Evaluation of the German Version of the Stroke Impact Scale (SIS) using Rasch Analysis

Purpose: The aim of this study is to evaluate the psychometric properties of the German version of the Stroke Impact Scale (SIS) using the Rasch model for ordered response categories.

Methods: Cross-sectional data from an ongoing international WHO cooperation study were used for this study. Data were collected within a multicenter study in Germany with 25 participating centers providing acute and rehabilitation services to stroke patients. Analyses were conducted using data from a convenience sample of 101 stroke patients receiving inpatient or outpatient services at one of the participating study centers.

Rasch analysis was conducted using RUMM2020 to evaluate (1) unidimensionality of the single domains and item fit, (2) response category functioning, (3) the targeting of the test, (4) reliability using Person Separation Reliability, (5) differential item functioning (DIF) for relevant patient groups, and (6) to compare the fit results of this study on the German SIS with the Rasch analysis results of the original SIS 2.0.

Results: Unidimensionality was confirmed in six SIS domains. Within the domains Emotion and Participation 4 items displayed model misfit. Response categories of 40 items showed threshold disordering. Data were purified by collapsing response categories and stepwise deletion of misfitting items. High values of Person Separation Reliability were found, with values above .90 in six domains. No DIF was detected with respect to age, gender, disability severity, and health care setting. Different fit results were found for the German SIS as compared to the original SIS 2.0.

Discussion: The results provide support for internal and construct validity, as well as reliability of the German version of the SIS, and also point out issues for further improvement and adaptation of the SIS. In future, the SIS still needs to prove its cross-cultural validity.

References:

Literatur beim Verfasser