Using Wear Time for the Analysis of Consumer-Grade Wearables' Data: Case Study Using Fitbit Data

Area of Science:

• Digital Health
• Human Behavior Monitoring
• Wearable Technology

Background:

• Consumer-grade wearables offer valuable real-world human behavior data.
• Maintaining user engagement and compliance (e.g., wear time) is a significant challenge, leading to data gaps and potential "wearables' abandonment".

Purpose of the Study:

• To quantify the impact of wear time requirements on study results using diverse population datasets.
• To emphasize the necessity of accounting for participant wear time in analyzing consumer-grade wearable data.
• To determine if all research questions require the same wear time compliance.

Main Methods:

• Analyzed 3 Fitbit datasets from 6 diverse population samples (caregivers, students, pediatric oncology patients).
• Assessed the sensitivity of average daily step count and heart rate to varying wear time definitions (Aim 1).
• Evaluated research question feasibility with lower compliance samples, focusing on average daily step count and average heart rate while walking (Aim 2).

Main Results:

• Population average daily step count estimates varied by up to 2000 steps based on analysis methods and wear time compliance.
• Low wear time samples (<15 hours/day) showed highest sensitivity to analysis method changes; individual step count differences exceeded 1000-3000 steps for a subset of participants.
• Average daily heart rate estimates were robust to wear time variations; some individuals with sufficient data for heart rate while walking lacked adequate data for daily step count.

Conclusions:

• Demonstrated a direct relationship between parameter estimates from consumer-grade wearables and participant wear time.
• Highlighted the critical importance of thorough wear time analysis for ensuring the relevance and reliability of wearable data findings.