It usually begins with something small and domestic. The front door is on the latch. The good coat is gone but the wallet is on the hall table. You walk to the corner, then to the bus stop, then back, doing the arithmetic that thousands of families do alone: ten more minutes, or the phone. Until recently there was almost nothing published to tell you what usually happens next. Now there is. In a retrospective analysis of police-recorded missing incidents from two forces in England and Wales over 84 months (January 2015 to December 2021), published in Age and Ageing on 13 May 2026, researchers led by Vasiliki Orgeta at UCL examined 1,540 incidents involving 1,286 people aged 65 and over. Of those, 414 incidents, involving 339 people, were dementia-related: 26.9 per cent of the older-adult total, and the whole sample amounted to 3.5 per cent of all missing incidents the two forces recorded in the period.
The geography of these episodes is intimate rather than dramatic. People with dementia in the data were mostly reported missing from domestic settings (77.5 per cent), then care homes (13.8 per cent), public places (4.8 per cent) and hospitals (3.9 per cent). Their median age was 80 (interquartile range 75 to 84), against 73 for older people reported missing without a recorded dementia diagnosis, and 45.7 per cent were women, a higher proportion than in the non-dementia group. Around 14 per cent of people with dementia had at least one repeat episode. The headline outcome is the one families brace for: 9.4 per cent of dementia-related incidents involved significant harm, including serious injury, with one recorded death. Put the other way, the great majority of these episodes ended with someone coming home.
What predicted the bad end is where this paper earns its keep. In regression models restricted to each person's first incident, three things stayed significant for people with dementia. Each additional year of age raised the odds of harm by a factor of 1.12 (95 per cent confidence interval 1.06 to 1.20). Being reported missing at night carried odds of 2.94 (1.34 to 6.48). Being missing for more than 24 hours carried odds of 9.91, with a confidence interval running from 2.20 to 44.68, which is a way of saying the direction is clear and the size is not: the true effect could be large or enormous, and the data cannot say which. This is observational police-record work, so none of it proves that darkness or delay causes injury. Time missing is also a symptom of a search that was going badly, not only a cause of the outcome. Still, if you were designing a night to worry about, the data has drawn it for you.
The other striking column is who actually found the person. Police located 46.3 per cent of people with dementia. Members of the public found 21.3 per cent. Family or friends found 13.3 per cent, care home or health service staff 9.7 per cent, and 8.7 per cent made their own way home. That is not a minor statistical footnote for someone living with this. It means a search is a public act carried out by strangers on a bus route, a shopkeeper who recognises confusion, a dog walker at the edge of a field. It also means the person who knows the old addresses, the childhood street, the workplace of 1974, is holding information that the search depends on, and holding it at three in the morning while frightened.
Which brings us to the part nobody publishes: the shame. The same UCL group interviewed 33 people (nine carers, six police officers, twelve healthcare professionals and six researchers) for a qualitative study in Age and Ageing in October 2025, and reported five themes, among them gaps in reporting and response pathways, and a response they described as fragmented, with unclear routes through it. Orgeta's own summary of the quantitative findings was blunt: "this risk is not adequately addressed by health and social care services". Researchers reporting a nationwide Swedish police registry study in BMC Geriatrics in March 2025, which found delayed reporting among the factors associated with harm and four people found dead, noted that carers carry strain and guilt alongside the physical risk. Meanwhile police guidance in England and Wales has for years told officers that going missing should be treated as an indicator that a person may be at risk of harm, and individual forces put it plainly in their own public notices: Avon and Somerset Police, publishing during Dementia Action Week in May 2026, said that if you are worried about someone who has gone missing, always call 999 in an emergency. The gap between that guidance and the hesitation on the doorstep is where families are improvising.
There is early evidence that improvising is not the only option. An impact assessment of the Avon and Somerset Dementia Safeguarding Scheme, which incorporates the Herbert Protocol form and allocates tracking and NFC devices, was published in Innovation in Aging in February 2026, using de-identified force data up to a census date of 8 June 2023. It reports reductions in dementia-related missing incidents after registration, and found that among first episodes, people graded low risk were located faster (about 2.73 hours) than those graded high risk (about 5.39 hours). It is a before-and-after evaluation in a single force using records collected for policing rather than research, not a randomised trial, and it cannot tell you what would have happened to those same people without the scheme. Whether a form, a device, or any particular plan fits your household is a conversation for your clinician, your local force and the person living with dementia, ideally on an ordinary Tuesday afternoon rather than at midnight.