Can Arts Therapies Do Harm? What Gets Reported, What Gets Noticed, and What Nobody Measures

Arts therapies are among the lower risk things offered anywhere in health care, and that is a comparison rather than a clean bill of health. People can be distressed by a session, can find something surfacing that they were not ready for, can spend money and hope on work that was never going to shift the thing they wanted shifted, and can meet somebody using a protected title they are not entitled to. None of that is common. All of it happens.

I have run singing sessions in care homes for years, and the honest reason I wanted this page written is a single afternoon. A woman in my Thursday group asked for a hymn, and by the second verse she was crying so hard that we stopped. Her daughter, who was visiting, thought I had done something wrong. I did not know for weeks whether I had. What I know now is that the answer was never mine to decide alone, and that the useful response was to tell the staff who knew her, which I did eventually and should have done that afternoon.

Where a claim about effect appears below it carries a tier (Tier 1 supported, Tier 2 promising but limited, Tier 3 practice and anecdote), explained on reading arts in health research.

What the research records about harm

Almost nothing, and the almost matters less than the nothing.

Trials in this field are built to detect benefit. Harms are rarely a named outcome, rarely defined in advance, and rarely reported when they are collected. In the reviews traced for this site, the closest thing to a harm outcome is dropout. The Cochrane review of dance movement therapy for depression, three studies and 147 participants, reported a non significant difference in dropout in one adult study, at low certainty, alongside a whole sample depression result of SMD -0.67 (95% CI -1.40 to 0.05) at very low certainty¹. That dropout line is the most safety relevant sentence in it, and all it establishes is that a similar proportion of people stopped attending in each arm.

Dropout is a blunt instrument. Somebody who leaves because the sessions upset them and somebody who leaves because the bus was cancelled appear in the same column. The correct reading is not that arts therapies have been shown to be safe. It is that the question has largely not been asked in a way that could produce an answer. Tier 3, of the nobody-has-looked-properly kind rather than the null-evidence kind, and the distinction is set out on reading arts in health research.

There is one thing the evidence does establish clearly, and it belongs here rather than in a section about benefit. Arts therapies can simply not work. The MATISSE trial randomised 417 people with established schizophrenia across four UK centres, 140 to group art therapy, 140 to activity groups and 137 to standard care, with 355 (85%) followed up at two years, and found no difference on either primary outcome: global functioning -0.9 (95% CI -3.8 to 2.1) and symptoms 0.7 (95% CI -3.1 to 4.6) against standard care². The one secondary difference ran the other way, with people referred to an activity group showing fewer positive symptoms at 24 months than those referred to art therapy, and attendance at art therapy groups was not associated with improvement³. Tier 3, of the null-controlled-evidence kind. That is a real finding about one referral, for one population, on two outcomes, and it is discussed properly on art therapy for psychosis.

The harms that are real and rarely written down

Four, and only one of them is what people imagine when they ask this question.

Distress in and after a session. Creative work gets underneath language, which is most of the argument for it, and it is also the mechanism by which it can bring something up faster than a person can hold it. A song, a smell of paint, a piece of movement, and somebody is somewhere they had not planned to be. Practitioners expect this. The skill is in noticing early, slowing down, and not leaving anybody alone with what has been opened. Difficult is not the same as harmful, and telling the two apart is the practitioner’s job.

Hope pointed at the wrong thing. Families arrive having watched a clip of a person with advanced dementia singing a whole verse, and hear that music therapy helps. The current Cochrane review of music based interventions in dementia, 30 studies and 1,720 participants randomised, found a small effect on depressive symptoms at SMD -0.23 (95% CI -0.42 to -0.04), moderate certainty, Tier 1, and moderate certainty evidence of essentially no effect on agitation at SMD -0.05 (95% CI -0.27 to 0.17), which is Tier 1, null⁴. There is also no evidence of an effect persisting four weeks after the sessions stop. A family that has been promised the agitation will improve has been set up for a disappointment nobody needed to hand them, and the setting up is the harm.

Time, money and displacement. If somebody is doing this instead of a treatment with better evidence for their problem, the loss is not the fee. That is the reason every efficacy page here is written the way it is.

The wrong person in the room. This is the preventable one. In the UK four titles are protected in law, art psychotherapist, art therapist, dramatherapist and music therapist, held on a register that listed 6,103 arts therapists in July 2026, searchable free by surname⁵. Using one of those titles without registration is a criminal offence. What is protected is the title and not the activity, so somebody can lawfully run a creative group and unlawfully call it therapy while doing it. The difference between the two halves of this field is set out on music therapy against community music, and the register check is covered on how to find a music therapist.

The harms that belong to the setting rather than the therapy

Consent, recording and the fate of the work.

In a care home, a ward or a school, the person in the session often cannot consent in the way a private client can. In England and Wales that is governed by law rather than by good intentions: capacity is assumed unless established otherwise, it is decision specific, and a decision made for somebody who lacks capacity must be in their best interests⁶. The everyday version of that is narrower than it sounds. Filming the man who sang. Passing round a photograph of somebody’s face. Putting a painting made in a session into a funding bid because it is moving. All three are done by people who mean well, and all three are the point at which a good programme causes harm. Consent and safeguarding in arts in health sets out what can and cannot be shared, and the short version is that the clip you most want to send is usually the one you should not.

The other setting harm is coercion, which rarely looks like it. A group activity in a home or a unit can drift into something people are moved into rather than asked. Sustained resistance is a communication, and a well run session reads it as one and backs off.

What to do if something has gone wrong

Say it early, to two people, in plain words.

• Tell the practitioner, that day if possible. What you saw, when it started, how long it lasted, what happened afterwards. A registered therapist has supervision precisely for this, and a good community leader will want to know.
• Tell the clinician who knows the person. Arts sessions sit alongside everything else going on, and only they can see the whole of it.
• If it is a title or conduct question, the HCPC takes concerns about registrants.
• If somebody is at risk right now, this is not a question about arts in health. It belongs with urgent clinical help, immediately.

And the thing I would say to the daughter in my Thursday group, five years late. Crying in a session is not proof that something went wrong, and it is not proof that something went right either. It is information, and the people who need it are the ones who know her. Nothing on this page is a reason to change, delay or decline any part of anybody’s treatment.