Advancement in wrist worn physical activity trackers: technological developments and measurement accuracy: a systematic review
TL;DR
Accuracy improved substantially over time, although variability across brands, activities, and populations persisted; Acceptability was generally high across brands, with long wearing time and good participant satisfaction.
Abstract
Background Wrist-worn physical activity trackers have advanced rapidly over the past two decades, evolving from simple accelerometer-based devices to multi-sensor systems integrating photoplethysmography, advanced movement classification, and machine-learning algorithms. Despite widespread use in both research and clinical settings, concerns remain regarding the accuracy and reliability of key metrics such as heart rate, energy expenditure, step count, and activity intensity. Methods This systematic review followed PRISMA guidelines (PROSPERO: CRD42024551740). MEDLINE, Scopus, PubMed, and SPORTDiscus were searched for studies evaluating the accuracy or acceptability of wrist-worn activity trackers in laboratory or free-living-simulated conditions. Eligible studies included healthy or general-population samples using any wrist-worn consumer or research-grade device. Reference standards included ECG, indirect calorimetry, doubly labelled water, manual step counting, and research-grade accelerometry. Of 1,659 records identified, 47 studies met inclusion criteria. Results and discussion Accuracy improved substantially over time, although variability across brands, activities, and populations persisted. Early heart-rate monitors (pre-2010) frequently reported Pearson correlation coefficient as low as 0.50 during exercise, while still being under the validity standard (0.90) at rest. Modern devices such as Apple Watch and Garmin vivosmart demonstrate markedly increased validity scores (typically above 0.80 during exercise), reflecting improvements in optical sensor quality and motion-artifact filtering. Energy-expenditure accuracy showed the least improvement; even recent devices exhibit 10%–40% error relative to indirect calorimetry, reflecting the inherent challenge of estimating metabolic rate without direct O₂ and CO₂ measurement. Step-count accuracy, prior to 2017 showed errors of up to 25%, now commonly falls below 5% for steady-state walking due to improved accelerometer precision and gait-pattern recognition. Classification of physical-activity intensity has also improved, with misclassification decreasing from >20% in early models to 10%–15% in current multi-sensor devices. Acceptability was generally high across brands, with long wearing time and good participant satisfaction.