Companion: A Pilot Randomized Clinical Trial to Test an Integrated Two-Way Communication and Near-Real-Time Sensing System for Detecting and Modifying Daily Inactivity among Adults >60 Years—Design and Protocol

Figures

• 

  Figure 1. Study flow diagram.

• 

  Figure 2. Conceptual model of Companion’s technical components, their integration, and function. (A) Flow diagram of the system’s components and their integration; (B) function of the system’s components.

• 

  Figure 3. SWaN and MUSS algorithm output merging scheme for generating composite prediction labels. Algorithms outputs are first filtered by SWaN’s non-wear, sleep, and wear classifications. SWaN’s wear event classifications are assigned corresponding to MUSS classifications with lying, sitting movement, and non-wear, which are grouped under a composite “sedentary” label.

• 

  Figure 4. Companion’s personalized and interactive two-way communication, behavior mapping with motion data and context, and activity type classification. (A) Health coaches first detect major behaviors (i.e., movement [M] vs. sedentary [S]) by visualizing periods of whole-body locomotion (i.e., steps) on the ActiGraph Centrepoint platform (A.1.). This knowledge is used in conjunction with overall wrist motion summary data (MIMS) to further identify periods of non-locomotion movement (A.2.), e.g., MIMS helps health coaches identify if periods inferred as sedentary during whole-body locomotion visualization (A.1.) are truly sedentary or may involve other types of non-locomotion physical activity. Raw actigraphy data are then processed and visualized on the backend using validated machine learning activity recognition algorithms to classify the signal into selected activity types and non-wear. Such a three-pass data processing and signal visualization strategy yields a reasonable understanding of when an older adult is active and sedentary (A.3.). (B) Health coaches use interactive two-way text-based communication to enquire about the context of major detected behaviors gathered from…