Causal Identification via DAG and ADMG Simplification in Wearable Parkinson’s Disease Studies

Parkinson's disease (PD) is a chronic neurodegenerative condition that is characterised by both motor and non-motor symptoms. During the last few years, wearable devices have started to be used in the clinical practice for monitoring patients' PD-related motor symptoms, during their daily activities. However, most studies to date rely on associations that may be confounded and clinically misleading when causal structure is ignored. The purpose of this study is to introduce non-parametric causal modelling using Directed Acyclic Graphs (DAGs) and Acyclic Directed Mixed Graphs (ADMGs) to identify causal effects of interest, such as PD diagnostic status from sensor data features, using large-scale population data. Although many standard techniques causal effect estimators exist, such as inverse probability weighting, causal bootstrapping, double machine learning, stratification, and average treatment effect estimators, their use depends on the structure of the graph. We demonstrate how complex causal graphs involving latent confounding and redundant variables can be systematically simplified using latent projection and graphical reduction while preserving the causal effect of interest. For PD, we demonstrate a plausible model which results in a DAG satisfying the backdoor criterion, thereby establishing the ignorability assumption and enabling the use of standard adjustment methods. In contrast, more general identification approaches, such as the ID algorithm, can identify causal effects in arbitrary graphs but often yield algebraically complex expressions that are difficult to estimate and interpret. This work also provides a practical and accessible guide for researchers to move from associative analyses toward causal inference in wearable-sensor studies.