Laboratory comparison of consumer-grade and research-established wearables for monitoring heart rate, body temperature, and physical acitivity in sub-Saharan Africa.

Abstract

Background: Consumer-grade wearables are becoming increasingly popular in research and in clinical contexts. These technologies hold significant promise for advancing digital medicine, particularly in remote and rural areas in low-income settings like sub-Saharan Africa, where climate change is exacerbating health risks. This study evaluates the data agreement between consumer-grade and research-established devices under standardized conditions.

Methods: Twenty-two participants (11 women, 11 men) performed a structured protocol, consisting of six different activity phases (sitting, standing, and the first four stages of the classic Bruce treadmill test). We collected heart rate, (core) body temperature, step count, and energy expenditure. Each variable was simultaneously tracked by consumer-grade and established research-grade devices to evaluate the validity of the consumer-grade devices. We statistically compared the data agreement using Pearson's correlation r, Lin's concordance correlation coefficient (LCCC), Bland-Altman method, and mean absolute percentage error.

Results: A good agreement was found between the wrist-worn Withings Pulse HR (consumer-grade) and the chest-worn Faros Bittium 180 in measuring heart rate while sitting, standing, and slow walking on a treadmill at a speed of 2.7 km/h (r ≥ 0.82, |bias| ≤ 3.1 bpm), but this decreased with increasing speed (r ≤ 0.33, |bias| ≤ 11.7 bpm). The agreement between the Withing device and the research-established device worn on the wrist (GENEActiv) for measuring the number of steps also decreased during the treadmill phases (first stage: r = 0.48, bias = 0.6 steps/min; fourth stage: r = 0.48, bias = 17.3 steps/min). Energy expenditure agreement between the Withings device and the indirect calorimetry method was poor during the treadmill test (|r| ≤ 0.29, |bias | ≥ 1.7 MET). The Tucky thermometer under the armpit (consumer-grade) and the Tcore sensor on the forehead were found to be in poor agreement in measuring (core) body temperature during resting phases (r ≤ 0.53, |bias| ≥ 0.8°C) and deteriorated during the treadmill test.

Conclusion: The Withings device showed adequate performance for heart rate at low activity levels and step count at higher activity levels, but had limited overall accuracy. The Tucky device showed poor agreement with the Tcore in all six different activity phases. The limited accuracy of consumer-grade devices suggests caution in their use for rigorous research, but points to their potential utility in capture general physiological trends in long-term field monitoring or population-health surveillance.