The results support the reliability and validity of SCIM III in a multi-cultural setup. Despite several limitations of the study, the results indicate that SCIM III is an efficient measure for functional assessment of SCL patients and can be safely used for clinical and research trials, including international multi-center studies.

Background: A third version of the Spinal Cord Independence Measure (SCIM III), made up of three subscales, was formulated following comments by experts from several countries and Rasch analysis performed on the previous version. Objective: To examine the validity, reliability, and usefulness of SCIM III using Rasch analysis. Design: Multicenter cohort study. Setting: Thirteen spinal cord units in six countries from North America, Europe, and the Middle-East. Subjects: 425 patients with spinal cord lesions (SCL). Interventions: SCIM III assessments by professional staff members. Rasch analysis of admission scores. Main outcome measures: SCIM III subscale match between the distribution of item difficulty grades and the patient ability measurements; reliability of patient ability measures; fit of data to Rasch model requirements; unidimensionality of each subscale; hierarchical ordering of categories within items; differential item functioning across classes of patients and across countries. Results: Results supported the compatibility of the SCIM subscales with the stringent Rasch requirements. Average infit mean-square indices were 0.79-1.06; statistically distinct strata of abilities were 3 to 4; most thresholds between adjacent categories were properly ordered; item hierarchy was stable across most of the clinical subgroups and across countries. In a few items, however, misfit or category threshold disordering were found. Conclusions: The scores of each SCIM III subscale appear as a reliable and useful quantitative representation of a specific construct of independence after SCL. This justifies the use of SCIM in clinical research, including cross-cultural trials. The results also suggest that there is merit in further refining the scale.Spinal Cord (2007) 45, 275-291.

Study design: Construction of an international walking scale by a modi®ed Delphi technique. Objective: The purpose of the study was to develop a more precise walking scale for use in clinical trials of subjects with spinal cord injury (SCI) and to determine its validity and reliability. Setting: Eight SCI centers in Australia, Brazil, Canada (2), Korea, Italy, the UK and the US. Methods: Original items were constructed by experts at two SCI centers (Italy and the US) and blindly ranked in an hierarchical order (pilot data). These items were compared to the Functional Independence Measure (FIM) for concurrent validity. Subsequent independent blind rank ordering of items was completed at all eight centers (24 individuals and eight teams). Final consensus on rank ordering was reached during an international meeting (face validation). A videotape comprised of 40 clips of patients walking was forwarded to all eight centers and inter-rater reliability data collected. Results: Kendall coecient of concordance for the pilot data was signi®cant (W=0.843, P50.001) indicating agreement among the experts in rank ordering of original items. FIM comparison (Spearman's rank correlation coecient=0.765, P50.001) showed a theoretical relationship, however a practical dierence in what is measured by each scale. Kendall coecient of concordance for the international blind hierarchical ranking showed signi®cance (W=0.860, P50.001) indicating agreement in rank ordering across all eight centers. Group consensus meeting resulted in a 19 item hierarchical rank ordered`Walking Index for Spinal Cord Injury (WISCI)'. Inter-rater reliability scoring of the 40 video clips showed 100% agreement. Conclusions: This is the ®rst time a walking scale for SCI of this complexity has been developed and judged by an international group of experts. The WISCI showed good validity and reliability, but needs to be assessed in clinical settings for responsiveness. Spinal Cord (2000) 38, 234 ± 243