You can’t work in mental health services for too long without hearing about Open Dialogue.
There certainly aren’t a shortage of bold claims about the model of care used to treat psychosis in a small pocket of Finland, for example, as the Open Dialogue UK website states:
‘They have the best documented outcomes in the Western World. For example, around 75% of those experiencing psychosis have returned to work or study within 2 years and only around 20% are still taking antipsychotic medication at 2 year follow-up’.
And as a well-publicised play inspired by the model suggests, there is a rumour that they have actually eradicated schizophrenia in Western Lapland. These are substantial claims – so substantial that a seriously hard objective look at both their model and the evidence for its success is required to verify them. I wasn’t able to find such an interrogative assessment online, so I thought I’d write one myself.
My aim is to contribute to a transparent, honest discussion about what the effective parts of Open Dialogue might be, and how confident we can be of that. There has been a lot of fervour about this approach, but we also need an impartial examination of the facts. Please do leave comments underneath the post.
I want to make a disclaimer straight away: Yes, I am a psychiatrist, but no, I am neither a particular fan of medication nor a particular hater of talking therapy. I try to appraise the evidence for both objectively. What I am not a fan of is a treatment of any kind being pushed harder than the evidence base should permit. Our patients deserve treatments that have been rigorously proven to work, whatever they may be.
Open Dialogue has been the standard psychiatric model in Western Lapland since the mid-1990s. In short, it sounds very much like a cross between a crisis team and intensive family therapy. When a referral comes in, whoever answers the phone is responsible for organising the subsequent care for that patient. The care involves a series of meetings at the patient’s home, at which 2 or 3 therapists as well as anyone else who is important in the patient’s life come together to share their experience of the crisis, develop an understanding of it and work out a plan. The team try to leave their assumptions and professional models at the door, and there is an added focus on avoiding ‘neuroleptic’ medication (antipsychotics) for as long as possible.
Their practice has 7 tenets:
1) On receiving a referral the first meeting with the patient should take place within 24 hours.
2) The social network of the patient should be invited to take part in the meetings from the outset and for as long as necessary.
3) The treatment should be adapted to the needs of the individual patient and their family, using the most suitable methods.
4) Mental health services will guarantee that certain people will take responsibility for the patient and organise a team that will make decisions with the patient and family about all aspects of treatment.
5) Psychological continuity is guaranteed; staff from different teams will be invited to participate as required.
6) A sense of security should be promoted by tolerating uncertainty, which mobilises the resources of the patient and family.
7) The team should focus on creating a dialogue between everyone in the meetings to create a new language for experiences previously unspoken.
There’s no denying it – all these principles are laudable. However, the most interesting thing that struck me about them is that we already value most of them in NHS mental health services, but struggle due to poor resourcing. They aren’t exclusive to the Open Dialogue model.
Seeing every patient in crisis within 24 hours? We’d love to do that. But in reality the more likely scene is of a nurse quietly holding back the tears of frustration as the duty phone won’t stop ringing. The NICE quality standards instruct that newly psychotic patients should be seen within 14 days and even’s that’s difficult to achieve. The Open Dialogue team clearly have enough staff and resources to be able to see people quickly, which may be an important part of why they do so well, but there’s nothing novel about the idea of early intervention.
Likewise, tailoring the treatment to the specific needs of the patient, taking their wishes and the wishes of their family into account, is also something we try hard to do, and would do better if we were simply properly resourced. Again, the NICE quality standards highlight family interventions, employment, education and support for carers and CBTp (as well as medication) as a range of choices which should add up to patient-centred care.
Similarly, continuity of care as a patient progresses through the system is a well-recognised aim for mental health professionals the world over. Sadly, the fragmentation of NHS mental health services has made this noticeably harder to achieve.
So, many facets of the Open Dialogue model are general qualities of a good service, qualities which could be achieved by improving the current systems we have rather than bringing in a whole new way of working.
However, some of the other facets of Open Dialogue are quite unique to their model. I think the aim to maintain a non-judgmental approach and foster a shared language between the patient and their family has great merit. I’m not keen on forcing people to take a certain narrative for their distress, and if they can find a way of understanding themselves that keeps them safe and well, then great. However, I do have reservations that this is suitable for all patients. Especially where I work in London, perhaps more so than in Lapland, we see an awful lot of extremely psychotic people. People whose mental states are an acute and serious risk to themselves or others, people whose experiences are plainly so extreme that to deny their pathological nature and take immediate action would be irresponsible.
