Industries / Health / Healthcare / Senior living and assisted care
Memory care
Memory care is housing and daily care for residents with dementia, often in a secure unit or a separate home, and covers how staff respond to distress, wandering and pain.
7
Milestones achieved, from 1992 to 2022.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1992
Personhood as the aim
Tom Kitwood and Kathleen Bredin argued in the journal Ageing and Society that dementia care should keep up a person's sense of self, which they called personhood, rather than only manage symptoms. Kitwood's group at the University of Bradford built Dementia Care Mapping on this idea: trained staff observe residents over a stretch of the day and use what they see to plan each person's care and the staff's training. The university still teaches the method and says England's National Institute for Health and Care Excellence (NICE) recognises it.
Credit
Tom Kitwood and Kathleen Bredin, University of Bradford
2009
England's pledge falls short
An independent review for England's Department of Health, published in October 2009 and known as the Banerjee report, found that antipsychotic drugs were used too much for people with dementia and that for most of them the risks outweighed the benefits. The government pledged to cut such use by two-thirds by November 2011. A University of Warwick study of prescribing data for 31,619 residents in 616 care homes found "no significant decline" from 2009 to 2012, and treatment that ran too long in over 77% of cases by 2012, up from 69.7% in 2009.
LessonA national pledge on its own did not change prescribing in care homes.
Credit
Sube Banerjee and the Department of Health; University of Warwick researchers
2009
A village for dementia
In December 2009 the Hogewey nursing home in Weesp, near Amsterdam, finished De Hogeweyk, a secure neighbourhood of 23 houses, each shared by six or seven residents with dementia, with streets, squares, shops, a restaurant and a theatre that residents can walk to freely. Staff wear ordinary clothes. The accounts read for this file describe the model but report no measured outcomes for residents.
hogeweyk.dementiavillage.comdementiaallianceinternational.orgen.wikipedia.org
Credit
the Hogewey management team, including Yvonne van Amerongen and Jannette Spiering; Vivium Care Group, which calls itself the initiator of the first dementia village
2017
A robot seal tested
PARO is a robot seal meant to calm and engage people with dementia. A trial in 28 care facilities in South-East Queensland, Australia, with 415 residents with dementia, compared 15-minute sessions three times a week for 10 weeks with PARO, with the same seal with its robotics switched off, and with usual care. On video, PARO improved mood and agitation more than usual care, but it beat the switched-off toy only on engagement, and a standard agitation questionnaire found no difference between the groups.
LessonA soft toy gave residents most of what the robot did.
Credit
Wendy Moyle and colleagues
2019
Antipsychotic use falls 41%
The US Centers for Medicare & Medicaid Services (CMS) and its partners in the National Partnership to Improve Dementia Care in Nursing Homes tracked the share of long-stay residents given an antipsychotic drug, leaving out residents diagnosed with schizophrenia, Huntington's disease or Tourette's syndrome, and posted each home's figure on the public Care Compare website. By CMS's count the share fell from 23.9% at the end of 2011 to 14% at the end of 2019, a fall of 41.3%.
Credit
CMS and the National Partnership
2020
Care mapping fails a trial
The EPIC trial assigned 50 care homes at random to Dementia Care Mapping or usual care and measured residents' agitation after 16 months. Among 675 residents analysed, it found no evidence of a difference: agitation scores were 2.11 points lower with mapping, within a 95% confidence range of 4.66 points lower to 0.44 points higher. Only a quarter of the homes completed more than one round of mapping.
LessonHomes short of staff time could not run the method often enough for it to work, by the authors' reading.
Credit
Claire Surr and colleagues
2022
A measure invites misdiagnosis
Because the CMS antipsychotic measure leaves out residents with schizophrenia, the US Department of Health and Human Services' Office of Inspector General (OIG) checked those diagnoses. It found that from 2015 to 2019 the number of residents reported with schizophrenia, and the number with no matching diagnosis in Medicare claims, each rose 194%, while anticonvulsant prescriptions rose and kept overall use of psychotropic drugs at about 80% of long-stay residents. A second OIG report, in March 2026, drew on 40 CMS inspections and found homes had diagnosed schizophrenia inappropriately to hide the misuse of antipsychotics and raise their star ratings.
LessonA measure that leaves out one diagnosis gives homes a reason to make that diagnosis.
