Using nonlinear dynamic analysis to differentiate fall status in older women.

Area of Science:

• Gerontology
• Biomechanics
• Data Science

Background:

• Falls pose a significant health risk for older adults.
• Nonlinear dynamic (NLD) analysis of gait reveals gait variability and complexity, offering insights into fall risk.
• Methodological inconsistencies in NLD analysis have hindered its clinical application.

Purpose of the Study:

• To identify Nonlinear Dynamic (NLD) measures and data sources that effectively differentiate between older adults who fall and those who do not.
• To evaluate the utility of different NLD metrics and data collection methods for fall risk assessment.

Main Methods:

• Thirty-four healthy older women (17 fallers, 17 non-fallers) underwent gait analysis on a treadmill under various conditions.
• Kinematic data were acquired using both motion capture and lower-back inertial measurement units (IMUs).
• Gait complexity and stability were quantified using Multiscale Entropy and Lyapunov Exponents (LyE), with subsequent analysis using PCA, logistic regression, and LDA.

Main Results:

• Short-term LyE (SLyE) from trunk anterior-posterior acceleration and sagittal-plane ankle angles were the most effective in distinguishing fallers.
• Receiver Operating Characteristic (ROC) analysis indicated high accuracy for ankle SLyE (AUC up to 0.88) and moderate accuracy for trunk SLyE (AUC up to 0.77).
• A Linear Discriminant Analysis (LDA) model achieved 85% cross-validated accuracy, with 82% sensitivity and 88% specificity.

Conclusions:

• Short-term Lyapunov exponent (SLyE) derived from ankle motion and trunk acceleration serve as reliable indicators of fall history in older women.
• The comparable performance of IMU and motion capture data supports the use of IMU-based NLD metrics for scalable fall risk screening.