Identifying Fall Risk Predictors by Monitoring Daily Activities at Home Using a Depth Sensor Coupled to Machine...

Area of Science:

• Gerontology
• Biomedical Engineering
• Rehabilitation Science

Background:

• Population aging presents significant human, economic, and social challenges related to fall prevention.
• Current fall-risk assessments are infrequent and often conducted only after a fall occurs.
• Home monitoring offers a potential solution to improve proactive fall prevention strategies.

Purpose of the Study:

• To identify behavioral parameters from daily activities that effectively differentiate individuals at high fall risk from those at low fall risk.
• To develop and validate an automated system for fall risk assessment using home monitoring data.

Main Methods:

• Utilized Microsoft Kinect sensors to capture daily activities of 30 rehabilitation patients.
• Extracted and analyzed behavioral parameters including gait speed, sit-stand transition time, and sitting duration.
• Employed statistical and machine learning algorithms for patient classification based on fall risk.
• Benchmarked automated assessments against established clinical tools like the Tinetti and Timed Up and Go tests.

Main Results:

• Step length, sit-stand transition, and total sitting time emerged as the most discriminative parameters for fall risk classification.
• Combining step length with sit-stand transition speed or total sitting time achieved error-less classification, matching clinician assessments.
• The developed algorithms accurately classified patients' fall risk levels.

Conclusions:

• Home monitoring systems analyzing parameters like step length and sit-stand transitions can effectively assess fall risk.
• Such systems can serve as a valuable complement to traditional clinical assessments.
• Implementing home monitoring can significantly enhance fall prevention efforts for aging populations.