Evaluating the construct validity of the Trunk Impairment Scale: a Rasch analysis of its subscales

Background: The Trunk Impairment Scale (TIS) evaluates static, dynamic sitting balance and trunk coordination in people after stroke. This study aimed to evaluate the construct validity of the subscales of the TIS. Method A total of 162 participants (mean age 67, SD=11), in all stages after stroke, were included in the study. Rasch analysis (Partial Credit Model) was used (RUMM2020 software). Results There were no disordered thresholds on any of the polytomous items. The static sitting balance subscale (SSB) did not fit the Rasch model (Chi-square=15.68; p=0.0004). We were unable to examine DIF due to a large number of extreme cases in the dataset. The dynamic sitting balance subscale (DSB) fitted the Rasch model (Bonferroni adjusted pvalue= 0.005) after splitting item 2 by Barthel categories to adjust for uniform DIF (Chisquare= 42.65; p=0.0052). The coordination subscale (COO) fitted the Rasch model and did not display uniform or non-uniform DIF (Chi-square=7.87; p=0.446). In DSB and COO fewer than 5% of the independent t-tests performed on person estimated locations were significant, further supporting unidimensionality of these subscales. Graphical exploration of thresholds for DSB and COO confirmed a hierarchy of dependent items. Person and items were well distributed across the continuum of balance impairment. Discussion We were unable to examine SSB in full and further work with patients with very severe static sitting balance impairments is needed. Item 2 of DSB needs to be given detailed attention before assessing people after stroke. Conclusion Construct validity of two subscales was confirmed using Rasch analysis. Further work is needed to examine SSB.