Why Some Fall Detection Misses Real Falls

The gap between staged and real falls

Most fall-detection algorithms are developed and demonstrated on simulated falls — a volunteer dropping onto a crash mat — where they can approach 100% sensitivity. But when researchers replayed 29 real-world falls recorded by wearable sensors on older adults through 13 of those published algorithms, average sensitivity collapsed to 57.0% (with a wide spread; the best reached 82.8%) (Bagalà et al., PLoS ONE, 2012). In other words, on the falls that actually happen, the typical impact-only detector missed roughly four in ten.

The cause is the mismatch this cluster keeps returning to: real older-adult falls are often low-impact. A simulated fall produces a clean, sharp jolt; a real fall — frequently a slow weight-shift or crumple — may peak far lower. In the same analysis, one real fall peaked around 2.3g, and catching real falls required dropping the impact threshold from about 3.5g to roughly 1.8g.

What actually improves detection: check what happens after

Lowering the impact threshold catches more real falls but also risks more false alarms — so the proven fix is to add a second signal: what the body does after the impact. Adding a post-fall posture check (is the person now lying down?) to an impact trigger lifted sensitivity to 97-98% with 100% specificity in one evaluation (Kangas et al., Gait & Posture, 2008). A short post-event stillness window works similarly well (Sucerquia et al., Sensors, 2018).

This is why the most honest and effective consumer approach is 'check-first': treat a possible fall as a question, not a verdict, and confirm with the person before doing anything. It tolerates a lower, more sensitive impact threshold — because a stumble or a dropped phone is resolved by the person simply tapping 'I'm OK', not by an unwanted emergency call.

Honesty is the point — no detector is perfect

The most important thing any fall-detection product can tell you is that it is best-effort, not a guarantee. It can miss a fall, and it can trigger on something that wasn't one. Stating that plainly is part of keeping you safe — it stops people relying on a single layer.

Phone fall detection is coming to the I'm Alive app in our next update (v2.3.0). When it arrives, it will detect a hard fall, check on you first with a full-screen "Are you okay?" and a countdown, and — only if you don't respond — alert the trusted contacts you chose, with your location. It never calls 911 and is not wired to a monitoring call center; help comes from your own people. It runs on your phone (no pendant or smartwatch to wear), it is opt-in and off by default, and it is best-effort, not a medical device. It is included from the Protect Me plan ($29.99/year), with location on the move on Protect Me On The Move ($39.99/year); the daily check-in is always free (Try It, $0), and there is a one-time Stay Connected lifetime upgrade at $4.99.

That honesty is exactly why the free daily check-in sits underneath detection. If a soft fall slips past the sensor, or your phone wasn't near you, a missed daily check-in still alerts your contacts. Two layers cover far more of the real-world falls than either does alone.