Abstract
INTRODUCTION: This study examined differential item functioning of the Geriatric Depression Scale - Short Form (GDS-SF) in the National Alzheimer's Coordinating Center (NACC) Uniform Data Set (UDS) to identify potential variables that produce measurement bias. METHODS: Data from 14077 individuals' first NACC visit were analyzed. Multiple indicator, multiple causes (MIMIC) models assessed differential item functioning (DIF) of the 15-item GDS-SF across race, Hispanic ethnicity, primary language, sex, and cognitive status (Clinical Dementia Rating [CDR] scale scores), while adjusting for educational attainment. RESULTS: Participants were on average 73 (SD = 9.1) years old and 54.4% women. The majority (13 of 15) of the GDS-SF items demonstrated DIF. For many items, participants with any level of CDR cognitive impairment were more likely to endorse depressive symptoms. DISCUSSION: Findings indicate the presence of widespread DIF by cognitive impairment severity such that individuals with even mild cognitive impairment may respond differently to certain items on this measure. HIGHLIGHTS: The Geriatric Depression Scale - Short Form (GDS-SF) showed differential item functioning (DIF) in 13 of 15 items across demographic and cognitive groups. Only two items-hopelessness and worthlessness-were invariant across all groups. Cognitive status (Clinical Dementia Rating [CDR]) most strongly influenced item endorsement patterns. Our study used a large, diverse sample (National Alzheimer's Coordinating Center Uniform Data Set [NACC UDS]) and robust DIF analytic methods. Findings highlight both reliable and problematic GDS-SF items for older adults.