This all links into the related discussion on whether psychosis is something we should be mainly trying to find meaning in, or remove. It’s not at all unusual to see a patient improve from their psychosis and come to the opinion that their psychotic beliefs were all absolute nonsense and they’re relieved to have had them medicated away instead of mused over. I wonder how Open Dialogue would address those patients.
Another issue would be that we see a lot of people who are so psychotic that they cannot engage in a series of meetings, either because they are unable or unwilling to talk, often with marked hostility or confusion. Also, many of our patients either lack any semblance of social network or possess one that actually contributes to their illness. Many would simply refuse to consent to the process, valuing their confidentiality, because they are individualists. The question mark over the cultural generalisability of Open Dialogue is huge.
Finally, I do find it hard to accept a model whose prior aim is to avoid antipsychotic medication for as long as possible. Whether you personally love them or hate them, there is a huge amount of evidence that especially in the early stages of psychosis antipsychotics are tremendously helpful for reducing symptoms and therefore distress. Their research papers (see below for links) note that at least historically they have tried to use only benzodiazepines for the first 3 weeks of treatment. This is a curious plan, as benzodiazepines expose patients to all kinds of side effects (oversedation, falls, paradoxical reactions) without any of the positives of antipsychotics.
As I said earlier, some huge claims have been made about Open Dialogue. They are essentially stating that schizophrenia, an illness that for hundreds of years no treatment (including many other talking therapies) has failed to adequately improve has been eradicated simply by a different method of talking. We owe it to our patients to examine these claims carefully.
I chose only to read the quantitative research into Open Dialogue, for the simple reason that only quantitative research could provide evidence for these claims. Furthermore, if we’re going to spend public money on Open Dialogue, quantitative research is the only thing that will prove that it works better than alternative options.
I need to say from the outset, I found the evidence base to be thin. Very, very thin. Whereas a single drug needs to be tested on hundred if not thousands of people to get anywhere near the marketplace, the amount of patients that Open Dialogue has been given to in quantitative studies appears to total 64, and the studies are of poor methodological quality.
The first study of interest is Seikkula et al. (2011a), which compares outcomes of Open Dialogue psychosis patients during two periods (1994-97 and 2003-05) with outcomes for psychosis patients seen by the team before they used the Open Dialogue model (1992-93). During this pre-Open Dialogue period the team still had a non-medication focus and the system of treatment had already been reorganised towards a needs-focused approach, but the specific Open Dialogue model had not been fully implemented. Seikkula et al. (2006) uses identical data but uses only the earlier Open Dialogue group.
The number of patients in each group was low – 46, 18 and 36 respectively – limited further by exclusions, including 9 from the Open Dialogue group that ended up with 18 patients, because patients refused to participate, could not be reached or follow-up or had died. There was no significant difference between the groups over the 2 year follow up period concerning:
• The number of patients given antipsychotics (26-50% started them across groups)
• The number of relapses (less than 30% in all groups)
• Employment status (62-78% studying or working across groups)
The mean Brief Psychiatric Rating Scale (BPRS) scores of patients in the 1994-97 group were significantly lower after 2 years than those in the 1992-3 group (30 vs 24), but this effect had disappeared by 5 years and the 2003-05 group showed no significant difference to the control at either point. More than 80% of patients in each group were free of psychotic symptoms at 2 year follow-up, but again, there was no significant difference between the Open Dialogue groups and the control. The amount of days patients spent in hospital was different between groups, with the control group (25 days) being worse than the 1994-97 group (9 days) and 2003-05 group (14 days).
So this study is the source of the statistics in the introductory quote, but overall, what these results make clear is that Open Dialogue is actually no better than the model that went before it, the simple needs-focused approach. The unusually good outcomes reported in the area must be due to factors other than the switch in model. Could it be that the team, who are dedicated, cohesive and not burnt out, get good results whatever the specifics of their model? Could it be that the area, a highly atypical one compared to the UK, has less of the stressors associated with emerging psychosis? Could it possibly be that all the patients who are going to do badly simply don’t seek out the local service or don’t engage with it, biasing their sample? And lets’ not forget that a third of the 2003-05 sample were excluded – are they very unwell?
