Cross-diagnostic validity in a generic instrument: an example from the Functional Independence Measure in Scandinavia

To analyse the cross-diagnostic validity of the Functional Independence Measure (FIM) motor items in patients with spinal cord injury, stroke and traumatic brain injury and the comparability of summed scores between these diagnoses.

Data from 471 patients on FIM motor items at admission (stroke 157, spinal cord injury 157 and traumatic brain injury 157), age range 11-90 years and 70 % male in nine rehabilitation facilities in Scandinavia, were fitted to the Rasch model. A detailed analysis of scoring functions of the seven categories of the FIM motor items was made prior to testing fit to the model. Categories were re-scored where necessary. Fit to the model was assessed initially within diagnosis and then in the pooled data. Analysis of Differential Item Functioning (DIF) was undertaken in the pooled data for the FIM motor scale. Comparability of sum scores between diagnoses was tested by Test Equating.

The present seven category scoring system for the FIM motor items was found to be invalid, necessitating extensive rescoring. Despite rescoring, the item-trait interaction fit statistic was significant and two individual items showed misfit to the model, Eating and Bladder management. DIF was also found for Spinal Cord Injury, compared with the other two diagnoses. After adjustment, it was possible to make appropriate comparisons of sum scores between the three diagnoses.

The seven-category response function is a problem for the FIM instrument, and a reduction of responses might increase the validity of the instrument. Likewise, the removal of items that do not fit the underlying trait would improve the validity of the scale in these groups. Cross-diagnostic DIF is also a problem but for clinical use sum scores on group data in a generic instrument such as the FIM can be compared with appropriate adjustments. Thus, when planning interventions (group or individual), developing rehabilitation programs or comparing patient achievements in individual items, cross-diagnostic DIF must be taken into account.