Industries / Health / Healthcare / Home health
Remote monitoring
Watching a person's health and safety at home between visits with alarm buttons, worn sensors, home measuring devices and room sensors, with someone ready to act when an alert comes in.
7
Milestones achieved, from 1974 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1974
A button that calls help
The gerontologist Andrew Dibner asked in 1972 what an older person living alone should do if they needed help. In 1974 he and his wife Susan, a sociologist, founded Lifeline Systems to sell a personal emergency alarm. The company says it was "the first to provide a personal alert service in North America"; that claim is the company's own.
Credit
Andrew and Susan Dibner
2008
Home telehealth at national scale
The US Veterans Health Administration's Care Coordination/Home Telehealth program had patients send symptoms and readings from home to care coordinators. Enrollment grew from about 2,000 patients in 2003 to 31,570 in 2007; among 17,025 patients, joining was associated with 25% fewer days in hospital and 19% fewer admissions, at about $1,600 per patient per year.
Lessonthese results compare patients before and after joining, not against a randomized control group.
Credit
Adam Darkins and colleagues, VHA Office of Care Coordination
2016
A large null result
The BEAT-HF trial at six California academic centers randomized 1,437 patients who had been in hospital for heart failure. Half got home monitoring of weight, blood pressure, heart rate and symptoms plus nurse phone coaching; readmission at 180 days was 50.8% with it and 49.2% without. By day 30 only 55.4% of the monitored group were sending data on more than half of days.
Lessonsending equipment home does not help if patients stop using it.
Credit
Michael Ong and colleagues
2018
Monitoring that worked, then paid
The German TIM-HF2 trial randomized 1,571 heart failure patients to remote management by a staffed telemedical center or to usual care. The share of days lost to unplanned heart-related hospital stays or death was 4.88% with remote management against 6.64% without, and deaths from any cause were lower (hazard ratio 0.70). In December 2020 the G-BA, the committee that decides what German statutory health insurance covers, made heart failure telemonitoring a covered service.
Credit
Friedrich Koehler and colleagues, Charité Berlin; G-BA
2018
Fall detection on a watch
Apple added fall detection to the Apple Watch Series 4, using a motion sensor that measures up to 32 g, and the watch offers to call emergency services after a hard fall. In an independent 2022 lab test, 25 young adults fell from a wheelchair onto a crash pad 300 times, and a Series 5 watch detected 14 of the falls, 4.7%.
Lessona detector tuned for falls from standing can miss other kinds of falls almost entirely.
Credit
Apple (feature); Libak Abou, Laura Rice and colleagues (test)
2019
Medicare pays, then audits
From 2019 Medicare began paying separately for remote monitoring of chronic illness. In September 2024 the Office of Inspector General of the US Department of Health and Human Services (OIG) reported that use had risen sharply, that about 43% of patients monitored did not get all three required parts of the service, and that Medicare did not record who ordered it.
Lessona new billing code drew fast growth before checks were in place.
Credit
CMS (payment); HHS-OIG (oversight)
2025
Alarms fail on digital lines
The UK is moving landlines from analogue to digital, with most expected to switch by the end of January 2027, and an estimated 2 million people use telecare, alarms linked to a monitoring center. The government's February 2025 action plan reports serious incidents, including a small number of deaths, where alarms failed after the switch, because users had not been identified before their line moved or the connection failed.
Lessona safety device is only as reliable as the network it runs on.
Credit
Department of Health and Social Care and Department for Science, Innovation and Technology
Watch for
Radar sensors end long lies
A long lie is when someone stays on the floor for a long time after a fall. Vayyar reports more than 316 of its wall-mounted radar sensors in care homes, sheltered housing and private homes across Essex, England, and no long lies or fall-related hospital admissions in the past 12 months at the sites taking part; the results are attributed to Essex County Council but published by Vayyar without a method.
Achieved whena body independent of Vayyar publishes fall counts, detection rate and false alerts per sensor-day for these sites.
Claimed by
Vayyar, January 2026
Room sensors cut falls
Inspiren, which sells the AUGi room sensor, cites operator results: 48% fewer falls over 10 months at one memory-care site, 63% fewer injurious falls in three months at another, and 22% fewer falls at Aegis Living. No evaluator or method is named.
Achieved whena controlled study by an outside evaluator compares fall rates with similar sites that lack the sensor.
Claimed by
Inspiren and its customers, September 2026
Open
01
Bench test of fall detectors
A published, repeatable test of five or more consumer fall detectors, such as recent Apple, Samsung and Google watches and two alarm pendants. Professional stunt performers fall onto crash mats in scripted ways (forward, backward, sideways, slow slide, fall from a chair), with video as the record of what happened, and volunteers wear each device in daily life to count false alarms per day.
The briefCredit · unclaimed02
Open radar fall detector
An open-source detector that uses a millimetre-wave radar development board to spot falls and long periods without movement, tested by actors in three mock room layouts and over weeks of normal volunteer living to count false alarms.
The briefCredit · unclaimed
Open 01 · the brief
Bench test of fall detectors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A published, repeatable test of five or more consumer fall detectors, such as recent Apple, Samsung and Google watches and two alarm pendants. Professional stunt performers fall onto crash mats in scripted ways (forward, backward, sideways, slow slide, fall from a chair), with video as the record of what happened, and volunteers wear each device in daily life to count false alarms per day.
Why it is still open
A 2018 review of 22 real-world studies found a median of 9 participants and 17.5 falls per study and called the field "in its infancy"; a 2025 systematic review still says standard test conditions are needed to compare systems. The one independent test of a consumer watch found covered a single device in wheelchair falls.
Built with what exists
retail devices; gymnastics crash mats; two video cameras; the measures the 2018 review recommends (sensitivity, precision, false alarms per hour of wear); a speech-to-text model such as ElevenLabs Scribe v3 to log spoken trial notes (see the Note "Fewer word errors").
First step
Write the protocol and safety plan, book a gym with mats, and run a 20-fall pilot on two devices with one stunt performer.
Done looks like
At least 40 falls per fall type per device, sensitivity with 95% confidence intervals, false alarms per 24 hours from at least 30 volunteer wear-days per device, and the protocol, sensor logs and labelled video published.
Open 02 · the brief
Open radar fall detector
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open-source detector that uses a millimetre-wave radar development board to spot falls and long periods without movement, tested by actors in three mock room layouts and over weeks of normal volunteer living to count false alarms.
Why it is still open
Open code exists: the mmFall project reports 98% of 50 falls detected with 2 false alarms, but in one apartment. Commercial radar results come from the vendors (see Watch for), and no public multi-room test reporting false alarms per day was found.
Built with what exists
a Texas Instruments mmWave evaluation board; the mmFall code as a baseline; a small edge AI chip to run the model on the device (see the Note "An open compiler").
First step
Buy one board and reproduce mmFall's published result on its released data before collecting anything new.
Done looks like
At least 90% of 100 actor falls flagged within 30 seconds across three layouts, at most 1 false alarm per 24 hours over 14 days of volunteer living, an alert within 5 minutes in a staged long lie, and code and data released.