Self-care Activities—Event Detection

As this study aimed to determine the feasibility of using wearable cameras, the analysis was exploratory and descriptive in nature. As highlighted above, this analysis focused on only those (n=8) participants who were readmitted to hospital during the 30-day study period. Images from at least 2 days immediately prior to hospitalization (T1) were a requirement for inclusion in this analysis. The comparison period (T2) for each participant was at least 2 days the same days of the week as T1 and selected depending on when in the study the participant was hospitalized (2 weeks before hospitalization or 2 weeks after discharge).

Images from the selected days for each participant were sequentially reviewed and coded by 1 author (SA) who had received training from 2 authors (RM and RN) experienced with camera image review. Many images were similar; therefore, images were scanned until a unique event (an image capturing an event of interest) was observed.

The research team developed categories for the events of interest after reviewing published literature on self-care and lifestyle risk factors [8]. The following 9 categories were identified: (1) sedentary behaviors—sit and eat, sit and screen (eg, television and computer), and sit and other tasks (eg, doing crosswords and reading); (2) physical activity (eg, cleaning, gardening, and walking); (3) activities of daily living (ADL; eg, cooking or washing); (4) presence of packaged foods; (5) weight monitoring; (6) other activities (eg, going to church); (7) shopping; (8) driving; and (9) taking medication. An event was determined by a sequence of 5 or more images (equating to 2 minutes) except for events in categories that were not time sensitive (eg, the presence of packaged foods, weight monitoring, and taking medication); these only required 1 image.

All unique events were coded into these categories using an Excel worksheet (Microsoft Corporation), and the image reference was noted for ease of future reference. For example, the image in Figure 2, showing a participant using an iron, would be coded as ADL. Events were coded into a single category except for images displaying the presence of packaged foods, which could occur alongside an activity category (eg, ADL and cooking). The Excel worksheet contained information about the number of unique events in each category per day for each participant but did not capture time spent on the activities.

Example wearable camera image (participant using an iron).