Table 3.

Predicting intervention outcomes in ARAT, BBT, and NHPT using data collected pre-intervention — detailed performance of best performing models

	Best performing models and feature sets
	Model	Feature set	Balanced accuracy (%)	Sensitivity (%)	Specificity (%)	Precision (%)
ARAT	Linear regression	1, 6	89	100	79	67
BBT	Decision tree	1	83	67	100	100
NHPT	Linear regression	5	73	67	79	57

Best performing models were selected according to the balanced accuracy and their ability to correctly identify the unexpected non-responder. Feature set nomenclature: (1) patient master data (ms type, chronicity, age, sex); (2) intervention group; (3) disability (EDSS, disability group); (4) conventional scales of body functions (motricity index, static fatigue index, monofilament index, symbol digit modality test, Fahn’s tremor rating scale); (5) digital health metrics of sensorimotor impairments (ten VPIT metrics); (6) conventional scale of activity (ARAT, NHPT, BBT)