Gerontology · Apr 28, 2026
Quantifying Reactive Dynamic Balance in Older Adults: An Exploratory Lag-Plot-Based Approach toward Fall Risk Estimation
By Feld L, Hackbarth M, Himmelmann L et al.
INTRODUCTION: Reactive dynamic balance in older adults has become a focus of research and targeted falls prevention interventions. However, an accurate and reliable measure to assess an individual's reactive dynamic balance performance is lacking. While the margin of stability (MoS) is commonly used in laboratory settings, its practical applicability for fall risk assessment is limited. This study proposes exploratory perturbation-based reactive balance measures based on lag plot analysis of acceleration data from a single inertial measurement unit worn at the lower back and evaluates their predictive value for future falls. METHOD: Sixty-four older adults (aged >60 years), all of whom had recently experienced a fall resulting in an emergency department visit, participated in a treadmill-based perturbation protocol including acceleration (right/left), deceleration (right/left), full stop, sway (right/left), and pitch perturbations. Besides age, Short Physical Performance Battery (SPPB), gait speed, cognition, and concerns about falling were assessed. Reactive dynamic balance responses were quantified, using an exploratory approach with individual lag plot ellipses derived from acceleration data in mediolateral and anteroposterior directions. Four unperturbed gait variability measures were calculated per participant, and four exploratory perturbation-based parameters as well as two MoS parameters were calculated per participant and perturbation type. Stepwise logistic regressions were performed to evaluate the predictive value of this exploratory approach for prospective falls, reporting odds ratio (OR) and McFadden's pseudo-R2. RESULTS: While the SPPB showed predictive relevance (OR = 0.65, p = 0.029, R2 = 0.14), unperturbed gait variability and MoS did not improve model performance, when controlling for age. In contrast, exploratory perturbation-based parameters for the acceleration-right perturbation significantly contributed to fall prediction. In particular, an effect was observed for the parameter associated with short-term variability (LMSSD2norm; OR = 0.39, p = 0.010, R2 = 0.20). These results indicate that older adults with higher short-term (stride-to-stride) variability in trunk acceleration during right-leg acceleration perturbations are less prone to falls. When allowing for an additional predictive parameter, the proportion of acceleration data deviating from unperturbed walking during the perturbation phase (POE) showed also a predictive value (OR = 1.13, p = 0.020, R2 = 0.28). CONCLUSION: The findings provide first evidence that the proposed exploratory perturbation-based acceleration parameters, especially LMSSD2norm, may quantify aspects of reactive dynamic balance performance relevant to fall risk. These measures hold promise for integration into wearable technologies for long-term monitoring and real-world fall risk assessment.