From Steps to Mobility Levels: Validating a Consumer-Grade Activity Monitor for Automated Recording of Patient Mobility in Hospitals.

Abstract

Patient mobility during hospitalization is essential for high-quality healthcare as mobility is linked to physical function and quality of life. The Johns Hopkins Highest Level of Mobility (JH-HLM) scale is a validated method to assess mobility in hospitalized patients. Although the JH-HLM is widely utilized, it has limitations including ceiling effects, unobserved mobility events going unrecorded, and the staff time needed to observe and document.We explored the feasibility of using a consumer-grade activity monitor (Fitbit) to predict JH-HLM scores and address these limitations.JH-HLM scores and step counts were recorded simultaneously using behavioral mapping and analyzed over 1-hour periods among inpatients. We predicted JH-HLM scores based on step counts by fitting ordinal logistic regressions, according to three categorizations of JH-HLM scores reflecting increasing mobility-granularity.We collected data for 189 patient-hours in a cohort of 20 participants. Step counts increased with higher JH-HLM mobility scores. When predicting JH-HLM scores from step counts, there was a trade-off between accuracy and mobility granularity: overall accuracy was 75% when categorizing patient-hours as immobility (JH-HLM of 1 to 5) or mobility (JH-HLM of 6 to 8); accuracy was 68% when categorizing immobility, shorter walking behavior (JH-HLM of 6 to 7), and longer walking behavior (JH-HLM of 8); accuracy was 61% when categorizing immobility and three progressively higher volumes of walking (JH-HLM of 6, 7 and 8).Step counts from the activity monitor could be used to predict whether a patient was immobile or mobile but may lack the sensitivity to accurately predict specific mobility levels.