A clinical study to assess fall risk using a single waist accelerometer

Abstract

Falls have various causes and are often associated with mobility impairments. Preventive steps to avoid falls may be initiated, if an increasing fall risk could be detected in time. The objective of this article is to identify an automated sensor-based method to determine fall risk of patients based on objectively measured gait parameters. One hundred fifty-one healthy subjects and 90 subjects at risk of falling were measured during a Timed 'Up & Go' test with a single triaxial acceleration sensor worn on a waist belt. The fall risk was assessed using the STRATIFY score. A decision tree induction algorithm was used to distinguish between subjects with high and low risk using the determined gait parameters. The results of the risk classification produce an overall accuracy of 90.4% in relation to STRATIFY score. The sensitivity amount to 89.4%, the specificity to 91.0% and the reliability parameter kappa equals 0.79. The method presented is able to distinguish between subjects with high and low fall risk. It is unobtrusive and therefore may be applied over extended time periods. A subsequent study is needed to confirm the model's suitability for data recorded in patients' everyday lives.

Phone Based Fall Risk Prediction

“…Gietzelt et al (2009) have generated a decision tree to identify high risk of falling based solely on two parameters, extracted from the acceleration signals of lower back trunk (see Figure 2.24) [59]. These parameters were kinetic energy and pelvic sway.…”

“…The pelvic sway was defined as the mean transversal acceleration deviation during walking [59]. The study of Tura et al (2010) has focused on the use of a single accelerometer to measure the gait symmetry and regularity, which are important parameters to consider for gait analysis [77].…”

“…where a is the magnitude length of the measured acceleration vector [59]. The pelvic sway was defined as the mean transversal acceleration deviation during walking [59].…”

“…Other measures were also calculated based on the description provided in [59], i.e. pelvic sway and kinetic energy.…”

“…This can be done through the quantification of the same parameters used in assessment scales, like the time to complete the test in TUGT (see [7]), quantification of mobility or gait parameters associated with fall risk (cf. [55], [56], [57], [58] and [59]) or using machine learning techniques, with the ground truth provided by clinically accepted scales (cf. [47], [32], [48], [6] and as suggested in [50]).…”