Internal consistency reliability of the fractionated and whole University of Pennsylvania Smell Identification Test

Abstract

The internal consistency reliability (ICR) of the 40-item University of Pennsylvania Smell Identification Test (UPSIT) and its 10-, 20-, and 30-item fractions was explored, as well as the relationships between the fractions and the entire 40-item test. Pearson correlation coefficients (rs) were computed among all independent combinations and permutations of the four 10-item UPSIT booklets using data from 774 subjects. The median r values of the 10- and 20-item combinations were used to establish the ICRs of the 10- and 20-item tests. The ICRs of the 30- and 40-item tests were estimated using the Spearman-Brown formula and the median rs of the 20-item combinations. Additional ICR estimates of the 40-item UPSIT were obtained from nonsymmetrical fractions using the Horst formula. The ICRs for the UPSIT and its 10-, 20-, and 30-item fractions were 0.922, 0.752, 0.855 and 0.898, respectively. No major sex differences emerged. Estimates of correlations between (1) single booklets and two-booklet combinations and (2) the 40-item UPSIT using Guilford's (1953) correction for nonindependence ranged from 0.812 to 0.871. Overall, these results indicate that (1) the UPSIT and its 10-, 20-, and 30-item fragments have very high ICRs and (2) individual UPSIT booklets or their combinations can be used to assess smell function in a reliable manner where extreme time constraints are present (e.g., in surveys and in brief neuropsychological test batteries).