Exploring rating scale responsiveness: does the total score reflect the sum of its parts?

Evaluating rehabilitation requires rating scales that detect change. The authors examined Barthel Index (BI) data from 1,495 patients at a neurorehabilitation unit to determine whether total scale responsiveness accurately reflects item responsiveness. Total score effect sizes were moderate to large (0.47 to 1.09). Item-level effect sizes (0.13 to 1.16) reveal floor (3.5 to 82.3%) and ceiling (9.7 to 95.4%) effects. Results suggest BI total score effect sizes may hide item-level weaknesses and may underestimate the impact of rehabilitation.