Addressing Data Quality Challenges in Observational Ambulatory Studies: Analysis, Methodologies and Practical Solutions for Wrist-worn Wearable Monitoring

TL;DR

An in-depth exploration of data analysis challenges tied to wrist-worn wearables and ambulatory label acquisition, using two real-world datasets (i.e., mBrain21 and ETRI lifelog2020) and introduces novel practical countermeasures, including participant compliance visualizations, interaction-triggered questionnaires to assess personal bias, and an optimized wearable non-wear detection pipeline.

Abstract

Chronic disease management and follow-up are vital for realizing sustained patient well-being and optimal health outcomes. Recent advancements in wearable sensing technologies, particularly wrist-worn devices, offer promising solutions for longitudinal patient follow-up by shifting from subjective, intermittent self-reporting to objective, continuous monitoring. However, collecting and analyzing wearable data presents unique challenges, such as data entry errors, non-wear periods, missing wearable data, and wearable artifacts. We therefore present an in-depth exploration of data analysis challenges tied to wrist-worn wearables and ambulatory label acquisition, using two real-world datasets (i.e., mBrain21 and ETRI lifelog2020). We introduce novel practical countermeasures, including participant compliance visualizations, interaction-triggered questionnaires to assess personal bias, and an optimized wearable non-wear detection pipeline. Further, we propose a visual analytics approach to validate processing pipelines using scalable tools such as tsflex and Plotly-Resampler. Lastly, we investigate the impact of missing wearable data on "window-of-interest" analysis methodologies. Prioritizing transparency and reproducibility, we offer open access to our detailed code examples, facilitating adaptation in future wearable research. In conclusion, our contributions provide actionable approaches for wearable data collection and analysis in chronic disease management.

Figures (15)

• Figure 1: mBrain application timeline of a dummy participant, showcasing contextual data, including user-defined semantic locations (e.g. “Work”).
• Figure 2: mBrain study interaction visualization of a single participant for a period of 90 days. Note: The figure consists of several subplots with a shared x-axis, each providing different layers of information about the participant's activity and interactions. Subplots (i) and (ii) showcase phone and wearable data sessions over time. Each bar on the x-axis represents a unique day. For the first four plots, the y-axis indicates the time of day. This format reveals patterns of data fragmentation and daily volumes across extended periods. Gray-shaded areas, consistent across all subplots, signify weekends. The mBrain study requires a minimum of eight hours of wearable data daily. This compliance is color-coded in the first two subplots: green represents days with more than 8 hours, while orange indicates less than 8 hours. The daily events subplot (iii) provides an overview of food intakes (dot-shaped markers; green: breakfast, blue: lunch, light blue: dinner, red-cross: skipped meal) and questionnaire interaction ($\nabla$: evening questionnaire, $\triangle$: morning questionnaire, #: stress event questionnaires). Subplot (iv) provides a visual record of the participant's headaches (depicted by red vertical bars) and medication intakes (indicated by green crosses). The final subplot (v), which denotes an interaction rate (%) as y-axis, illustrates the frequency of participant interactions with stress (light blue) and activity (light green) timeline events derived from the wearable data stream.
• Figure 3: ETRI lifelog 2020 study interaction visualization of a single participant for a period of 28 days. Note: Similar to the mBrain interaction plot (Figure \ref{['fig:study_interaction']}), the ETRI interaction visualization utilizes stacked bars to depict phone (i) and wearable (ii) data sessions. Accommodating the fact that participants manually labeled intervals, the label subplot (iii) also uses a bar interval representation, indicating periods for which social and affective labels are present. Remark how the phone and wearable session are a subset of the label session data. In alignment with Figure \ref{['fig:study_interaction']}, session bars are color-coded in orange when fewer than 8 hours of data are available for the corresponding day.
• Figure 4: Example of an mBrain alert message, shown to the study coordinators when no wearable data is received from a participant.
• Figure 5: (a) Screenshot of questions in the mBrain study’s morning questionnaire evaluating implicitness for headache and medication events. (b) Notifications activated based on responses to the implicitness questions