Robots in hospitals and care homes: what published trials have actually measured
Randomised trials and hospital pilots of social and service robots: what they measured, how many people took part, and what the results do and do not show.
Szerző: a humanoid4rent szerkesztősége · megjelent 2026. november 8.
Published evidence on robots in hospitals and care homes exists, but it is thinner than the marketing suggests. A 2026 meta-analysis of 34 randomised trials with 2,476 participants found a statistically significant benefit only for neuropsychiatric symptoms, and only at low certainty. Hospital pilots show delivery robots saving staff time, while humanoids such as Pepper and NAO have mostly been tested in small, short studies.
Why care settings matter beyond healthcare
Hospitals and care homes are among the few places where robots that talk to people have been studied with control groups, validated questionnaires and published methods. Event venues, trade fairs and shopping centres rarely produce that kind of data. So for anyone deciding whether visitors will accept a robot, the clinical literature is one of the more reliable reference points available.
It also shows what usually goes wrong: cramped spaces, lifts, short attention spans, missing staff time and very small samples. This article summarises recent peer-reviewed studies, gives their sample sizes and explains what the results can and cannot support.
The broad picture: 34 randomised trials pooled
The widest recent summary is a systematic review and meta-analysis by Kim and Jeon, published in the Journal of Medical Internet Research on 30 June 2026. It pooled 34 randomised controlled trials of care robots published between 2015 and 2024, with 2,476 participants in 17 countries. Sixteen trials used humanoid platforms such as NAO, Pepper, Kaspar and CommU; eleven used PARO, the robotic seal; seven used other companion or pet robots. Settings were split mainly between nursing homes and hospitals.
Of seven outcome areas pooled, only neuropsychiatric symptoms showed a statistically significant improvement, with a moderate effect size. The authors rated that finding as low certainty and noted it did not hold up in most of their sensitivity analyses. Depression, agitation, quality of life, cognition, stress or pain, and social-communication skills all showed no significant pooled effect, at very low certainty.
The authors conclude that the evidence does not yet support routine clinical use. They describe care robots as adjuncts that need a facilitator and that add to human care rather than replace it. They also point out a problem built into this research: participants always know whether they are with a robot, so blinding is impossible.
One trial in detail: CARESSES and Pepper in care homes
The EU-funded CARESSES project, which ran from January 2017 to January 2020 with about 2.08 million euros of EU funding and was coordinated by the University of Genoa, built software to let a SoftBank Pepper robot adapt its conversation to a resident's cultural background. It is a useful case because the system ran autonomously during the trial rather than being steered by a hidden operator.
The trial results were published in the International Journal of Social Robotics. According to the paper, 33 residents in nine care homes in England and Japan completed the study, split into three groups: a culturally adaptive robot (12), a basic robot (11) and care as usual (10). Each robot participant had up to 18 hours of sessions over two weeks.
Scores for emotional wellbeing on the SF-36 health survey improved significantly in the adaptive-robot group compared with care as usual. Loneliness scores improved slightly, but the change was not statistically significant. The authors describe their own support for the approach as cautious. They note that the sample was small and underpowered, and that only residents without severe dementia or frailty took part.
Children in hospital: anxiety around surgery
Paediatric care is where social humanoids have the most consistent results, though still from few studies. A systematic review and meta-analysis in the Journal of Clinical Nursing, published in November 2026, covered seven studies with 663 children and adolescents in Italy, Turkey, Spain and Canada. Robots included NAO, Pepper, Sanbot Elf and iRobiQ, mostly used as distraction before or after procedures.
Pooling two studies with 302 participants showed significantly lower anxiety after surgery. The reduction in anxiety before surgery was not significant. The authors found little or no effect on pain, fear or stress, and rated the methodological quality of the studies as low to moderate.
Humanoids at the hospital front desk
The use closest to an event setting is reception and wayfinding. The HOSPER study at the IRCCS INRCA geriatric hospital in Ancona, Italy, is testing NAO and Pepper as reception robots for 200 older visitors, 100 per robot. The robots greet people, use speech recognition to find out which ward they need and give directions. Pepper also shows a map on its chest tablet.
The protocol, published in JMIR Research Protocols in February 2026, measures perceived service quality, usability and how people rate the interaction. Recruitment ran from January to December 2025, and the authors expect to report results in 2027. Until then there are no outcome data, and the study should not be quoted as showing that the service works.
Hospital logistics: where the numbers are firmest
The strongest operational data come from robots that move things rather than talk. A six-month study at Shaanxi Provincial People's Hospital in Xi'an, China, published in Scientific Reports in April 2026, compared a fleet of ten autonomous mobile robots with manual delivery of drugs and specimens. The authors report delivery times 32 to 36 per cent shorter and 7.3 times as many delivery trips as the manual group, with fewer omissions and higher nurse and patient satisfaction.
The authors themselves note that this was a single hospital with its own infrastructure, so the figures may not transfer elsewhere. These robots are wheeled carriers, not humanoids, and their results say nothing about how people respond to a talking robot.
What staff say, and why integration is hard
A mixed-methods pilot at East Tallinn Central Hospital in Estonia ran a temi v3 robot for two weeks in summer 2024 for visitor guidance, light deliveries and corridor patrols. The paper in JMIR Nursing shows a common weakness: of 45 health care workers invited, 20 completed the first survey and only 5 the second. All five who responded afterwards considered the robot safe, but a group that small cannot support general conclusions, as the authors acknowledge.
