Using wearable devices to generate real-world, individual-level data in rural, low-resource contexts in Burkina Faso, Africa: A case study

Abstract

BackgroundWearable devices may generate valuable data for global health research for low- and middle-income countries (LMICs). However, wearable studies in LMICs are scarce. This study aims to investigate the use of consumer-grade wearables to generate individual-level data in vulnerable populations in LMICs, focusing on the acceptability (quality of the devices being accepted or even liked) and feasibility (the state of being workable, realizable, and practical, including aspects of data completeness and plausibility).MethodsWe utilized a mixed-methods approach within the health and demographic surveillance system (HDSS) to conduct a case study in Nouna, Burkina Faso (BF). All HDSS residents older than 6 years were eligible. N = 150 participants were randomly selected from the HDSS database to wear a wristband tracker (Withings Pulse HR) and n = 69 also a thermometer patch (Tucky thermometer) for 3 weeks. Every 4 days, a trained field worker conducted an acceptability questionnaire with participants, which included questions for the field workers as well. Descriptive and qualitative thematic analyses were used to analyze the responses of study participants and field workers.ResultsIn total, n = 148 participants were included (and n = 9 field workers). Participant's acceptability ranged from 94 to 100% throughout the questionnaire. In 95% of the cases (n = 140), participants reported no challenges with the wearable. Most participants were not affected by the wearable in their daily activities (n = 122, 83%) and even enjoyed wearing them (n = 30, 20%). Some were concerned about damage to the wearables (n = 7, 5%). Total data coverage (i.e., the proportion of the whole 3-week study duration covered by data) was 43% for accelerometer (activity), 3% for heart rate, and 4% for body shell temperature. Field workers reported technical issues like faulty synchronization (n = 6, 1%). On average, participants slept 7 h (SD 3.2 h) and walked 8,000 steps per day (SD 5573.6 steps). Acceptability and data completeness were comparable across sex, age, and study arms.ConclusionWearable devices were well-accepted and were able to produce continuous measurements, highlighting the potential for wearables to generate large datasets in LMICs. Challenges constituted data missingness mainly of technical nature. To our knowledge, this is the first study to use consumer-focused wearables to generate objective datasets in rural BF.

Widening the lens of population-based health research to climate change impacts and adaptation: the climate change and health evaluation and response system (CHEERS)

“…In 2020, five weather stations were installed across the Nouna CHEERS site to represent various agro-ecological zones. A sensor-based sub-cohort, stratified by age and gender, collected data on daily activity, sleep, heart rate, indoor temperature, and humidity [refer to ( 41 , 42 )]. Remote sensing approaches, including agricultural yield models, were employed to estimate crop productivity ( 43 , 44 ).…”

Implementation of precision medicine in healthcare—A European perspective

“…The term RWD refers to health data that are collected during the course of disease, including treatment and outcome data [45]. These may be derived from individual patient records or registry‐based studies and may also include data collected via health apps or by wearables [46, 47]. Such data reflect to a certain extent the reality of treatment and associated outcomes outside the context of clinical studies which—by design and intentionally—carefully control numerous parameters and factors.…”

On Monetizing Personal Wearable Devices Data: A Blockchain-based Marketplace for Data Crowdsourcing and Federated Machine Learning in Healthcare

“…In contrast, data brokers accumulate large amounts of data and use it to create products for surveillance and marketing [8] . In addition to privacy, creating machine learning models that work effectively for individuals from different backgrounds is inhibited by the inaccessibility of inclusive datasets to researchers [9] .…”