A Bayesian network analysis of gait speed change upon transition to uneven surfaces in older adults

Gait adaptability, defined as the ability to adjust walking performance to environmental challenges, likely reflects complex interactions among the central nervous system (CNS) and other physiological systems. However, the drivers of lower gait adaptability in older adults are poorly understood.

We applied a Bayesian network framework to understand multisystem interactions contributing to percent change in gait speed (%GSC) on transition from even to uneven surface in 159 older adults. Neuroimaging measures include gray matter volume, white matter hyperintensities, striatal dopaminergic (DA) neurotransmission, and functional connectivity. Other measures were obtained for domains important for locomotor control: health history, lifestyle, psychological well-being, cognition, and musculoskeletal and peripheral nervous systems. The Bayesian network captured relationships among these variables as a network, quantifying their direct and indirect dependencies. The predictive accuracy of %GSC from the Bayesian network was compared with that of multivariate linear regression using 10-fold cross-validation.

Participants exhibited slower gait on uneven compared to even surfaces (mean %GSC = -6.32%). The Bayesian network outperformed linear regression in predicting %GSC and identified four direct paths to %GSC from body mass index, muscle strength, striato-cortical sensorimotor connectivity, and purpose in life. Indirect paths to %GSC showed interrelations among CNS and non-CNS variables, including striatal DA neurotransmission, total grey matter volume, medications, proprioception, and sex.

Gait adaptability in older adults is influenced by interactions among functional connectivity, body composition, muscle strength, and psychological well-being. Strengthening both neural and physical systems through targeted interventions may mitigate declines in gait instability and preserve mobility with aging.