Defining Valid Activity Monitor Data: A Multimethod Analysis of Weight-Loss Intervention Participants' Barriers to Wear and First 100 Days of Physical Activity.

Christina Hernandez,Meghana Reddy,Melanie Jay,Stephanie L. Orstad,Lauren Gerchow,Nikhil R. Patel,Dawn K. Wilson

doi:10.3390/informatics8020039

Abstract

Despite the popularity of commercially available wearable activity monitors (WAMs), there is a paucity of consistent methodology for analyzing large amounts of accelerometer data from these devices. This multimethod study aimed to inform appropriate Fitbit wear thresholds for physical activity (PA) outcomes assessment in a sample of 616 low-income, majority Latina patients with obesity enrolled in a behavioral weight-loss intervention. Secondly, this study aimed to understand intervention participants' barriers to Fitbit use. We applied a heart rate (HR) criterion (≥10 h/day) and a step count (SC) criterion (≥1000 steps/day) to 100 days of continuous activity monitor data. We examined the prevalence of valid wear and PA outcomes between analytic subgroups of participants who met the HR criterion, SC criterion, or both. We undertook qualitative analysis of research staff notes and participant interviews to explore barriers to valid Fitbit data collection. Overall, one in three participants did not meet the SC criterion for valid wear in Weeks 1 and 13; however, we found the SC criterion to be more inclusive of participants who did not use a smartphone than the HR criterion. Older age, higher body mass index (BMI), barriers to smartphone use, device storage issues, and negative emotional responses to WAM-based self-monitoring may predict higher proportions of invalid WAM data in weight-loss intervention research.

Highlights

Wearable activity monitors (WAMs), such as the Fitbit®®, are widely available and highly acceptable behavior change tools [1,2]
The primary reason for this discrepancy was that the Fitbit device did not store minute-by-minute data for longer than one week; if the device was not synced with the Fitbit app, the data were not uploaded and the heart rate data were automatically deleted from the device
Few differences in demographic and health-related characteristics emerged between participants with and without valid wear weeks, suggesting other explanatory factors may be relevant in predicting valid wearable activity monitor (WAM) data

Summary

Introduction

Wearable activity monitors (WAMs), such as the Fitbit®®, are widely available and highly acceptable behavior change tools [1,2]. WAMs capture continuous physical activity (PA) behavior in real time and provide immediate feedback to the user on goal progress. WAMs are commonly deployed in PA research to change behavior and to assess PA, given the precision and objectivity of their accelerometer function when compared to self-reported PA measures [8]. Fitbit validity results for free-living moderate-vigorous physical activity (MVPA) range from moderate to strong correlation with research-grade accelerometers (Spearman’s r = 0.56 [12], 0.86 [13], and 0.88–0.91 [14]). Despite the benefits of objective data, continuous measurement of PA over time results in large quantities of data to process. Determining an appropriate analytic approach requires skills in PA assessment and data analysis software [15], and benefits from interdisciplinary decision making and inquiry into participants’ lived experience

Objectives

Methods

Results

Discussion

Conclusion