18-1 · Patient safety · 2026-11-14
Fall Prediction and Prevention
Replay acceleration and gait records to measure warning lead time and distinguish falls from everyday movement.
Research question and target
Evaluate a warning policy on timestamped sensor streams. Separate detection of falls from prediction before impact.
- Internal target
- At least 92% early-warning sensitivity in the high-risk group
- Programme
- Patient safety
- Evidence on this record
Evaluation data
- UP-Fall Detection Dataset 2019
Sensors 19(9):1988 · 17 publisher-listed groups
Manifest locked - KFall 2021
DOI 10.3389/fnagi.2021.692865
Manifest locked - FARSEEING observed falls and eICU-CRD 2.0
2016 FARSEEING study · eICU 2.0
Blocked
Sources and eligibility
Contains enacted falls and everyday activities. Performance for spontaneous falls in hospitals requires hospital data.
Healthy volunteers performed falls and everyday activities. Verify impact time and observation coverage before each event.
Multicentre critical care records. A reference linking individual gait sensors to real fall time and impact remains unverified.
Boundary cases
- Event-level assessment of a warning at least three seconds before impact
- False alarms per hour from movements such as picking up objects or turning in bed
- Abstention with sensor displacement, device removal or missing samples
Common synthetic checks
- Information cutoff and warning lead time · 6/6 matched
What the official call asks
Clinical efficacy and field acceptance are evaluated through multi-site clinical validation, with field measures such as fall incidence, work efficiency and alarm fatigue.
Proposal requirements, printed pages 94–96 · All twelve requirements
Interpretation and next evaluation
A laboratory fall label and an advance alert for a patient are different endpoints. Measure lead time only for records that cover at least three seconds before impact.
- Next check
- Register participant and event splits, alert lead time and the time denominator for false alarms.