Evidence Review · June 2026

Non-Wearable AI Fall Detection: Technology, Evidence and International Adoption

A review of how privacy-first AI fall detection works, the demographic and clinical evidence behind it, how health systems and governments internationally are trialling and adopting it, and the implications for Irish health and social care.

← Back to the Technology Explained hub

Contents

  1. Summary
  2. How AI machine vision detects a fall
  3. The demographic driver
  4. The evidence: why call-button alarms underperform
  5. International government and health-service adoption
  6. Market trajectory
  7. Implications for Irish health and social care
  8. References

Summary

For decades, the standard technology for protecting an older person living alone has been the personal alarm pendant: a button worn on the body and pressed in an emergency. A consistent body of evidence shows that this model fails precisely when it is most needed, because a person who has fallen is frequently unable to press it.4 A newer category of technology addresses this directly: artificial intelligence that recognises a fall automatically, from how a body moves, without requiring the person to wear or activate anything.

This review describes how that technology works, summarises the demographic and clinical evidence behind it, and documents where governments and public health services internationally are now funding, trialling and deploying it. It concludes with implications for Irish health and social care. The central finding is that privacy-preserving, non-wearable AI fall detection has moved from a consumer novelty to an approach being adopted by public health systems, and that the designs being favoured share a common principle: detect the event without identifying the person or retaining any image.

How AI machine vision detects a fall

"Machine vision" in this context does not mean a surveillance camera. The intelligence is not in recording an image, it is in interpreting movement. An AI model trained on large volumes of annotated examples learns to distinguish ordinary activity (sitting, lying down, walking) from a fall, and to recognise when a person is on the floor and not getting up.

In privacy-first implementations, no watchable video is ever produced. The system reduces every person in view to an anonymous moving skeleton (a "stick figure") and operates only on that abstraction; there is no face and no identifying detail. Processing is performed locally on the device, and the only data that leaves the home is an alert. Different products achieve this with different hardware, optical or infrared sensors, millimetre-wave radar (which produces no image of any kind), or under-mattress and ambient sensors, but the shared design choices are an on-device AI layer and the deliberate non-retention of identifiable imagery.

How privacy-first AI fall detection works The model reads movement, not images.
Four steps of privacy-first AI fall detection Step 1: an AI sensor monitors the room. Step 2: an AI model reads body position and motion. Step 3: every person is reduced to an anonymous stick figure, with no video kept. Step 4: a fall is recognised and an alert is sent. No footage is recorded, transmitted, or shown to anyone. 1 · Sense An AI sensor monitors the room 2 · Interpret AI reads body position and motion 3 · Anonymise Each person becomes a stick figure 4 · Alert Fall recognised, responder notified No video is recorded, transmitted, or shown to anyone. Only an alert leaves the home.

Key point: the clinical value derives from the AI interpreting movement, not from retaining a picture. This is what allows immediate detection while preserving the dignity and privacy of the monitored person.

The demographic driver

The underlying driver is global population ageing. The World Health Organization reports that the number of people aged 60 and over will roughly double between 2020 and 2050, from 1 billion to 2.1 billion, and that by 2030 one in six people worldwide will be over 60.1 The population aged 80 and over is projected to triple, to approximately 426 million.1

Global population aged 60 and over Source: World Health Organization, Ageing and Health (2024), drawing on UN World Population Prospects
Global population aged 60 and over, 2020 to 2050 In 2020 there were 1 billion people aged 60 and over. By 2030 this rises to 1.4 billion. By 2050 it reaches 2.1 billion. 2020 1.0 billion 2030 1.4 billion 2050 2.1 billion

Implication: a larger older population, a shrinking working-age carer pool, and a strong preference for ageing at home together create sustained demand for technology that supports safe independent living. AARP reported in 2024 that 75% of over-50s wish to remain in their current home as they age.5

The evidence: why call-button alarms underperform

Falls are a major cause of injury and death in older adults. The WHO estimates approximately 684,000 fatal falls occur globally each year, making falls the second leading cause of death from unintentional injury.2 Between one in four and one in three people aged over 65 fall each year.23

