Abstract
Daily activity can be characterized by metrics such as duration, intensity, frequency. Combining these activity metrics into a temporal pattern (barcoding), and calculating its complexity can provide a global view of physical behaviour. We hypothesize that complexity increases following rehabilitation (n=524 inpatients). Using an accelerometer fixed on the thigh, the activity of one full day was recorded at the beginning of the rehabilitation and at follow-up 14 days later. The activity profile (i.e. time spent sitting, standing, and walking) was obtained. Activity patterns were built considering 14 different levels of intensity and the entropy of these levels was calculated (Lempel-Ziv metric). Complexity increased significantly, confirming its consistency with the activity profile. Complexity was correlated with Barthel Index, age, gender, and severity of mobility impairments. Complexity analysis of daily activity is deemed suitable to assess rehabilitation outcome and it is highly correlated with a standard clinical score.