Identification of measurement differences between English and Spanish language versions of the Mini-Mental State Examination. Detecting differential item functioning using MIMIC modeling.

BACKGROUND: Knowledge of the extent to which measurement of adult cognitive functioning differs between Spanish and English language administrations of the Mini-Mental State Examination (MMSE) is critical for inclusive, representative, and valid research of older adults in the United States.
OBJECTIVES: We sought to demonstrate the use of an item response theory (IRT) based structural equation model, that is, the MIMIC model (multiple indicators, multiple causes), to evaluate MMSE responses for evidence of differential item functioning (DIF) attributable to language of administration.
SUBJECTS: We studied participants in a dementia case registry study (n = 1546), 42% of whom were examined with the Spanish language MMSE.
RESULTS: Twelve of 21 items were identified as having significant uniform DIF. The 4 most discrepant included orientation to season, orientation to state, repeat phrase, and follow command. DIF accounted for two-thirds of the observed difference in underlying level of cognitive functioning between Spanish- and English-language administration groups.
CONCLUSIONS: Failing to account for measurement differences may lead to spurious inferences regarding language group differences in level of underlying level of cognitive functioning. The MIMIC model can be used to detect and adjust for such measurement differences in substantive research.