The limitation of the traditional pendant alarm is well documented. A Cambridge cohort study published in the British Medical Journal in 2008 examined what happens when older people fall while alone. It found that in 80% of falls where the person was alone, the call alarm was not used to summon help; and in 97% of "long lies" (an hour or more on the floor), no alarm was activated.4

80% / 97% Proportion of falls (person alone) in which the call alarm was not used, and proportion of "long lies" in which no alarm was pressed, respectively. (Fleming & Brayne, BMJ, 2008)4

The "long lie", the period spent on the floor undiscovered after a fall, is independently associated with worse outcomes, including higher rates of hospital admission and subsequent admission to long-term care.4 A device that depends on a conscious, capable person activating it therefore fails in exactly the circumstances, disorientation, injury, loss of consciousness, that define a serious fall. This is the central rationale for automatic, non-wearable detection.

International government and health-service adoption

Adoption is no longer confined to private consumers. Health services, research agencies and municipal governments are funding pilots, amending policy, and deploying these systems at scale. Selected recent examples are summarised below.

Spain

Valencia and Madrid

The regional government of Valencia (Generalitat Valenciana), with European Union co-funding, is running a pilot of a computer-vision system ("Verif-AI") that detects falls and changes in posture from a camera feed without recording any images, evaluated on the night shift in two care homes.6 Separately, Madrid City Council operates one of Europe's largest advanced-telecare programmes, including approximately 24,000 fall detectors and an AI layer that learns each household's routine, with a predictive-AI pilot in 200 homes (sensor- and wearable-based rather than vision-based).7

United Kingdom

NHS deployment and national policy

In England, a vision-based monitoring system (Oxevision, Oxehealth) using an infrared camera and AI is in use across approximately half of NHS mental-health trusts; it measures pulse and breathing contactlessly and flags falls without providing a continuous viewable video feed. An economic evaluation across five NHS trusts reported a 48% reduction in bedroom falls on older-adult wards, with substantial reductions in fall-related emergency attendances.8 At national level, NHS England announced in 2025 a nationwide rollout of an AI tool to predict and help prevent falls in home care.9

Japan

Policy incentives for monitoring technology

Japan, the world's most-aged society, designates "monitoring" as a priority field for long-term-care technology through joint policy of its economy and health ministries (METI and MHLW), and a 2021 reform of long-term-care insurance permits care homes using watch-over sensors to adjust night staffing.10 Commercial systems for this market include ceiling-mounted near-infrared camera-and-AI units for bed-exit and fall detection.

United States

Federally-funded research and peer-reviewed outcomes

In the United States, a camera-and-AI fall-detection system (SafelyYou), developed with funding from the National Institute on Aging, is deployed across 34 states. A peer-reviewed study of 11 memory-care communities reported a 41% reduction in falls and a 69% reduction in fall-related emergency-department visits; the system retains video only in the short window around a detected event.11

Singapore

National housing and health deployment

Singapore's Housing & Development Board, with the Ministry of Health and Agency for Integrated Care, has piloted and rolled out in-home fall detection, subsidised by up to 80%. Notably, Singapore selected privacy-preserving radar and similar non-image sensors rather than cameras.12

Wider context

Computer-vision fall-detection systems are deployed in care homes in the Netherlands and Belgium with public research backing; South Korea launched a national AI-care strategy in 2026; and Australia, following its Royal Commission into aged care, funded a national project with its science agency to predict falls from ambient sensors. Within the European Union, the AI Act15 and the General Data Protection Regulation together shape permissible designs: a system that detects a fall without identifying the individual biometrically and without retaining footage is more readily compliant, which reinforces the trend toward anonymised, on-device processing.

Market trajectory

Investment mirrors the policy direction. The dedicated fall-detection systems market was valued at approximately USD 447 million in 2023, projected to reach approximately USD 748 million by 2030 (a CAGR of about 8%).13 The broader categories within which AI fall detection sits are growing considerably faster.

