What the research says about phoning someone with dementia
People ask what KindPhone is based on. The research here is thinner than any of us would like, and that applies to KindPhone as much as to anything else. This page sets out what has actually been studied, what it found, and where we are working on judgement rather than evidence.
We have read the papers rather than the summaries. Where a study is good, we say so. Where it does not reach as far as people claim, we say that too.
A familiar voice on a recording
The closest thing to real evidence for any of this is simulated presence therapy: playing a person with dementia a recording of a family member, usually made by that family, on the theory that a familiar voice is reassuring.
The Cochrane review of it found three trials, judged them low quality, found mixed results, and said it could not tell either way. One detail is worth knowing, because it tends to get skipped: in those trials the comparison was sometimes not silence but a recording of somebody else's voice. The trials could not show that a family member's voice did more good than a stranger's.
There is also a 2024 trial in the International Journal of Neuroscience, which is the one usually cited as proof. It is a real study with real results: eighty-five patients, allocated at random, and the group given recordings alongside their usual nursing did better on agitation, low mood and anxiety, while their carers reported less strain.
It is worth reading closely, though. It was all done in one place, with eighty-five people, and the group it was compared against got "routine nursing". One of the things measured was how long patients spent physically restrained, so these were people in a care setting rather than at home. And what was played to them was exactly what the name says: a recording, made by the family, played back. Nobody in that study was holding a conversation with a synthesised copy of their daughter.
We think that is an encouraging result about familiar voices in a care home. We do not think it stretches to a live, synthetic conversation with someone living alone in her own house, and we would be uneasy about anyone who said it did.
Not correcting, not arguing
The practice of going along with how someone sees the world, rather than putting them right, comes largely from Naomi Feil's validation therapy. It is widely taught, and most carers arrive at something like it on their own.
Its Cochrane review found three trials that fitted, could not combine their results, and found no clear difference. There is not enough evidence to draw a conclusion either way.
KindPhone borrows the practice anyway, because carers and clinicians find it humane, and because the alternative of arguing with someone about what year it is only makes things worse. But borrowing a practice is not the same as having evidence for it, and we would rather say so than dress it up.
The part that is well evidenced
The strongest study in this field is the Care Ecosystem trial, published in JAMA Internal Medicine in 2019. Seven hundred and eighty families, twelve months, support over the phone. Quality of life improved, there were fewer trips to A&E, and carers were less depressed and less worn down. It is good work and the result is real.
It also deserves to be described accurately: the thing on the other end of the phone was a trained human being: a care navigator who rang regularly and knew the family. That is what worked. So the best evidence in this area is really evidence for a person on the telephone.
We take that seriously, and it is why KindPhone is deliberately narrow. It is not a companion, and it is not a substitute for you. It answers the phone in the gap before you can ring back.
The UK guidance we work to
Most services in this area cite American or Canadian dementia guidance. We are a UK service, so we work to UK sources. On the question that matters most here, they are more careful than the overseas ones.
NICE guideline NG97 on dementia sets the frame: care built around the person, and carers treated as people with needs of their own, not just as a way of looking after somebody else.
The Alzheimer's Society is direct about untruths. An outright lie, it says, should only be told where anything else would cause significant physical or psychological harm. It should be treated "a bit like antipsychotic drugs: used only under very select conditions and even then only with extreme care."
What is truth? (Mental Health Foundation, funded by the Joseph Rowntree Foundation, 2016) is the most useful document we have found, because the panel included people living with dementia and family carers rather than only clinicians. It accepts that untruths can be right when they really are in someone's interests. But it draws a line at building make-believe worlds. It names things like a painted seaside wall with wave sound effects, or a bus stop where no bus will ever come. It says those should be avoided.
That line is a large part of the reason KindPhone works the way it does.
What we do because of all this
- KindPhone never claims to be a person. It does not use a family member's voice, and if someone asks who they are speaking to, it tells them.
- It does not argue. It will not correct someone about the date, or point out that she has already asked. She can ask the same question twenty times and get a kind answer twenty times.
- You decide what is true. The facts it can give and the topics it may discuss are the ones you have written down. It is not inventing a comforting world of its own.
- We say where we are unsure. An AI that composes its own sentences can get things wrong. That risk is set out on the front page and in the geeky details, not buried.
If you work in dementia care and think we have read any of this wrongly, we would like to hear from you. Better to be corrected than to be confidently wrong about somebody's mum.
Sources
- Abraha I, et al. Simulated presence therapy for dementia. Cochrane Database of Systematic Reviews, 2017. CD011882.pub2
- Duan Q, Liu X, Zhang A. Effects of simulated presence therapy on agitated behavior, cognition, and use of protective constraint among patients with senile dementia. International Journal of Neuroscience, 2024. doi:10.1080/00207454.2024.2346154
- Neal M, Barton Wright P. Validation therapy for dementia. Cochrane Database of Systematic Reviews. CD001394
- Possin KL, et al. Effect of collaborative dementia care via telephone and internet on quality of life, caregiver well-being, and health care use: the Care Ecosystem randomized clinical trial. JAMA Internal Medicine, 2019. Full text
- Dementia: assessment, management and support for people living with dementia and their carers. NICE guideline NG97. nice.org.uk/guidance/ng97
- What is truth? An inquiry about truth and lying in dementia care. Mental Health Foundation, 2016. mentalhealth.org.uk
- Is it okay to lie to someone with dementia? Alzheimer's Society. alzheimers.org.uk