Abstract
Purpose: Monitoring home exercise using accelerometry in patients with peripheral artery disease (PAD) may provide a tool to improve adherence and titration of the exercise prescription. However, methods for unbiased analysis of accelerometer data are lacking. The aim of the current post hoc analysis was to develop an automated method to analyze accelerometry output collected during home-based exercise.
Methods: Data were obtained from 54 patients with PAD enrolled in a clinical trial that included a home-based exercise intervention using diaries and an accelerometer. Peak walking time was assessed on a graded treadmill at baseline and 6 mo. In 35 randomly selected patient data sets, visual inspection of accelerometer output confirmed exercise sessions throughout the 6 mo. An algorithm was developed to detect exercise sessions and then compared with visual inspection of sessions to mitigate the heterogeneity in session intensity across the population. Identified exercise sessions were characterized on the basis of total step count and activity duration. The methodology was then applied to data sets for all 54 patients.
Results: The ability of the algorithm to detect exercise sessions compared with visual inspection of the accelerometer output resulted in a sensitivity of 85% and specificity of 90%. Algorithm-detected exercise sessions (total) and intensity (steps/wk) were correlated with change in peak walking time (r = 0.28; r = 0.43).
Conclusions: An algorithm to assess data from an accelerometer successfully detected home-based exercise sessions. Algorithm-identified exercise sessions were correlated with improvements in performance after 6 mo of training in patients with PAD, supporting the effectiveness of monitored home-based exercise.