Validation of two short forms of Stroke Impact Scale: unidimensionality and reliability

Purpose We examined the unidimensionality and Rasch reliability of both Jenkinson's and MacIsaac's eight-item short versions of the Stroke Impact Scale (SF-SIS), a questionnaire for assessing overall health-related quality of life (HRQOL). Methods This study was a secondary data analysis in which 263 persons with stroke completed the SIS. The 263 persons, on average, had age of 60 years, mild stroke, and moderate disability of self-care. The unidimensionality of both versions was validated via testing of model fitting and principal component analysis (PCA) of residuals using the Rasch analysis to determine the Rasch reliability and measures. Results The eight items in both SF-SIS versions met the criteria of infit and outfit MNSQ (0.6), indicating good data-model fit. The PCA showed that no dominant factors existed in the residuals of the items. The person reliability of Jenkinson's and MacIsaac's SF-SIS versions was 0.80 and 0.79, respectively. The Rasch measures (i.e., person measure in logits) ranged from -1.06 to 1.87 in Jenkinson's SF-SIS and -0.82 to 1.88 in MacIsaac's version. Conclusions The unidimensionality of both versions was supported. The Rasch measures of both appear valid for representing overall HRQOL levels. Both versions also showed acceptable reliability for research purposes.

Automated summary

This study examined the unidimensionality and Rasch reliability of Jenkinson's and MacIsaac's short versions of the Stroke Impact Scale (SF-SIS), which assesses overall health-related quality of life (HRQOL). The results showed that both versions met the criteria for good data-model fit and had no dominant factors in the residuals. The Rasch measures also indicated valid representation of overall HRQOL levels. Both versions showed acceptable reliability for research purposes.