Related Experiment Video
Updated: Aug 6, 2026
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
Evaluating the Discriminative Validity of a Smartphone-Based Accelerometer Application During the Modified Clinical
Kanokporn Pooranawatthanakul1, Nichaporn Khumduang1, Nattawan Utoomprurkporn2
1Department of Physical Therapy, Faculty of Allied Health Sciences, Chulalongkorn University, Bangkok, Thailand.
Abstract:
Background: Objective, accessible tools are needed to screen fall risk in vestibular patients. Smartphone-based accelerometry (ACC) offers a practical, scalable solution.ObjectivesThis study evaluated the preliminary discriminative validity of smartphone ACC application during the Modified Clinical Test of Sensory Interaction on Balance (mCTSIB) to differentiate vestibular patients with and without a fall history.MethodsIn this cross-sectional study, 42 vestibular patients were classified as fallers (n = 16) or non-fallers (n = 26) based on a 12-month retrospective self-reported history. Postural sway was quantified via resultant root mean square (RMS) acceleration using a custom Android application during four mCTSIB conditions. Area under the curve (AUC) was calculated using receiver operating characteristic analysis, with DeLong's test comparing AUCs. Spearman's rank correlation assessed associations with Mini-BESTest.ResultsAUCs ranged from 0.681 to 0.798, indicating acceptable-to-good preliminary discriminatory capacity, though wide confidence intervals were present and condition AUCs did not differ statistically. The eyes closed-firm condition demonstrated high specificity (96.1%) and a strong positive likelihood ratio (LR+=14.63), suggesting utility as a highly specific "rule-in" screening task, while eyes open-foam yielded the highest AUC (0.798). Significant negative correlations occurred between application RMS metrics and total Mini-BESTest scores across all conditions (ρ = -0.410 to -0.525, p < .05), with eyes closed-foam specifically associating with reactive postural control and dynamic gait subscores.ConclusionsThe application demonstrates preliminary discriminative validity for screening fall risk within vestibular cohorts. Due to modest sensitivity, wide confidence intervals, and the study's cross-sectional nature, this tiered pathway remains strictly exploratory; prospective longitudinal validation is warranted.
