Preliminary Feasibility and Development of a Heart Rate-Based Mobility and Activity Scale for Hospitalized Older Adults (MAS)

Background Mobility is a critical component of health status in hospitalized older adults. Adoption of routine mobility tracking in acute and subacute settings is hampered by lack of automated and standardized measurements. Advances in smartwatch technology and machine learning provide the opportunity to use heart rate (HR) and HR variability data to quantify mobility and activity. Methods In this pilot study, we recruited 30 older adults aged 65 years and older in a tertiary care geriatric ward to develop (n = 8) and validate (n = 30) the automated Mobility and Activity Scale (MAS). Twelve features based on smartwatch HR data were used in a random forest model to predict 5 activity levels (0 = sleep to 4 = walking with at least a moderate effort or > 20 minutes). We examined concurrent validity with Hierarchical Assessment of Balance and Mobility (HABAM), gait speed, and functional status, as well as discriminant validity with frailty and multimorbidity. We assessed acceptability of watch wearing for patients and care staff. Results Participants’ mean (SD) age was 86 years (8), 18 (60%) were female, and mean follow-up was 8.3 (5.2) days. Mean (SD) HABAM score was 36 (18) and gait speed was 0.53 (0.26) m/s. Across the cohort, mean (SD) MAS score was 1.2 (1.0) overall and 2.1 (0.7) for 10 most active hours of the day. MAS scores were moderately correlated with HABAM (r = 0.43 [95%CI = 0.07,0.69]) and functional status (r=-0.31 [95%CI=-0.60,0.06]), but not with gait speed (r = 0.02 [95%CI=-0.39,0.42]). MAS scores had no association with frailty or multimorbidity. Smartwatch wearing was acceptable. Conclusions Smartwatch-derived HR data may quantity hourly mobility and activity of hospitalized older adults and facilitate automated and real-time monitoring.