest that some people receiving Open Dialogue describe feeling more listened to, understood and involved in their care, although experiences are not universally positive (Tribe et al., 2019; Sunthararajah et al., 2022).
Psychiatric admission should not itself be treated as a failure: for some people it is necessary, wanted and potentially lifesaving, and lower admission rates could theoretically reflect altered thresholds rather than better outcomes. Equally, avoiding an unwanted or unnecessary admission while providing safe and acceptable community care is meaningful. Future work should therefore examine hospital use alongside symptoms, functioning, risk, coercion, adverse events and service-user preferences.
The substantial reduction in acute service use also makes the economic evaluation of Open Dialogue particularly important. A separate health-economic analysis from ODDESSI is planned, which should help establish whether these differences translate into better value for mental health services from an economic perspective.
ODDESSI raises a broader question about how we evaluate complex mental health service interventions. If an intervention aims to change relationships, continuity and decision-making as well as symptoms, relapse may not be the only outcome that matters. Patient-reported recovery measures may be particularly useful in capturing dimensions of recovery that symptom-focused measures can miss (Garrido-Cervera et al., 2025). Future trials should retain rigorous prespecified outcomes while also asking which components of Open Dialogue change people’s lives, for whom, and whether those changes can be delivered safely and equitably.
The next question is therefore not simply, “Does Open Dialogue work?” It is: what changes when services organise themselves around dialogue, continuity and relationships, and which of those changes are worth sustaining?
That seems a much more useful dialogue to have.
Statement of interests
Diamantis Toutountzidis is an Open Dialogue practitioner and mentor and is currently training to become an Open Dialogue trainer. He also works as an Open Dialogue lead practitioner for a newly established company providing residential mental health care in the UK. These professional roles represent a potential competing interest in relation to Open Dialogue. He used AI to help with grammar.
Editor
Edited by Simon Bradstreet.
Links
Primary paper
Pilling S, Craig T, Clarke K, Chu P et al. (2026). Open Dialogue versus treatment as usual for adults presenting in crisis to mental health services in England (the ODDESSI Trial): a multisite cluster-randomised trial. Lancet Psychiatry
Other references
Bakhtin, M. (1984). Problems of Dostoevsky’s Poetics. Edited and translated by Caryl Emerson. University of Minnesota Press.
Bergström T, Seikkula J, Alakare B, et al. (2018). The family-oriented Open Dialogue approach in the treatment of first-episode psychosis: Nineteen-year outcomes. Psychiatry Research, 270, 168-175.
Buus N, Jacobsen EK, Bojesen AB, et al. (2019). The association between Open Dialogue to young Danes in acute psychiatric crisis and their use of health care and social services: A retrospective register-based cohort study. International Journal of Nursing Studies, 91, 119-127.
Freeman AM, Tribe RH, Stott JCH, Pilling S. (2019). Open Dialogue: A review of the evidence. Psychiatric Services, 70(1), 46-59.
Garrido-Cervera JA, Ruiz-Granados MI, Cuesta-Vargas, AI, Sánchez-Guarnido, AJ (2025). Patient-reported outcome measures (PROMs) for recovery in mental health: A scoping review. Community Mental Health Journal, 61, 890-898.
Pilling S, Clarke K, Parker G, James K, Landau S, Weaver T, Razzaque R, Craig T. (2022). Open Dialogue compared to treatment as usual for adults experiencing a mental health crisis: Protocol for the ODDESSI multi-site cluster randomised controlled trial. Contemporary Clinical Trials, 113, 106664.
Seikkula J, Aaltonen J, Alakare B, Haarakangas K, Keränen J, Lehtinen K. (2006). Five-year experience of first-episode nonaffective psychosis in Open Dialogue approach: Treatment principles, follow-up outcomes, and two case studies. Psychotherapy Research, 16(2), 214-228.
Sunthararajah S, Clarke K, Razzaque R, et al. (2022). Exploring patients’ experience of peer-supported Open Dialogue and standard care following a mental health crisis: qualitative 3-month follow-up study. BJPsych Open, 8(4), e139. doi:10.1192/bjo.2022.542.
Tribe RH, Freeman AM, Livingstone S, Stott JCH, Pilling S. (2019). Open Dialogue in the UK: qualitative study. BJPsych Open, 5(4), e49.
Photo credits
Photo by Priscilla Du Preez 🇨🇦 on Unsplash
Photo by Towfiqu barbhuiya on Unsplash
Photo by Verne Ho on Unsplash
Photo by Road Trip with Raj on Unsplash
Photo by Ilya Chunin on Unsplash