Care and research
Every number we put in a film, a page or a document carries the study it came from, the number of people in it, and a link you can open. Five of the six sources below are free to read in full.
We also publish what these studies do not show. That section is at the bottom and it is the more important half.
What was measured: agitation symptoms reported by care home staff using the Cohen-Mansfield Agitation Inventory, a 29 item questionnaire. The study did not measure needs, and we do not describe these behaviours as expressions of unmet need. That is a widely used interpretation, not a finding of this study.
Study: MARQUE cross-sectional study. 1,483 residents across 86 English care homes, recruited 2014 to 2015.
What was measured: the same inventory, scored above 45, which is the usual threshold for clinically significant agitation. 569 of 1,424 residents with complete data.
Worth knowing: the rate rises with dementia severity. 15% in very mild dementia, 33% in mild, 45% in moderate or severe. Forty percent is the average across the whole sample.
Study: MARQUE, as above.
What was measured: trained observers recorded what each resident was doing in 20 minute slots from 7am to 7pm across two days. 47.5% of observations were sitting or lying with no activity, or resting and sleeping. The authors' own phrase is "doing nothing at all". Some of that half is daytime sleep.
Worth knowing: residents cared for in bed were excluded, so this figure describes the more active end of the population.
Study: 46 people living with dementia in five long-term care homes in Germany. 2,760 observation points.
What was measured: the same care homes study looked at what agitation levels were associated with. Staffing ratios, time spent in activities, and the number of family visits per month were each not associated with lower agitation. Lower quality of life was.
The authors' own conclusion: "More staffing time and activities as currently provided are not associated with lower agitation levels."
Worth knowing: this is a cross-sectional study, so it cannot tell us which way the relationship runs. The authors raise two possibilities themselves. Residents who are already less agitated may be the ones approached for activities. And families may visit more often when someone is more unwell.
Study: MARQUE, as above.
What was measured: a trial across 69 UK nursing homes randomised 847 residents to either a person centred care programme or usual care. 553 were still in the study at nine months. The programme improved agitation and quality of life, and total health and social care costs were lower than usual care.
Worth knowing: the effect sizes were small, and the paper says so. The programme was three things together, not one: staff training in person centred care, 60 minutes a week of tailored activity per resident, and a review of antipsychotic medication. The programme itself cost more to deliver. The saving is in the total.
Study: WHELD cluster randomised controlled trial, 2018.
What was measured: a randomised trial gave socially isolated adults aged 75 and over a 30 minute video conversation four times a week for six months. The comparison group received a weekly 10 minute telephone check in. Among participants with mild cognitive impairment, the conversation group scored 1.75 points higher on a cognitive screening test at six months.
Worth knowing, and it matters: the conversations were held with trained human partners, not with AI. The conversation partners were deliberately rotated. 186 people were randomised, but this result rests on 31 participants with mild cognitive impairment who completed their six month assessment in person before March 2020, because the pandemic changed how testing was done. There was no difference in participants with normal cognition. The trial excluded people with dementia.
Study: I-CONECT randomised controlled trial, 2024.
What was measured: playing a recording of a familiar voice is an established approach in dementia care, called simulated presence therapy, first described in 1995. In a randomised trial, 85 patients with dementia received either routine nursing or routine nursing plus simulated presence. The group receiving simulated presence had lower agitation, lower anxiety and lower caregiver burden.
Worth knowing: this is the one source on this page that is not free to read in full, and the published abstract reports a single p value covering all of its outcomes rather than one for each. The wider evidence base for simulated presence is early. A 2020 Cochrane review of three trials and 144 participants rated the evidence very low quality and drew no conclusion about efficacy. Not everyone responds to a recording the same way.
Study: 85 patients, 40 receiving routine nursing and 45 receiving routine nursing plus simulated presence.
What this is: a seven stage scale describing cognitive decline, published in 1982 and still in common use. Stage 3 is mild cognitive impairment. Stage 6 is moderately severe dementia. Stage 7 is severe dementia, in which verbal ability is lost.
KindredMind is built for stages three through six. Stages one and two do not need it. At stage seven, a conversation companion is no longer the right thing.
These are the conclusions the studies above cannot support, and we do not make them.
Every number carries its source. If a figure appears on this site, in a film, in a document or in a conversation with a partner, it comes with the study, the year and the number of people in it.
If we cannot source it, we remove it. In September 2026 we removed two statistics from this site that we had been unable to trace to a primary source. They were plausible and they were in wide circulation. Neither was good enough.
We publish the limits. Every research review on this site carries a section on what the study did not prove, including about us.
References 1, 2, 3, 4 and 7 are free to read in full. Reference 5 is available as an abstract only.