LAUSR

OBJECTIVE Evaluate the psychometric properties of the SCI-SET and PRISM using Rasch analysis to optimize their validity and efficiency. DESIGN Rasch analysis of the SCI-SET and PRISM represents a secondary analysis of data collected as part of a collaborative research project of the SCI Model Systems Centers. The overall survey was organized into four sections: 1) participant demographics and injury characteristics, 2) participant experiences of spasticity, 3) SCI-SET, and 4) PRISM. Participants were recruited from the community via multiple avenues. Data were collected and managed via an online survey tool using a secure web-based data management application. SETTING Participating SCI Model Systems Centers. PARTICIPANTS Most participants (N = 1,239) have lived with their injury for more than two years and used a wheelchair as their primary mode of mobility. Majority of the sample (58%) sustained cervical injuries. INTERVENTIONS None. MAIN OUTCOME MEASURES SCI-SET and PRISM. RESULTS The SCI-SET demonstrated strong measurement properties with acceptably high reliability and point-measure correlations, and no evidence of multidimensionality. However, respondents underused some rating scale categories. Analyses of the PRISM demonstrated three distinct subscales relating to the physical, psychological, and social influences of spasticity; respondents underused some rating scale categories. Combining underused rating scale categories for both spasticity instruments resulted in increased reliability and reduced respondent burden compared to the original versions. Both the modified SCI-SET (person separation reliability = 0.93) and PRISM (person separation reliability = 0.85, 0.89, 0.83 for physical, psychological, and social subscores) display strong measurement properties. CONCLUSIONS Measurement properties of the SCI-SET and PRISM improved from use of Rasch model methods. The SCI-SET required minor revisions, whereas the PRISM required definition of subscores. Both modified spasticity measures demonstrated adequate psychometric properties and correlations among the modified measures were high, providing evidence of convergent validity. We recommend use of the modified SCI-SET and PRISM measures in future studies.