A data-driven framework for selecting and validating digital health metrics: use-case in neurological sensorimotor...

Area of Science:

• Digital health
• Neuroscience
• Rehabilitation technology

Background:

• Digital health metrics offer potential for understanding impaired body functions, particularly in neurological disorders.
• Clinical integration of these metrics is hindered by insufficient validation of numerous abstract measures.
• The Virtual Peg Insertion Test (VPIT) is a technology-aided assessment for upper limb sensorimotor impairments.

Purpose of the Study:

• To propose and validate a data-driven framework for selecting clinically relevant digital health metrics.
• To apply this framework to the VPIT to identify a core set of validated metrics for upper limb sensorimotor impairments.
• To demonstrate the framework's ability to address challenges in clinical integration of digital health metrics.

Main Methods:

• Developed a data-driven framework incorporating pathophysiological motivation, demographic confound modeling, and clinimetric property evaluation (discriminant validity, structural validity, reliability, measurement error, learning effects).
• Applied the framework to 77 VPIT metrics from 120 neurologically intact and 89 affected individuals.
• Utilized a step-by-step selection procedure based on clinical evidence.

Main Results:

• Successfully selected 10 clinically relevant core metrics from the VPIT.
• These core metrics demonstrated valid, reliable, and informative assessment of multiple sensorimotor impairments.
• The selected metrics identified subtle impairments missed by conventional scales and covered arm/hand sensorimotor deficits comprehensively.

Conclusions:

• The proposed framework offers a transparent, evidence-based alternative to traditional metric selection algorithms.
• Validated core metrics for the VPIT were established, facilitating their integration into neurorehabilitation trials.
• This approach can significantly improve the clinical integration of digital health metrics in neurological assessments.