Another paper which seems to be the source of so many large claims about Open Dialogue is Seikkula et al. (2003). As far as I can tell, this is the only time that the results of an Open Dialogue team have been compared to those of a different team. The paper uses the same Open Dialogue group from 1994-97 and the same pre-Open Dialogue comparison group from 1992-93 but adds in another comparison group – one from a neighbouring Finnish community that weren’t using Open Dialogue. They investigate the outcomes of only the patients in each group with schizophrenia, not other psychotic disorders, as only that subgroup of patients was comparable in demographics across treatment arms. This raises immediate questions – are the neighbouring community’s patients older, more unemployed, less married? In any case, the resultant groups are absolutely tiny – just 14 patients in the neighbouring community control, versus a barely better 22 and 23 in the Open Dialogue groups.
The results show that the Open Dialogue patients were in a much better state that the control group after 2 years. Only 6 out of 23 patients had relapsed compared to 10 out of 14 in the control group, 15 out of 23 were working or studying versus 3 out of 14 in the control group, and 8 of 23 had needed antipsychotics at some point compared to all 14 in the control group. Days spent in hospital were fewer in the Open Dialogue group, but interestingly BPRS scores weren’t significantly different between the groups (25 vs 27).
The limitation of this study are eye-popping. Anything as minor as the psychiatrist in the control group having a very paternal style of practice (i.e. favouring medication and longer stays in hospital) could have biased the whole thing. Heck, there could be a crack den or a genetically unfortunate family in the control area that skewed the results. The sample sizes are tiny and a group of 14 people is in no way an adequate representation of ‘treatment as usual’ in the rest of mental health services. The Open Dialogue team seemed to keep their patients off medication, out of hospital and in work but their actual levels of psychosis – the BPRS scores – weren’t any different, so are they treating psychosis or just providing awesome social support?
The final study of note is Seikkula et al (2011b). In this study, the notes of every patient aged 15-44 who presented to mental health services in Western Lapland between 1984 and 1994 are examined, the patients are retrospectively diagnosed based on those notes, and the relative rates of diagnoses between two groups (1985-89 and 1989-94) are compared to see if the rates of schizophrenia dropped after Open Dialogue was brought in. The paper states that a comparison between pre- and post-1990 was chosen as family therapy programmes began in the area in 1989. This seems to reconcile poorly with the previous papers that used the 1992-3 group as the ‘pre Open Dialogue’ group, effectively admitting that the 1992-93 group is very similar to Open Dialogue proper.
The authors found that the mean annual incidence of schizophrenia fell from 25 to 10 between the 1985-89 and 1990-94 groups. The fewest cases diagnosed in any one year was 3, in 1994. Brief reactive psychoses rose from 1 to 7 cases per year. The paper also notes that there were no patients staying longer than a year in hospital after 1992. It is from this paper, it seems, that the claim that schizophrenia has been eradicated has emerged. The authors hypothesise that so many of the local population must have been involved with their family meetings that the beneficial psychological effects may have spread, actually preventing new cases of schizophrenia emerging in the first place.
Sadly, there are several more mundane possibilities. The local economic climate may have picked up, or illicit drugs might have been better controlled. The professionals who originally saw the patients, working for a service that is so enthusiastic about reducing psychosis, may have been selective in their note taking at the time, leaving out or not fully exploring key psychotic symptoms. After all, the difference between a ‘psychotic episode’ and ‘schizophrenia’ is only a clinician’s judgement of how soon the psychosis ended. Furthermore, retrospectively diagnosing patients from their notes isn’t exactly accurate and using old notes to establish a regional incidence rate isn’t close to watertight. The claim that the Open Dialogue model has disseminated amongst the local populus, creating a psychosis-resilient social environment, is a huge one and there is no reason to believe it without intense further study. There’s not even a lot of evidence that individual interventions like medication and CBT work for people at ultra-high risk of psychosis, so it’s a real long shot hypothesis.
My thoughts on Open Dialogue can be summed up in two themes:
Firstly, their model is admirable, but many facets of it, like continuity of care and timeliness in intervention, have been recognised as key components of good care for many decades. The reason we struggle to achieve them in the NHS is because we’re so short on resources, not because we lack the Open Dialogue model. Other facets of the model are more unique but there are huge issues around their generalisablity on social, cultural and individual levels.
Secondly, the evidence which suggests that Open Dialogue works better than any other treatment is slim to none. We have no idea if it’s the general facets of the model that cause improvements (if there are any), or more specific aspects of the dialogical approach. If we’re going to make significant investments in the model in this country, using public money, we need to have more convincing data. A multi-centre UK trial is in the offing and I look forward to the results with an objective eye.