[Unidimensionality of a functional measure for patient with an injured upper limb]

The construction of an instrument including a number of tests requires an analysis of its structure and its unidimensionality (which allows calculation of global score), and the determination of the difficulty level of various tests. This study examined a tool including 67 tests designed to evaluate the functional ability of patients with an injured upper limb. The patients seen in a rehabilitation centre during 12 months (173 subjects) were evaluated by the occupational therapists familiar with the tool. The statistical analyses were made using the principal component analysis method (PCAM), the Cronbach's coefficient and the Rasch model. The PCAM showed 3 principal factors which explained 44%, 10% and 4% of the total variance respectively in the case of patients with injured dominant limb. The predominance of the first axis and the high ratio of first by second eigenvalues suggested the unidimensionality of the tool. The Cronbach's value of 0.97 attested the good congruence of the items. The results obtained with the Rasch model seemed to be consistent with the hypothesis of the unidimensionality of the tool. This analysis also provided the difficulty scale of various tests. Similar results were obtained in patients with injured non dominant limb or with all the sample. The methods used provide complementary results.