000369095_sm_table.docx (12.89 kB)
Supplementary Material for: Wearable Sensor-Based In-Home Assessment of Gait, Balance, and Physical Activity for Discrimination of Frailty Status: Baseline Results of the Arizona Frailty Cohort Study
datasetposted on 2014-12-24, 00:00 authored by Schwenk M., Mohler J., Wendel C., D''Huyvetter K., Fain M., Taylor-Piliae R., Najafi B.
Background: Frailty is a geriatric syndrome resulting from age-related cumulative decline across multiple physiologic systems, impaired homeostatic reserve, and reduced capacity to resist stress. Based on recent estimates, 10% of community-dwelling older individuals are frail and another 41.6% are prefrail. Frail elders account for the highest health care costs in industrialized nations. Impaired physical function is a major indicator of frailty, and functional performance tests are useful for the identification of frailty. Objective instrumented assessments of physical functioning that are feasible for home frailty screening have not been adequately developed. Objective: To examine the ability of wearable sensor-based in-home assessment of gait, balance, and physical activity (PA) to discriminate between frailty levels (nonfrail, prefrail, and frail). Methods: In an observational cross-sectional study, in-home visits were completed in 125 older adults (nonfrail: n = 44, prefrail: n = 60, frail: n = 21) living in Tucson, Ariz., USA, between September 2012 and November 2013. Temporal-spatial gait parameters (speed, stride length, stride time, double support, and variability of stride velocity), postural balance (sway of hip, ankle, and center of mass), and PA (percentage of walking, standing, sitting, and lying; mean duration and variability of single walking, standing, sitting, and lying bouts) were measured in the participant's home using validated wearable sensor technology. Logistic regression was used to assess the most sensitive gait, balance, and PA variables for identifying prefrail participants (vs. nonfrail). Multinomial logistic regression was used to identify variables sensitive to discriminate between three frailty levels. Results: Gait speed (area under the curve, AUC = 0.802), hip sway (AUC = 0.734), and steps/day (AUC = 0.736) were the most sensitive parameters for the identification of prefrailty. Multinomial regression revealed that stride length (AUC = 0.857) and double support (AUC = 0.841) were the most sensitive gait parameters for discriminating between three frailty levels. Interestingly, walking bout duration variability was the most sensitive PA parameter for discriminating between three frailty levels (AUC = 0.818). No balance parameter discriminated between three frailty levels. Conclusion: Our results indicate that unique parameters derived from objective assessment of gait, balance, and PA are sensitive for the identification of prefrailty and the classification of a subject's frailty level. The present findings highlight the potential of wearable sensor technology for in-home assessment of frailty status.