Conference Agenda

Purpose: The 24-hour human movement behaviour is inherently complex and heterogeneous, making it challenging to represent with a single metric. Accelerometers provide more information than simply time spent at different intensity levels. From raw acceleration signals, a wide array of indicators can be extracted, reflecting multiple dimensions such as intensity, duration, volume, frequency, accumulation pattern, timing, regularity, and complexity. This study aimed to identify distinct movement behaviour phenotypes and examine how variations in these phenotypes relate to cardiometabolic health.
Methods: In this cross-sectional study, 1009 adults from Luxembourg (aged 25–79) with at least four valid days of accelerometer data were included. A reversed graph embedding technique reduced the complexity of movement behaviours into a two-dimensional tree structure, allowing the identification of distinct phenotypes. Linear regression models were applied to explore associations between the tree dimensions, movement behaviour indicators, and cardiometabolic health biomarkers.
Results: Four movement behaviour phenotypes emerged: inactive, irregularly active, regularly active, and regularly low active. Both tree dimensions showed favourable associations with cardiometabolic outcomes, indicating that greater overall activity volume—regardless of its regularity—may benefit health. Among the indicators, average 24-hour acceleration and total step count were the strongest contributors to phenotypic variation.
Conclusion: This study demonstrates that a novel dimensionality reduction approach can effectively capture the multifaceted nature of movement behaviours, highlight key movement behaviour indicators, and distinguish between meaningful phenotypes. The findings enhance understanding of behavioural heterogeneity and emphasize movement dimensions most relevant to cardiometabolic health.