The Smell Identification Test as a measure of olfactory identification ability in schizophrenia and healthy populations: a Rasch psychometric study.

This study examines University of Pennsylvania Smell Identification Test (UPSIT; R. L. Doty, 1995) performance in 133 controls and 54 chronic, medicated outpatients with schizophrenia (SZ) using item-response theory modeling. Results show that UPSIT items contribute to 1 factor, cover a range of 8 standard errors of measurement, and articulate 3 ability levels. Although it is not difficult enough to discriminate among persons of above-average ability, the test has diagnostic utility in detecting moderate impairment. Independent of item difficulty, 13 items differentiate patients from controls. When 45 patients and 45 controls were matched on gender and age, patient accuracy remained significantly reduced. The findings support the test's utility and demonstrate how traditional data analysis is insensitive to complexities in test performance.