Fall history: Is a minimum setup quantification possible?

Introduction: Falls in the elderly represent a major issue in today's society, as confirmed in the Horizon 2020 European Programme, which puts major emphasis on active ageing and independent living. Questionnaires about risk factors for falling are commonly used for fall risk assessment in the clinic, but their predictive value is limited [1,2]. Objective methods, suitable for clinical application, are hence needed to obtain a subject specific quantitative assessment of fall risk. Since falls in older adults often occur during walking and the trunk is known to play a critical role in gait, analysis of trunk kinematics during gait could represent a viable solution to the development of such methods. In this study, nonlinear measures such as harmonic ratio (HR), index of harmonicity (IH), multiscale entropy (MSE) and recurrence quantification (RQA) of trunk accelerations were calculated. These measures are not dependent on step detection, hence excluding an important source of error. The aim of the present study was to investigate the association between the aforementioned measures and fall history in a large sample of subjects (fallers and non-fallers) aged 50 or more. Methods: 131 subjects (62.4 ± 6.1 years old) participated in the study. Participants walked on a treadmill at 4 km/h for 12–17 min, wearing a single inertial sensor located on the trunk, below the shoulder blades. Data of 3 min of consecutive walking were acquired. Fall history was obtained by supervised self-report; a subject was classified as a faller if at least one fall had occurred in the 12 months prior to the experiment. Accelerations of the trunk in the anterior–posterior (AP) and medio-lateral (ML) directions were analyzed. Measures of smoothness (HR, IH), complexity (MSE) and regularity (RQA) were calculated according to established techniques [3–7]. MSE was implemented for scale factors from 1 to 6. For RQA, recurrence ratio (RR), determinism (DET), average diagonal length (avg_length) and maximum diagonal length (max_length) were calculated. Log transformed parameters were then used as inputs for backward stepwise logistic regression models, to classify subjects as fallers or non-fallers. Results: 42 subjects (32.3%) had at least one fall in the previous year. The majority of falls had occurred during locomotion. HR and IH parameters showed poor or no correlation with fall history. MSE for scale factor 3 and RQA max_length in the AP direction showed association with fall history. With MSE, the model correctly classified 71.8% of the subjects (sensitivity 21.4%, specificity 95.5%), while with max_length the model correctly classified 71% of the subjects (sensitivity 16.7%, specificity 96.6%). Discussion: Complexity (MSE) and regularity (RQA max_length) of trunk acceleration in the AP direction during walking are associated with fall history; subjects with high complexity or low regularity were found to be more likely to have experienced a fall in the previous year. Conversely, HR and IH showed poor or no correlations with fall history; analysis in the frequency domain was hence not be useful to this aim. In conclusion, MSE and RQA could represent useful measures for assessing fall risk without requiring step detection. Future research should address the physiological implications of complexity and regularity of trunk kinematics, to try to better understand the mechanisms behind fall risk and plan appropriate therapy and rehabilitation.