Effectiveness of a Smartwatch App in Detecting Induced Falls: Observational Study

Area of Science:

• Gerontology
• Biomedical Engineering
• Wearable Technology

Background:

• Older adults face significant fall risks, leading to severe health consequences and increased healthcare costs.
• Existing smartwatch fall detection algorithms lack rigorous, blinded assessment and a full understanding of factors influencing their effectiveness.

Purpose of the Study:

• To evaluate the accuracy of a new fall detection algorithm integrated into a smartwatch.
• To identify factors influencing the algorithm's performance in detecting simulated falls.

Main Methods:

• A cross-sectional study involved 22 healthy adults performing induced forward, backward, left, and right falls and near falls.
• A novel smartwatch algorithm's detection was compared against a gold standard video recording by a blinded assessor.
• Data were collected using three different smartwatches across two operating systems, analyzing 226 distinct falls.

Main Results:

• The smartwatch algorithm achieved 77% sensitivity and 99% specificity for falls, with 89% overall accuracy.
• For near falls, sensitivity was 89% and specificity 100%, yielding 93% accuracy.
• Detection accuracy was higher when falls occurred on the same side as the smartwatch wrist.

Conclusions:

• The smartwatch algorithm demonstrates high accuracy in detecting falls and near falls, though effectiveness is influenced by fall direction.
• While some devices showed trends toward better sensitivity, results were not statistically significant.
• Findings provide a benchmark for similar smartwatch fall detection technologies.