Open Dialogue did not delay relapse after crisis, but admissions and satisfaction favoured it

Design

Multicentre, cluster-randomised superiority trial, assessor-masked

Population

494 adults in crisis across 30 general practice clusters in five NHS trusts

Primary outcome

Time to first relapse after recovery from the index crisis, over 2 years

Effect

Marginal HR 0.95 (95% CI 0.67-1.32); fewer admissions and re-referrals with Open Dialogue

ODDESSI was a cluster-randomised superiority trial in five NHS mental health trusts in London and southern England (Lancet Psychiatry, 26 August 2026). Thirty clusters of general practices were randomised to Open Dialogue, a service-wide model built on network meetings with the person, their family and social network and usually two consistent practitioners, or to usual functional-team care. 494 adults presenting in crisis were enrolled and followed for two years.

The primary outcome, time to first relapse after recovery from the index crisis, did not differ (marginal HR 0.95, 95% CI 0.67-1.32). On secondary outcomes, Open Dialogue was associated with lower probabilities of psychiatric admission and of re-referral to crisis or secondary services, and with better self-rated recovery, quality of life and satisfaction. Social network size and quality did not change. Serious adverse events were fewer in the Open Dialogue arm (105 vs 281), and 97% overall were judged unrelated to the intervention.

For services considering Open Dialogue, the honest reading is that it did not prevent relapse but may change how crisis care is used and experienced. Those secondary gains are worth testing further rather than treating as proven.

• Open Dialogue did not reduce relapse after crisis compared with standard community team care.
• Admissions and re-referrals were less likely with Open Dialogue, a secondary finding worth weighing in service planning.
• Involving the family and social network early in crisis care was acceptable and not associated with safety concerns.
• Continuity of the same two practitioners is the core of the model; partial adoption may not reproduce these results.
• In Indian practice, where families are often already present at crisis assessment, network meetings may be easier to start.

Why it matters

It is the first large randomised test of Open Dialogue, and it separates what the model changes (bed use, experience) from what it does not (relapse).

Don't overread it

The admission and satisfaction gains are secondary outcomes from a trial that missed its primary endpoint.

The statistics, in plain English

An HR of 0.95 with a CI of 0.67 to 1.32 means the data are compatible with a third fewer relapses or a third more, so no effect was shown. Secondary outcomes that reach significance after a negative primary result are hypotheses for the next trial, not conclusions.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.