Beyond motion: a four-layer framework for free-living energy expenditure estimation

Abstract

Wearable energy expenditure estimates now inform fitness decisions and metabolic-disease trial endpoints. Against doubly labeled water (DLW), validation studies report activity-component error of 19%–100%; against indirect calorimetry, wrist devices show laboratory medians above 20% and errors exceeding 100% in clinical populations. For motion-dominant pipelines, a substantial component of the residual error is mechanistic rather than purely algorithmic: the relevant metabolic state is absent from the motion signal, so better models of that signal alone cannot recover it. A four-layer framework decomposes that state into a circadian resting baseline, the thermic effect of food (TEF), walking economy, and thermoregulatory thermogenesis. Each layer names a measurable signal and the sensing or calibration step that recovers it. The roadmap runs from physiologists and sensor developers working against calorimetric references to the modeling teams downstream.