Fall detection: what actually works, and what it cannot do

Fall detection is the first thing families ask about, and it is the area where marketing has run furthest ahead of what the technology reliably does. That is worth addressing directly, because a family that believes a device will always detect a fall may make worse decisions than one that understands its limits.

Falls matter enormously. Among older adults they are a leading cause of injury, and the outcome frequently depends less on the fall itself than on how long someone lies there afterwards. Reducing that interval is a genuine and worthwhile goal. The honest question is which technologies actually reduce it.

The three approaches, and how they really compare

Wearable accelerometers

A pendant, wristband or watch detects the acceleration signature of a fall and raises an alert, usually with a cancel window in case of a false positive.

Strengths: works anywhere including outdoors and away from home, and can combine automatic detection with a manual button, which is often the more reliable path.

Real limitations: it must be worn to work, and adherence is the single biggest failure — a pendant on a bedside table protects nobody, and bedside tables are where pendants end up. Detection is imperfect in both directions: a slow slide down a wall to the floor may not register as a fall, and vigorous activity sometimes does. It needs charging. And many people dislike wearing something that marks them as frail, which is a real objection rather than vanity.

Radar and mmWave sensors

Ceiling or wall-mounted sensors that detect the position and posture of a person using millimetre-wave radar, and infer a fall from a rapid change to a floor-level position that persists.

Strengths: nothing to wear, nothing to charge, nothing to remember. It works in a bathroom, where falls are common and cameras are unacceptable, because radar produces no image. Coverage is continuous within the room.

Real limitations: it only covers the rooms where you fit sensors, so it is per-room cost and complete outside the home is impossible. Performance varies considerably between products, and independent evaluation is scarce. Furniture, pets and multiple occupants complicate detection, and some units need careful zone configuration to avoid false alerts from a hallway or a corridor visible through a doorway.

Floor and bed sensors

Pressure mats beside a bed, or bed occupancy sensors, which do not detect the fall itself but detect the circumstances around it — someone who got out of bed and did not return.

Strengths: simple, cheap, very reliable at what they do, and completely non-intrusive. Night-time is when a large share of falls happen, and this addresses that window well.

Real limitations: narrow scope. It tells you someone left the bed and has not come back; it cannot tell you what happened.

What about ordinary smartwatches?

Several mainstream smartwatches include fall detection, and for someone who already wears one daily this is a reasonable option — the adherence problem solves itself, because the watch is worn for other reasons. Two caveats. Detection on general-purpose watches is tuned conservatively to avoid annoying a mostly younger user base, so it may be less sensitive than a purpose-built device. And the emergency call typically goes to the emergency services rather than to family, which is right for a serious fall and heavy-handed for a stumble. Check what the alert actually does and who receives it before relying on it.

Preventing falls beats detecting them

This is the part that gets least attention and delivers most benefit. Detection shortens the time on the floor; prevention avoids the fall. Several of the most effective measures are lighting automations that cost very little.

Night-time lighting

A large proportion of falls at home happen at night, on the route between bed and bathroom, in the dark, in the first disoriented seconds after waking. Motion-triggered low-level lighting along that route — dim, warm, no bright overhead light — is arguably the single highest-value intervention discussed anywhere in this category. It preserves night vision, does not fully wake the person, and removes the hazard rather than reporting it.

Consistent, adequate daytime light

Ageing eyes need considerably more light to see an edge or a step clearly. Automating lights to reach a proper level in the late afternoon, rather than relying on someone noticing it has grown dim, addresses a hazard that builds gradually and is easy to miss.

Stairs

Motion-triggered lighting at both ends of a staircase, with a longer timeout than you would use in a hallway. A stairway that goes dark halfway up is a serious hazard, and this is the one place where a generous timeout is unambiguously correct.

Remove the reasons to move in the dark

A bedside control for lights and heating, a water carafe within reach, a phone that can be answered without getting up. Every removed night-time journey is a removed opportunity to fall.

What happens after an alert

A detection system with no defined response is theatre. Decide these things before installing anything.

• Who is contacted, in what order? Two people minimum. One person is asleep, driving or abroad.
• How does the person cancel a false alarm? There must be an easy way, or false positives will lead to the system being switched off.
• Can somebody actually get in? This is the practical detail most often missed. If a responder arrives and cannot enter, the alert achieved nothing. A key safe with a code held by two people, or a smart lock with a time-limited code for a responder, solves it.
• Is there a professional response option? Family-based response works well where family is nearby and available. Where it is not, a monitored service with trained operators is a different and often more appropriate product, and worth pricing.

What we would actually do

If we were setting this up for a relative, in this order:

1. Night-time route lighting first. Cheapest, least intrusive, and it prevents rather than reports.
2. A bed occupancy sensor with an alert if someone is out of bed for an unusually long period overnight.
3. A manual call button or two — wall-mounted in the bathroom and by the bed. Simple, understood, and it works for every emergency, not only falls.
4. A key safe, so anyone responding can get in.
5. Then either a radar sensor in the highest-risk room, or a wearable if the person is genuinely willing to wear it — not if they merely agreed to stop the conversation.

Notice that four of those five are cheap, and none of them requires anyone to trust a detection algorithm. Which is rather the point: the reliable parts of this are unglamorous, and the glamorous part is the least reliable.

Falls are also a medical matter as much as a technical one. Balance, medication side effects, blood pressure, vision and footwear all contribute, and several are treatable. A conversation with a GP or a falls clinic will frequently do more than any sensor, and is the right first step if falls have already started happening.