Supporting someone with psychosis: what the latest research says

A review of 29 studies after a first episode found relapse estimates of 28% at one year, 43% at around one and a half to two years, and 54% at three years [1]. These were pooled results from different studies, not one group followed through all three points.

A 2017 review found 58% reached remission, meaning symptoms had largely settled, in studies averaging five and a half years of follow-up. A different, overlapping set of studies found 38% met definitions of recovery combining symptoms and everyday functioning, over an average of just over seven years [2]. Definitions and results varied. These research categories do not capture everything that makes a life meaningful or predict one person's future. A relapse does not mean recovery is out of reach.

The same 29-study review found associations between relapse and not taking prescribed medication (about four times the risk), ongoing drug or alcohol problems (about three times), critical comments from carers (a little over twice), and greater difficulties before the illness began (about twice) [1]. These are findings across groups, not a way to decide why one person became unwell. They do not establish that relatives cause psychosis or relapse.

Communication is one potentially changeable part of a much larger picture. Medication concerns and substance use need support from the person’s clinicians, and families need help with their own strain. A relative cannot control all these factors or guarantee that another episode will not happen.

A 2022 analysis of 90 trials, with 10,340 participants, found most structured family interventions reduced relapse over about a year. Family psychoeducation, which helps people understand psychosis and manage it together, had the strongest estimate compared with usual care. But confidence in the different comparisons ranged from moderate to very low, so it is not a reliable ranking of what will work best for one family. The review found no clear benefit from only one or two sessions or crisis-only work; that is not proof they can never help [3].

NICE says family intervention should run for three months to a year, with at least ten planned sessions [4]. The national audit published in 2025 found variation in access to recommended care [5]. If you've only been offered a meeting or two, ask whether a full course of family intervention is available. If you're a friend, ask the team how you could be involved, with the person's agreement.

A 2021 review of 30 studies looked at the emotional atmosphere at home. Frequent critical comments from relatives roughly doubled the odds of relapse, while warmth was associated with odds about a third as high. [6]. These observational findings do not establish that relatives cause relapse. Psychosis can put the whole household under enormous strain, and difficult exchanges may reflect worry, exhaustion and a lack of support. Relatives do not cause psychosis by failing to find the right words.

The useful response is to ask for structured family intervention: space for everyone to understand what is happening, communicate and solve problems with professional help. It should support you too. You do not need to monitor every conversation or treat a setback as evidence that you have failed.

Relapses can follow changes such as disturbed sleep or feeling more suspicious. The EMPOWER study in Glasgow and Melbourne randomly allocated 73 people to an app with peer support and clinical follow-up, or to usual care. Of those with relapse data, 8 of 33 in the app group and 13 of 28 in usual care relapsed over a year [7]. The app was part of a supported programme, and this small trial tested whether a larger study was workable. It did not establish that an app alone halves relapse.

While they're well, ask whether they'd like to discuss their early signs with their team and write down what should happen if those signs appear. You can ask whether they'd like you to have a copy. A plan should make it easier to get help, rather than leave you responsible for spotting every change.

Medicines are a decision between the person and their prescriber. In the UK RADAR trial, 253 adults with recurring psychosis either continued their usual antipsychotic treatment or reduced it gradually with clinical support. Over two years, severe relapse requiring hospital admission was more common in the reduction group: 25% against 13%. The trial found no clear improvement in social functioning, its main outcome [8]. These findings concern recurring psychosis and should not be treated as a prediction for every person after a first episode.

If they're thinking of stopping, encourage them to discuss their reasons and plan any change with their prescriber. If they stop, cut down or restart smoking, or switch from cigarettes to vaping, tell the prescriber promptly. Tobacco smoke affects the processing of some medicines, so a change can require monitoring or a dose adjustment [9].

A review of 24 studies found more relapses among people who continued using cannabis after psychosis began than among those who stopped or did not use it. Hospital stays were longer compared with non-users. People who stopped did not have a clearly different relapse rate from non-users [10]. That does not mean stopping removes all risk, and these observational findings cannot prove cause.

