Degrees of xerostomia? A Rasch analysis of the Xerostomia Inventory.

OBJECTIVE(S): The global demographic changes resulting in an ageing population require attention on xerostomia, as its prevalence appears to increase with age. The Xerostomia Inventory (XI) is a 11-item instrument developed to evaluate the symptoms and behavioural components of xerostomia, while a shortened 5-item version named Summated Xerostomia Inventory (SXI) was later proposed. The aim of the present study was to evaluate the construct validity of the XI and whether the SXI can provide a shortened version. Since previous studies focused only on dimensionality and reliability, we employed modern psychometric methodology to investigate properties such as differential item functioning (DIF) and targeting. STUDY DESIGN: The XI was applied to 164 middle-aged to older adults who participated in a randomized controlled trial to investigate the effects of alcohol-containing mouth rinse in Singapore. The psychometric properties of the XI were investigated with the Rasch model (Partial Credit Model). Overall model fit was evaluated with a summary chi-square statistic. Item fit was evaluated with the Fit Residual, and values between -2.5 and 2.5 are considered acceptable. DIF by sex was evaluated through a two-way ANOVA of the residuals. RESULTS: After collapsing the categories of "Hardly ever" and "Fairly often", the test of global fit (chi(2) (30) = 34.32, P = .27) indicated overall fit to the Rasch model. Since Fit Residuals were between -2 and 2, the fit of individual items was also adequate. No DIF was found between men and women, and targeting was adequate (mu = -0.56). CONCLUSION: The current study expanded the evidence on the XI and SXI validity and provides new implications for practice: a 3-point categorization ("Never," "Occasionally" and "Very often") should be preferred rather than the original 5-point categorization; the XI and SXI scores can be compared between men and women and will reflect true differences in xerostomia rather than measurement bias.