Industries / Health / Healthcare / Senior living and assisted care
Daily care
Daily care is the hands-on help residents get with washing, dressing, toileting, eating and moving, planned from a regular assessment of what each person can still do.
7
Milestones achieved, from 1991 to 2024.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1991
One assessment for every resident
The US Omnibus Budget Reconciliation Act of 1987 (OBRA) rewrote the rules for nursing homes paid by Medicare and Medicaid. Its rules took effect on 1 October 1990, and from spring 1991 every resident had to be assessed with one standard form, the Resident Assessment Instrument (RAI), whose core is the Minimum Data Set (MDS), and given a care plan built from it. Research nurses compared residents in 254 nursing homes in 10 states before (1990) and after (1993): records became more accurate, care plans more complete, indwelling urinary catheters fell from 9.8% to 7% of residents, and more residents with bowel incontinence had toileting programs.
Credit
US Congress (OBRA 1987); evaluation by C. Hawes, V. Mor and colleagues
2003
Ten residents to a house
The first Green House homes opened in 2003 in Tupelo, Mississippi. Each is a small house for about ten residents, with private rooms and bathrooms, a homelike kitchen, dining room and hearth, and a changed staffing model for nurses and aides. A two-year study compared residents of the first four houses with residents of two ordinary nursing homes: quality of life, as residents reported it, was better in the houses, and quality of care, measured from MDS data, was at least as good and better for change in daily functioning. The project counts nearly 400 such homes in 35 US states and Australia.
Credit
The Green House Project; evaluation by R. A. Kane and colleagues, University of Minnesota
2004
Lifts cut handling injuries
Lifting a resident from bed to chair by hand strains carers' backs. Six US nursing homes brought in mechanical lifts and repositioning aids in January 1998, with a zero-lift policy (no resident lifted by hand) and training, and were compared over three years before and three years after. Workers' compensation claims for resident-handling injuries fell to 0.39 times the earlier rate, the USD 158,556 spent on equipment and training was recovered in under three years, and assaults on staff during transfers fell by 72% by the compensation data.
Lessonthe equipment came with a rule and training, and the study measured the three together.
Credit
J. W. Collins and colleagues at the US National Institute for Occupational Safety and Health (NIOSH)
2013
Turning less often, same skin
Bed sores (pressure ulcers) form when the weight of a still body cuts off blood to the skin, so staff turn bedbound residents on a schedule, day and night. The TURN trial assigned 942 residents at moderate or high risk on the Braden Scale (a standard bed-sore risk score) in 20 US and 7 Canadian nursing homes, all on high-density foam mattresses, to turning every 2, 3 or 4 hours for three weeks. Only 19 (2.0%) developed a sore, all superficial, with no significant difference between schedules (2.5%, 0.6% and 3.1%).
Lessonon a good foam mattress, turning every two hours prevented no more sores than turning every four hours, a finding the authors say matters for nursing staff time and cost.
Credit
N. Bergstrom and colleagues, University of Texas Health Science Center at Houston
2014
A safety standard for robots
ISO 13482 is the international safety standard for personal care robots. Panasonic's Resyone, an electric care bed whose one half splits off and becomes a reclining wheelchair, so that a resident moves from bed to chair without being lifted, was certified to it by 2014: its designers reported the certification in a paper received by the Journal of the Robotics Society of Japan on 28 May 2014. Panasonic says Resyone was the world's first product certified to ISO 13482; that claim is the company's own.
Credit
Panasonic (Yohei Kume, Shohei Tsukada, Hideo Kawakami)
2023
Lifting robots fall short
Japan's Robot Care Project, run by the research institute AIST, aimed to cut caregivers' back injuries to zero with lifting robots, both wearable ones such as Cyberdyne's HAL exoskeleton and machines with arms such as Fuji's Hug. The anthropologist James Wright, who watched Hug's introduction at a Japanese nursing home, found such lifting aids "far from commercially successful", without significant impact on Japanese eldercare, and that staff saw the machine as getting in the way of the touch between carer and resident that they valued.
Lessona machine that saves effort can still go unused if it adds steps and comes between carer and resident.
Credit
James Wright, in Robots Won't Save Japan (Cornell University Press)
2024
Robots arrive with more staff
Economists built a panel of Japanese nursing homes recording which robots each used and for which tasks. Homes that adopted robots employed and kept more care workers rather than fewer, most clearly for non-regular care workers and for monitoring robots; staff moved toward tasks that need a human touch, and the homes recorded fewer uses of restraint and fewer pressure ulcers.
Credit
Yong Suk Lee, Toshiaki Iizuka and Karen Eggleston, NBER working paper 33116, published in Labour Economics
Watch for
A humanoid puts on socks
Waseda University's AIREC humanoid robot learned to put socks on from 12 demonstrations recorded on a mannequin's foot, then dressed the feet of 10 volunteers. By the team's own count it succeeded in 42 of 50 attempts against a background it had seen in training and 37 of 50 against a new one, where an established learning method (Action Chunking with Transformers) managed 33 and 0. Japan's Moonshot research program has since funded the AIREC project to make diaper changing, which includes turning, wiping and dressing a person, possible by robot by 2030.