Credit
HHS Office of Inspector General
Watch for
AI pain checks reach America
People with advanced dementia often cannot say they are in pain. PainChek, an Australian company, sells a phone app that uses AI to read facial expressions linked to pain, alongside a checklist that staff fill in. On 6 October 2025 the US Food and Drug Administration classified it as a new kind of Class II device, "pain assessment software in non-communicative adults", for observational assessment of non-verbal patients with dementia. In August 2026 PainChek said a US study of 105 residents in skilled nursing facilities in Iowa and New York found its scores "matched well" with the Abbey Pain Scale, an established observation chart, and gave no agreement figures. Australia's government funded up to 100,000 licences, and by PainChek's sponsored account 993 Australian aged care facilities had access by May 2021.
Achieved whena study run by people with no stake in PainChek publishes agreement figures between the app and a standard pain scale in care homes outside Australia.
Claimed by
PainChek, August 2026
A humanoid that runs games
Andromeda Robotics' humanoid Abi, which talks with residents, plays music and runs games, and which its makers say speaks up to 90 languages, began a year-long observational study with up to 25 residents of the memory care area at Eskaton Village near Sacramento.
Achieved whenthe study publishes measured effects on residents' engagement or wellbeing, or on staff work, in a peer-reviewed paper.
Claimed by
UC Davis Betty Irene Moore School of Nursing, May 2026
Open
01
Exit alarm bench test
A mock doorway, in a lab or an empty building, fitted in turn with at least five commercial exit alarms of different kinds (door contacts, pressure mats, infrared beams and radio-tag wander systems), and 200 scripted exits by actors: walking out alone, following a visitor out, pushing a wheelchair, waiting at the door before leaving. For each system, record which exits it catches, how many seconds it takes to alert, and how many false alarms it raises when staff and visitors pass normally.
The briefCredit · unclaimed02
Photo check of contrast
A phone app that photographs a finished room beside a printed colour reference card and reports the contrast between surfaces that dementia design guides ask to stand apart, such as a toilet seat against the floor, a handrail against the wall and a door against its frame. It would give the contrast as light reflectance value (LRV), the share of light a surface reflects on a scale from 0 for black to 100 for white.
The briefCredit · unclaimed
Open 01 · the brief
Exit alarm bench test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A mock doorway, in a lab or an empty building, fitted in turn with at least five commercial exit alarms of different kinds (door contacts, pressure mats, infrared beams and radio-tag wander systems), and 200 scripted exits by actors: walking out alone, following a visitor out, pushing a wheelchair, waiting at the door before leaving. For each system, record which exits it catches, how many seconds it takes to alert, and how many false alarms it raises when staff and visitors pass normally.
Why it is still open
The only bench test found, by Applegarth and colleagues in 2013, measured 14 alarm products for loudness, power use, battery life and pitch, not whether they catch real exits, found "wide performance variability" and called for standard test protocols. A University of East Anglia systematic review of 25 studies with 814 participants found "no overall robust evidence" that any intervention to prevent people with dementia exiting or getting lost works.
Built with what exists
Off-the-shelf door, mat, beam and radio-tag alarms; ceiling cameras for a time-stamped record; open-source person detection such as Ultralytics YOLO to time each crossing from the video; actors; a protocol registered before testing on the Open Science Framework.
First step
Write 20 exit scenarios with a dementia nurse educator, then pilot the timing method with two low-cost door alarms on one doorway.
Done looks like
At least five systems tested on the same 200 scripted exits and 200 ordinary passes, with the share of exits caught, the median seconds to alert and the false alarms per 100 passes for each, published with the protocol and the video.
Open 02 · the brief
Photo check of contrast
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone app that photographs a finished room beside a printed colour reference card and reports the contrast between surfaces that dementia design guides ask to stand apart, such as a toilet seat against the floor, a handrail against the wall and a door against its frame. It would give the contrast as light reflectance value (LRV), the share of light a surface reflects on a scale from 0 for black to 100 for white.
Why it is still open
The computer tools found score floor plans, not built rooms. UNSW's Plan-EAT, made with the University of Stirling and the University of Wollongong, assesses the plans of proposed or existing care facilities, and a 2024 TU Delft thesis by F. Y. J. Alsaggaf trained models on 256 floor plans to rate wayfinding, with 70 to 80% accuracy. The University of Stirling's IRIDIS app, reported in June 2017, asks users questions about a home or care setting and asks them to take photographs, but no account found says it measures contrast from the photos or checks itself against a colorimeter. No tool that measures contrast in a finished room from photos was found.
Built with what exists
A phone camera; a printed colour reference card such as the Calibrite ColorChecker; OpenCV for colour correction and measurement; a handheld colorimeter as the reference measurement.
First step
Measure 20 paint and flooring samples with a colorimeter, then photograph them in pairs under the three kinds of light.
Done looks like
On 100 surface pairs in a mock bathroom and bedroom, photographed under daylight, warm LED and fluorescent light, the app's contrast figure is within 5 LRV points of the colorimeter's for at least 90 pairs.