Investigating In-Situ Personal Health Data Queries on Smartwatches

Abstract

Smartwatches enable not only the continuous collection of but also ubiquitous access to personal health data. However, exploring this data in-situ on a smartwatch is often reserved for singular and generic metrics, without the capacity for further insight. To address our limited knowledge surrounding smartwatch data exploration, we collect and characterize desired personal health data queries from smartwatch users. We conducted a week-long study (N = 18), providing participants with an application for recording responses, containing their query and current activity related information, throughout their daily lives. From the responses, we curated a dataset of 205 natural language queries. Upon analysis, we highlight a new preemptive and proactive data insight category, an activity-based lens for data exploration, and see the desired use of a smartwatch for data exploration throughout daily life. To aid in future research and the development of smartwatch health applications, we contribute the dataset and discuss implications of our findings.

In Summary

To better understand how people want to use their smartwatch to explore health data, we conducted a week-long study with 18 participants, collecting 205 natural language queries about their health data. Our analysis revealed new ways users want to explore their data and highlighted the potential for smartwatches to offer more proactive and activity-based insights. We provide this dataset and discuss how it can help improve future smartwatch health applications.

Key Findings

• Queries were coded into 6 insight categories: Current Status and Value, Historical or Trend, Combination or Comparison, Goals or Performance, Preemptive and Proactive, and Contextul
  • Current Status or Value insight was often about more than just the simple metrics captured. We found a broader defnition of current metrics desired than that of previous work; these included heart rate zones, total values from activities throughout the day, peak values or fuctuations of metrics throughout the activity, and aggregated values such as perceived exertion
  • Contextual exploration was utilized by participants to fnd cause and efect between a range of data. While our captured queries from participants garnered similar contextual information to that found in previous work, we also note that participants were inclined to look for cause and efect using their own collected data as context (rather than simply using time, weather, location, etc.).
  • Preemptive and Proactive queries, a new form of insight brought forward from our study compared to previous works [1, 12, 13, 45], make up ~20% of our collected data from 15 of 18 participants. Our participants were looking for a wider range of infuential exploration from their smartwatch, such as to help choose a workout for the day, plan an activity based on goals, or to pick up on elements that they alone may not be able to predict.
• Queries were desired on the smartwatch throughout the day, not only before, during, and after a tracked activity (~50%). When related to a tracked activity, and in relation to our insight categories:
  • Preemptive and Proactive as well as general Goal insight was most often queried Before an activity.
  • Current Status or Value insight was most queried During an activity.
  • Current Status or Value, Historical or Trend, and Combination or Comparison insight categories were most often queried after an activity.

In More Detail

To view our raw data collected and/or the tutorial slides used during our user study please click here. Please review our full paper (linked above) for study details, methodologies, and complete results.