Objective
To examine the fourth version of the Spinal Cord Independence Measure for reliability and validity.

Design
Partly blinded comparison with the criterion standard Spinal Cord Independence Measure III, and between examiners and examinations.

Setting
A multicultural cohort from 19 spinal cord injury units in 11 countries.

Participants
A total of 648 patients with spinal cord injury.

Intervention
Assessment with Spinal Cord Independence Measure (SCIM IV) and Spinal Cord Independence Measure (SCIM III) on admission to inpatient rehabilitation and before discharge.

Main outcome measures
SCIM IV interrater reliability, internal consistency, correlation with and difference from SCIM III, and responsiveness.

Results
Total agreement between examiners was above 80% on most SCIM IV tasks. All Kappa coefficients were above 0.70 and statistically significant ( P <.001). Pearson's coefficients of the correlation between the examiners were above 0.90, and intraclass correlation coefficients were above 0.90. Cronbach's alpha was above 0.96 for the entire SCIM IV, above 0.66 for the subscales, and usually decreased when an item was eliminated. Reliability values were lower for the subscale of respiration and sphincter management, and on admission than at discharge. SCIM IV and SCIM III mean values were very close, and the coefficients of Pearson correlation between them were 0.91-0.96 ( P <.001). The responsiveness of SCIM IV was not significantly different from that of SCIM III in most of the comparisons.

Conclusions
The validity, reliability, and responsiveness of SCIM IV, which was adjusted to assess specific patient conditions or situations that SCIM III does not address, and which includes more accurate definitions of certain scoring criteria, are very good and quite similar to those of SCIM III. SCIM IV can be used for clinical and research trials, including international multi-center studies, and its group scores can be compared with those of SCIM III.