An observational study at a large academic medical centre in New England, published in the Journal of Medical Internet Research in June 2026, explains why deployment is difficult. Across 53 hours of observation in four hospital units, researchers recorded 78 barriers to robot integration, including cramped rooms, changing light and temperature, constant interruptions and equipment that does not work together. They concluded that current robots suit auxiliary and background tasks rather than core bedside work.
Reading these studies with care
A few patterns repeat across the literature:
- Small samples. Several trials have a few dozen participants or fewer, and attrition can be severe.
- Short exposure. Two weeks is common. Novelty effects may fade, and few studies follow people for months.
- Selected participants. Trials often exclude people with severe cognitive impairment or frailty.
- No blinding. Everyone knows the robot is there, which can inflate positive ratings.
- Staff in the loop. Many interventions rely on a facilitator, so it is hard to separate the robot's effect from the person running it.
None of this means the robots do not work. It means positive headlines about robots in care should be checked against the sample size, the control group and the specific outcome measured.
What this means for event organisers
- Expect interest, not transformation. The most consistent clinical findings are modest gains in engagement, mood or anxiety in specific groups. Plan a robot as an attraction or guide that supports staff, much as hotel and hospitality pilots do.
- Wayfinding is the closest match. Reception and direction-giving, as in the HOSPER study with Pepper and NAO, resemble what a robot does at a venue entrance, but outcome data are still pending.
- Keep a person with the robot. The trials consistently rely on facilitators. Budget for an operator and plan crowd flow, using the safety planning guide for public events.
- Measure your own results honestly. Small, short samples mislead in hospitals and at events alike. Set clear measures in advance, as described in measuring the return of a robot at an event.
- Mind cameras and microphones. Social robots record speech and images. Review the privacy and GDPR points before the event; this article is general information, not legal advice.
Frequently asked questions
Do randomised trials show that social robots improve wellbeing in care homes?
Some trials report improvements, but pooled evidence is weak. A 2026 meta-analysis of 34 randomised trials with 2,476 participants found a significant benefit only for neuropsychiatric symptoms, rated low certainty. Depression, quality of life, agitation and cognition showed no significant pooled effect. Individual trials such as CARESSES found better emotional wellbeing scores, but with only 33 participants who completed the study.
Which humanoid robots have been studied most in healthcare?
SoftBank's NAO and Pepper appear most often. In the 2026 meta-analysis by Kim and Jeon, 16 of 34 trials used humanoid platforms including NAO, Pepper, Kaspar and CommU. NAO and Pepper also feature in paediatric anxiety studies and in the HOSPER reception study at an Italian geriatric hospital. Full-size walking humanoids have hardly been tested in published clinical trials.
Where do hospital robots have the strongest measured results?
In logistics rather than conversation. A six-month study at a hospital in Xi'an, China, published in Scientific Reports in 2026, reported that a fleet of ten autonomous delivery robots cut drug and specimen delivery times by 32 to 36 per cent compared with manual delivery. The authors note it was a single-centre study, so results may not transfer to other hospitals.
Why are results from robot trials in care settings often uncertain?
Samples are small, exposure is short and participants cannot be blinded because they know a robot is present. Many studies exclude people with severe dementia or frailty, and interventions usually rely on a staff facilitator. Attrition can be high: in one Estonian hospital pilot only 5 of 45 invited staff completed the follow-up survey. Reviewers therefore rate most findings as low or very low certainty.
How this article was made: written by the humanoid4rent editorial team with AI assistance; every factual statement was checked against the sources listed below on 2026-11-08. Spotted an error? Write to info@humanoid4rent.com.
Források
A cikk állításait az írás napján ezeken az oldalakon ellenőriztük.
- Care Robots as Emerging Health Technologies: Systematic Review and Meta-Analysis of Randomized Controlled Trials – Journal of Medical Internet Research (via Europe PMC)
- Culture Aware Robots and Environmental Sensor Systems for Elderly Support (CARESSES) – CORDIS, European Commission
- The CARESSES Randomised Controlled Trial: Exploring the Health-Related Impact of Culturally Competent Artificial Intelligence Embedded Into Socially Assistive Robots and Tested in Older Adult Care Homes – International Journal of Social Robotics (via Europe PMC)
- Effectiveness of Social Robots in Relieving Children's and Adolescents' Anxiety, Pain, Fear and Stress in Perioperative Nursing: A Systematic Review and Meta-Analysis – Journal of Clinical Nursing (via Europe PMC)
- Evaluating a Social Robot Reception Service in an Italian Geriatric Hospital (HOSPER Study): Protocol for a Usability Study – JMIR Research Protocols (via Europe PMC)
- Application management and effectiveness analysis of intelligent logistics robots in hospital drug and specimen delivery scenarios – Scientific Reports (via Europe PMC)
- Evaluating Social Assistive Robots in Clinical Nursing Care: Mixed Method Pilot Study on Health Care Workers' Perceptions and Adoption – JMIR Nursing (via Europe PMC)
- Evaluating Nursing Work Systems and Identifying Barriers for Robotic Technology Integration: Observational Study – Journal of Medical Internet Research (via Europe PMC)
További írások a blogról
Szeretne robotot a következő rendezvényére?
Egyszer írja le a időpontot, a helyszínt és azt, hogy mit tegyen a robot. A szolgáltatók olyan ajánlatokkal válaszolnak, amelyeket közvetlenül össze tud hasonlítani.
Ajánlatkérés küldése