Projected annual growth by segment (CAGR to 2030) Sources: Grand View Research (fall detection, ambient assisted living); MarketsandMarkets (remote patient monitoring)
Projected compound annual growth rate by market segment to 2030 Ambient assisted living is projected to grow at roughly 27 percent a year. Remote patient monitoring at roughly 13 percent. Fall-detection systems at roughly 8 percent. Ambient assisted living to roughly $38bn by 2030 26.8% Remote patient monitoring to roughly $57bn by 2030 12.7% Fall detection systems to roughly $748m by 2030 7.7%

Observation: ambient assisted living (~27% CAGR) and remote patient monitoring (~13% CAGR) are growing several times faster than the narrow fall-alarm category, indicating that investment is consolidating around AI-enabled, in-home monitoring.1314

Implications for Irish health and social care

Ireland exhibits the same demographic trajectory and the same stated policy preference for ageing in place that underpin international adoption. The evidence that pendant alarms are not activated in the majority of unwitnessed falls applies equally in an Irish setting.4

Several considerations follow for Irish health and social care stakeholders:

  • Evidence base. Non-wearable AI fall detection is supported by peer-reviewed outcome data811 and is being adopted by comparable public health systems, notably the NHS. It is a candidate for structured evaluation in Irish community and residential care settings.
  • Privacy and compliance. Designs that detect a fall without identifying the individual and without retaining footage align with GDPR and the emerging EU AI Act framework,15 reducing data-protection risk relative to conventional camera monitoring.
  • Equity and the analogue-to-digital transition. As legacy analogue telecare is retired internationally, there is an opportunity to specify digital, AI-capable monitoring in its place rather than replicating button-based systems with known limitations.
  • Deployment model. International examples span home settings (Spain, Singapore) and institutional care (UK, Japan, US), indicating applicability across the continuum from community dwelling to residential care.

Note on sources and claims. Demographic and falls figures are drawn from the World Health Organization and US CDC. Outcome figures cited for specific systems (Oxevision; SafelyYou) are from a peer-reviewed economic evaluation and a peer-reviewed study respectively, and are reproduced as reported; some involve authors or data affiliated with the developers, and should be interpreted accordingly. Market figures are third-party analyst projections and vary by scope. This document describes publicly reported deployments and does not constitute an endorsement of any specific product.

References

  1. World Health Organization. Ageing and health (fact sheet), 2024. who.int/news-room/fact-sheets/detail/ageing-and-health
  2. World Health Organization. Falls (fact sheet), 2021. who.int/news-room/fact-sheets/detail/falls
  3. US Centers for Disease Control and Prevention. Older Adult Falls Data, 2026. cdc.gov/falls/data-research
  4. Fleming J, Brayne C. Inability to get up after falling, subsequent time on floor, and summoning help: prospective cohort study in people over 90. BMJ, 2008;337:a2227. pmc.ncbi.nlm.nih.gov/articles/PMC2590903
  5. AARP. 2024 Home and Community Preferences Survey, 2024. aarp.org/pri/topics/livable-communities/housing
  6. Generalitat Valenciana. Verif-AI pilot in residential care (Ivace+i / EU FEDER), 2026. comunica.gva.es
  7. Ayuntamiento de Madrid. Teleasistencia Avanzada programme, 2025. madrid.es
  8. Oxehealth / NHS trusts. Economic evaluation of vision-based monitoring (Oxevision). PLOS Digital Health, 2024. journals.plos.org/digitalhealth
  9. NHS England. Nationwide roll-out of AI tool that predicts falls, 2025. england.nhs.uk
  10. Ministry of Economy, Trade and Industry (METI) & Ministry of Health, Labour and Welfare (MHLW), Japan. Priority Fields in the Use of Technologies for Long-term Care, 2024. meti.go.jp/english/press/2024/0628_004.html
  11. US National Institute on Aging-funded study. Real-time video detection of falls in dementia-care facilities. JMIR Aging, 2021. ncbi.nlm.nih.gov/pmc/articles/PMC8277400
  12. Singapore Housing & Development Board. Assistive Living Technologies. hdb.gov.sg
  13. Grand View Research. Fall Detection Systems Market; Ambient Assisted Living Market. grandviewresearch.com
  14. MarketsandMarkets. Remote Patient Monitoring Market, to USD 56.94bn by 2030. marketsandmarkets via PR Newswire
  15. European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act), in force 1 August 2024. artificialintelligenceact.eu

Prepared by SmartCare Living as an evidence review for discussion with health and social care stakeholders, June 2026. Figures are reproduced from the cited public sources. This document is provided for information and does not constitute clinical, legal or procurement advice.