Evaluation of an Automated Falls Detection Device in Nursing Home Residents.

Abstract : To determine the concordance between falls recorded using an investigational fall detection device and falls reported by nursing staff in a nursing home. Six-month prospective study. Hebrew SeniorLife nursing home units in Boston, Massachusetts. Nursing home residents with a documented history of at least one fall within 12 months before consent (N = 62, mean age 86.2 ± 8.1, 66% female). Subjects continuously wore an automated falls detection device on a pendant around their neck. The device contained triaxial accelerometers set to detect a rapid change in position that was interpreted as a fall. Healthcare staff reported daily falls, defined as unexpected events in which residents were found on the floor, and the number of these falls was compared with the number of falls recorded according to the device. Seven of 37 residents whom nursing staff found on the floor had a fall recorded according to the device (19%). The device did not identify any of the clinical fall events in 23 of the 37 fallers (62%). The device detected 17 of 89 total falls that nursing staff recorded (sensitivity 19%) within an 8-hour time window. Of 128 fall events that the device recorded, 17 were concordant with nursing reports (13%) within an 8-hour time window, and 111 (87%) were false positives. There is poor concordance between falls recorded using the investigational fall detection device and falls to the floor that nursing home staff report.