Patient Safety

Fall Detection

Sensors, cameras, and the surprisingly hard problem of predicting a bed exit.

9 min readUpdated April 15, 20263 topics

Detection vs prediction

Fall detection — recognizing that a fall has occurred — is a solved problem with modern accelerometers and vision. Fall prediction — recognizing that a patient is about to try to stand up — is much harder, because the pre-fall behaviors are subtle and vary by patient.

Sensor comparison

SensorDetectsPredicts (bed exit)PrivacyNotes
Bed load cellsBed exit eventWeight shift precursorHighBaseline standard of care
Chair padChair exitLimitedHighSimple, cheap, high FPR
Wearable accelerometerFall after eventPoorHighCompliance dependent
RGB camera + CVFall + pre-fall poseModerateRequires privacy design
Depth camera / radarFall + poseModerateHighPreferred where imagery is sensitive

The workflow around the alert

The determining factor is not the model. It is what happens in the ten seconds after an alert fires: who receives it, on what device, and whether they can get to the room in time. Route to the assigned nurse's phone first, escalate to charge and rapid response after documented no-answer.

Fall-risk assessment tools that fold into the room's technology:

  • Morse Fall Scale (US default).
  • Hendrich II (widely used).
  • STRATIFY (UK origin).
  • Johns Hopkins Fall Risk Assessment Tool.

Standards & frameworks

  • AHRQ Preventing Falls in Hospitals toolkit.
  • CMS Hospital-Acquired Conditions — falls with injury are a non-payable condition since 2008.
  • The Joint Commission Sentinel Event Alert 55 on preventing falls.

References & further reading

  1. 1AHRQ — Preventing Falls in Hospitals: A Toolkit for Improving Quality of Care
  2. 2CMS Hospital-Acquired Conditions (HAC) list
  3. 3The Joint Commission — Sentinel Event Alert 55: Preventing falls and fall-related injuries

Help us improve

Spotted something wrong, outdated, or unclear? Let the editors know.