Purpose: This study aims to assess the feasibility of using a contact-free, passive sensor system to continuously monitor falls among older adults in long-term care facility.

Background/significance: Falls are a significant health concern for adults over 60, with one-third over 65 experiencing a fall annually and 10% of these resulting in serious injury. Falls are the leading cause of injury and death in this age group, contributing to 65,000 hip fractures annually. Each year, the US treats nearly 3 million older adults for fall-related injuries, as the risk escalates with age. Among those aged 90 and older, 30% who cannot get up after a fall remain on the ground for over an hour, significantly increasing the risk of serious injuries and hospitalizations. This underscores the urgent need for effective fall monitoring and assistance solutions.

Methods: The research was conducted in two long-term care facilities in Missouri and Arizona, concentrating on the remote monitoring of older adults, including those with dementia and/or mobility aids. Sensors built by Signify were installed in the residents' rooms, including a ceiling light integrated with a radar sensor in the bedroom and a time-of-flight sensor in the bathroom. These sensors tracked falls in the environment for 6 to 9 months, and data collected was used to evaluate fall detection algorithm. We also recruited and trained healthy adults as stunt actors to perform both fall and non-fall activities, mimicking the typical movement patterns of older persons, to obtain an adequately sized fall for validation. Their training encompassed executing several scripted falls, including those from standing and seated positions and falls resulting from tripping or falling. Using these controlled simulations made sure that the fall detection algorithms were tested in a thorough and organized way, since real falls among residents might not happen often enough to be reliable. Sensitivity and specificity are reported for each group.

Results: This study enrolled 20 residents (70% female, mean age 83.46 years, SD = 8.64) who were at high risk for falls and 13 stunt actors who executed a total of 1,125 scripted activities, comprising 662 fall actions and 463 non-fall actions. The radar sensor exhibited a sensitivity of 74.42% and a specificity of 100%, based on the analysis of 650 stunt simulation actions, including both falls and non-falls. The ToF sensor demonstrated a sensitivity of 95.24% and a specificity of 100%, based on the analysis of 104 stunt simulation actions. The radar sensor demonstrated 100% sensitivity for actual falls among residents over a 6- to 9-month period, with a false alarm rate of less than one per sensor month.

Conclusions/implications: This study demonstrates that contactless sensors proficiently detect falls, attaining high sensitivity and specificity in both actual and simulated conditions. This non-invasive, continuous fall monitoring approach has potential for improving safety, lowering healthcare expenses, and enhancing outcomes for high-risk groups in long-term care contexts. Despite problems such as gradual falls, movements from low heights, and occasional device offline, improving detection algorithms can significantly augment system efficacy and reliability.

Chang-Chun Chen PhD, RN

Appearances