Current Correlations between Disease Modifiers & Physiologic Measurements from the MS Mosaic Study

Background: Many people with multiple sclerosis suffer from mobility impairments or sleep disturbance, but very little is known about whether these symptoms correlate with their current disease modifier.

Objectives: To evaluate whether physiologic measurements of sleep and activity are predictive of current disease modifier use.

Methods: MS Mosaic is a longitudinal study (NCT02845635) that combines data from a mobile platform with existing biomarkers and then utilizes machine learning methods to help reveal a more comprehensive picture of MS.  Continuously collected data from participants’ daily symptom surveys, medication diaries, and mobile sensors is analyzed through a Bayesian generative hierarchical model that uses a Dirichlet process at a higher level and then represents the observed data at a lower level, providing a particular patient’s “physiometry sub-group” membership and attempts to predict their current disease modifier use (if any).

Results: Disease subtypes can be discovered by combining the symptom/medications diaries and sensor-based physiometry measurements from the MS Mosaic app with machine learning clustering methodology. 

Conclusions: The ability of these identified “physiometry sub-groups” to predict a given patient’s current disease modifier is described.