Gait model predicts health phenotypes across 18 body systems

Researchers developed a model that analyzes three-dimensional skeletal motion recorded with a single depth camera during five motor tasks. The study included 3,328 deeply phenotyped adults and tested whether learned representations of whole-body movement could predict health measures across 18 body systems. The authors compared these representations with conventional engineered gait features and with a baseline model using age, body mass index (BMI), visceral adipose tissue area and height.

Learned representations predicted age with a Pearson correlation of 0.69, BMI with a correlation of 0.91 and visceral adipose tissue area with a correlation of 0.83. They significantly predicted 1,887 of 3,210 tested phenotypes. Compared with the baseline, gait representations improved prediction across all 18 systems in males and 17 of 18 in females, and also improved prediction of clinical diagnoses and medication use. Leg movement contributed most to metabolic and frailty predictions, while torso movement captured sleep and lifestyle phenotypes.

The findings support further study of gait as a multisystem biomarker, but do not establish that the model can diagnose disease or improve patient outcomes. The reported results are predictive associations in the studied cohort; external validation and evaluation in clinical workflows would be needed before routine use.