A smartphone-based online system for fall detection with alert notifications and contextual information of real-life falls.

Only 23 participants and 37 reported falls were available, so the performance estimate is based on a small number of events. Some diagnosis groups were very small, such as 2 with stroke and 2 with TBI, so results cannot be applied to one diagnosis. The system missed 10 falls, including cases where the phone was not carried and falls to one knee. It produced 45 false alerts, so alerts need confirmation. The detection model was trained on simulated lab falls from 17 participants, not on the real-life falls in this study. The phone had to be carried during awake hours and needed cellular signal to send alerts. Battery life averaged 12.2 hours while running the system, so it may not support all-day monitoring. The system did not record whether falls were indoors or outdoors, limiting weather interpretation. Device model and operating system were fixed, so performance on personal phones was not tested. Faller reports of activity before falls may be biased.

The supplied text states that the work was supported by the National Institute of Biomedical Imaging and Bioengineering.

Do not treat this as a proven fall-prevention or emergency alert system for patients. It detected 27 of 37 reported falls and raised 45 false alerts, so it missed some falls and may alert for non-falls.