Achieved whena group outside Waseda repeats the sock task at a similar success rate, or a care home trial measured by an outside party reports how often the robot completes the task.
Claimed by
Tetsuya Ogata, Shigeki Sugano and colleagues, Waseda University, May 2025 (paper revised January 2026; Moonshot project selected in fiscal year 2026)
China tests care robots
China's Ministry of Industry and Information Technology and Ministry of Civil Affairs launched a 2025 to 2027 pilot that pairs robot makers with care providers. Robots for care of disabled people and people with dementia, companionship, health and daily living must each be tried for at least six months, in at least 20 care institutions or communities, or 200 households for home products.
Achieved whenthe ministries or an outside evaluator publish which robots were tried in care institutions and what they measured.
Claimed by
the two ministries, June 2025
Open
01
Sock and stocking robot test
A repeatable test, with open data, of low-cost robot arms that pull socks and tight elastic support stockings onto mannequin legs of several sizes, trained from recorded demonstrations. No person is dressed.
The briefCredit · unclaimed02
Test set for care assessments
A public set of at least 200 made-up resident records (admission notes, therapy notes, shift reports) written by nurses, each coded for MDS Section GG, the items that score self-care and mobility, by two certified MDS coders, plus a script that scores any AI tool that drafts those codes, item by item.
The briefCredit · unclaimed03
Bench test of wetness sensors
A lab test of smart incontinence pads, which tell staff by app when a pad needs changing, run on a seated and a lying mannequin with a pump that delivers saline at set rates, comparing each alert with the weighed amount of liquid in the pad and counting false alerts when the mannequin moves with a dry pad.
The briefCredit · unclaimed
Open 01 · the brief
Sock and stocking robot test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A repeatable test, with open data, of low-cost robot arms that pull socks and tight elastic support stockings onto mannequin legs of several sizes, trained from recorded demonstrations. No person is dressed.
Why it is still open
The only published robot sock result found comes from the AIREC team testing its own research humanoid, socks only, with its builders expecting care-home use around 2030. No public test kit, shared data set or low-cost repeat was found, and no robot result for support stockings. A US patent granted to Torchward Ltd in June 2026 describes a motor-driven frame that pulls socks and compression stockings up a leg, but no published test of it was found.
Built with what exists
Hugging Face's LeRobot library (see the Note "No download"); two low-cost arms such as the SO-101 for socks and a stronger arm with a wrist force sensor for stockings; NVIDIA's GR00T 2 robot model (see the Note "Five demonstrations per task") or Runway's Praxis-1 open weights (see the Note "Open robot model weights"); shop-display mannequin legs; a force gauge.
First step
Buy two mannequin legs and an arm kit, record 50 teleoperated sock demonstrations, and train a first model.
Done looks like
On five mannequin feet from 22 cm to 28 cm long, 100 filmed attempts per garment, a success being heel seated and cuff above the ankle; success rate with a 95% confidence interval and peak pulling force per attempt; videos, recordings and code published.
Open 02 · the brief
Test set for care assessments
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of at least 200 made-up resident records (admission notes, therapy notes, shift reports) written by nurses, each coded for MDS Section GG, the items that score self-care and mobility, by two certified MDS coders, plus a script that scores any AI tool that drafts those codes, item by item.
Why it is still open
Several vendors sell AI that suggests MDS codes, among them DolphinCare, Nanonets Health, which claims 97% accuracy in classifying patients for payment, and PointClickCare. An August 2026 review of AI documentation tools for skilled nursing found no independent study of the accuracy of AI-drafted MDS codes.
Built with what exists
CMS's RAI manual for MDS 3.0; commercial AI models and small open ones that can be adapted on one graphics chip, such as Liquid AI's LFM3 (see the Note "One GPU to fine-tune"); ElevenLabs Scribe v3 to turn staged spoken shift handovers into text (see the Note "Fewer word errors").
First step
Write 20 records that cover easy and hard Section GG items and have two coders code them separately.
Done looks like
200 records with agreement between the two human coders reported per item, at least three AI tools scored against the agreed codes, and records, codes and scores published.
Open 03 · the brief
Bench test of wetness sensors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A lab test of smart incontinence pads, which tell staff by app when a pad needs changing, run on a seated and a lying mannequin with a pump that delivers saline at set rates, comparing each alert with the weighed amount of liquid in the pad and counting false alerts when the mannequin moves with a dry pad.
Why it is still open
The trials found are sponsored by the maker. In a 2025 cluster randomized trial with 108 residents in Germany and Canada, the TENA SmartCare Change Indicator did not significantly improve care time against usual care (30 against 16 minutes a day saved). No independent test of how accurately such sensors detect wetness was found.
Built with what exists
TENA SmartCare and two other makers' pad sensors from distributors; a peristaltic pump; a lab scale; saline; two mannequins; the makers' apps.
First step
Buy one sensor kit and a case of pads, and run five wetting runs to settle the protocol.
Done looks like
For each product, 30 wetting runs at three flow rates with liquid volume at alert against pad capacity, false alerts over 24 hours of dry movement, and all data published.