Predicting physical activity from self-reported mobility in individuals with transtibial amputation: A validation...

Area of Science:

• Rehabilitation Engineering
• Biomedical Engineering
• Wearable Technology

Background:

• Patient-reported outcome measures (PROMs) like the Prosthetic Limb User's Survey of Mobility (PLUS-M) offer insights into mobility but lack objective measurement.
• Accelerometry provides objective, free-living physical activity data, complementing subjective PROMs.

Purpose of the Study:

• To investigate the relationship between the PLUS-M PROM and accelerometer-measured steps in individuals with transtibial amputation.
• To develop and validate a predictive model for estimating daily steps using PLUS-M scores.

Main Methods:

• Participants with transtibial amputation completed the PLUS-M and wore an activPAL accelerometer for seven days.
• LASSO regression was employed to develop a prediction model using training data (n=80) and validated on holdout data (n=26).

Main Results:

• A moderately high correlation (r=0.77) was observed between predicted and actual steps in the holdout data.
• The model overestimated steps and did not meet the predefined equivalence threshold (±10% of mean step count).
• A root mean square error of 1,380 steps (approx. 33% of mean actual steps) was found.

Conclusions:

• Combining PROMs with accelerometer data shows potential for approximating physical activity in prosthetic users.
• Current model limitations, including overestimation and significant error, preclude its use in clinical decision-making.
• Further research and model refinement are necessary to improve accuracy and clinical applicability.