Using Average And Variability Metrics Of Accelerometer-measured Daily Movement Behaviors For Frailty Detection

Frailty is associated with major adverse health outcomes in older adults. While movement behaviors (MBs) showed promise for frailty detection, few studies examined the combined role of average and variability metrics of daily MBs. PURPOSE: This study aimed to compare frailty detection models using only average metrics versus using both average and variability metrics of daily MBs. METHODS: This cross-sectional study included 44 older adults (79.6 ± 9.3 years; 37 females) residing in senior communities. Fried Frailty Phenotype (FFP) was defined as meeting 3 or more of unintentional weight loss, exhaustion, low physical activity, slowness, and weakness. Participants wore an ActivPAL on the thigh for 10 consecutive days. The mean and standard deviation (SD) were computed for daily: 1) overall activity (activity score, daily steps, and number of sit to stand), 2) time in postures (standing, stepping, sitting, and lying), 3) time in sitting bouts >30 min, 4) steps in >1 min bouts, 5) time in cadence >75 steps/min. Random forest models were trained using age, sex, BMI, and either mean daily MBs (model 1) or both mean and SD of daily MBs (model 2) to classify FFP and its 5 components. RESULTS: Eleven (25%) participants were classified as frail. The average wear time was 9.1 ± 1.5 days. Model 1 achieved an AUC (95% CI) of 0.89 (0.77-1.00) for identifying FFP, while model 2 showed a similar AUC of 0.88 (0.75-1.00). Both models achieved comparable AUCs for identifying individual FFP components, with AUCs ranging from 0.81 to 0.94 for model 1 and 0.83 to 0.94 for model 2. However, the importance ranking indicated that variability metrics were ranked higher than average metrics in identifying unintentional weight loss, exhaustion, slowness, and weakness (Table). CONCLUSIONS: Although adding variability metrics of daily MBs did not improve the overall model performance in frailty detection, these metrics may capture unique aspects of frailty phenotypes not reflected by average daily MBs alone.