You could ask what they notice about sleep, fear or voices when they use cannabis, without turning it into an argument. Their care team or a local drug service can help if they want to cut down.

Individual Placement and Support (IPS) places employment specialists within mental health services to help people find paid work and support them once they start [11]. An analysis of 27 trials involving people with severe mental illness, rather than psychosis alone, found employment was about twice as likely with IPS as with standard vocational support [12]. It does not guarantee a job or establish the same benefit for returning to study.

The national psychosis audit found variation in access to supported employment and education [5]. If work or a course is something they'd like, ask what employment or education support the team offers. They do not have to feel completely well before asking.

A 2023 review of 15 studies found exercise improved negative symptoms of schizophrenia, such as low motivation and reduced expression, by a moderate amount on average. The effect on hallucinations and unusual beliefs was smaller [13]. Some studies were small or of poor quality, and these averages do not identify one best exercise programme.

A walk, swim or another activity they enjoy may be worth offering if it feels manageable to them. Exercise can sit alongside treatment; it should not become another demand or a test of how hard someone is trying.

A 2023 review found that education and support for carers of people with schizophrenia could reduce burden and improve wellbeing, though evidence quality ranged from very low to moderate [14].

In a separate English trial, 407 carers were offered either COPe-support, an online programme with information and professional and peer forums, or a basic information website. There was no clear difference in wellbeing at 20 weeks, the main outcome [15]. That does not show all online support is ineffective, or prove face-to-face courses work better. Ask what local or online support is available and choose something you can realistically use.

In England, the standard is that people with a first episode start treatment from an early intervention team within two weeks of referral, and that the treatment follows NICE guidance [16]. That package includes family intervention, CBT for psychosis and help with work or study. This is a service standard, not a guarantee of the actual wait. If parts of the package haven't been offered, ask what is available and when. If they are becoming more unwell, seek help rather than waiting for the appointment.

In Open Dialogue, the team brings the person and people important to them into meetings from early in a crisis. The NHS ODDESSI trial included 494 adults with a range of mental health diagnoses. It found no clear difference from usual care in time to relapse, its main outcome, over two years. Some secondary findings favoured Open Dialogue, including hospital admissions and satisfaction with care [17]. Those results are encouraging, but this was not a trial solely of people with psychosis and it did not show that Open Dialogue prevents more relapses.

NICE allows some NHS use of these therapies while more evidence is collected, under specific conditions and with clinical support. Its guidance on AVATAR and SlowMo is for adults; separate guidance covers gameChangeVR for people aged 16 and over with psychosis and severe avoidance of going out [18][27]. Local availability varies.

gameChange uses virtual reality to practise situations outside the home. In a trial of 346 people, average benefits for avoidance and distress were small at six weeks. There was no clear overall benefit at six months, although people with more severe avoidance appeared to benefit more [19][27].

AVATAR lets someone practise responding to a digital version of a distressing voice, voiced by a therapist. In a 345-person trial, distress about voices improved at 16 weeks, but the difference from usual care was not statistically clear at 28 weeks [20].

SlowMo combines sessions with an app to help people examine worrying thoughts. Some results, including paranoia at 12 weeks, improved. However, its main measure of paranoia at 24 weeks did not show a statistically clear benefit over usual care [21]. That matters when weighing the early positive findings.

A 2025 study tested how AI models responded to mental health scenarios and found that some responses reinforced delusional ideas [22]. It tested model responses, not whether using a chatbot caused psychosis in patients, and its results concern the systems tested at that time. A chatbot is not a substitute for mental health care. If conversations with one seem to be adding to their fear or confusion, ask what has been happening and encourage them to tell a clinician.

The Mental Health Act 2025 provides for a "nominated person" chosen by the patient to replace the nearest-relative role in England and Wales. The changes are being introduced in stages [23][24]. Do not assume the new role already applies to someone's care; ask the ward or an independent mental health advocate which rules currently apply.

The University of Bristol has free guides for nearest relatives, developed with families who have held the role [25].

Next check due: 5 November 2026. Details to check: Commencement of nominated-person provisions and digital therapy guidance.