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

Accelerometry-based physical activity monitoring can complement information obtained via patient-reported outcome measures (PROMs) by permitting objective, free-living measurements of mobility. The purpose of this study was to examine the relationship between the Prosthetic Limb User's Survey of Mobility (PLUS-M) PROM and accelerometer-measured steps and to develop and validate an equation for predicting steps. Participants with transtibial amputation completed the PLUS-M and wore an activPAL for seven days. LASSO regression was used to build a prediction model in training data (<i>n</i> = 80), and performance was evaluated in holdout data (<i>n</i> = 26). There was a moderately high correlation (<i>r</i> = 0.77) between model-predicted and actual steps in the holdout data. However, the model overestimated steps (<i>t</i>₂₅=-2.09, <i>p</i> = 0.046) and failed to meet the predefined equivalence threshold of ± 10% of the actual mean step counts (CI: 0, 968.82 steps; <i>p</i> > 0.05). Additionally, a root mean square error of 1,380 steps was noted between the model and activPAL-measured steps, representing roughly 33% of the mean actual step count. While the moderately high correlation highlights the potential of combining PROMs with accelerometer data to approximate physical activity, these limitations indicate the model is not yet suitable for clinical decision-making. Further refinement is required to improve accuracy and enhance practical application.