INS NYC 2024 Program

Poster	Poster Session 11 Program Schedule 02/17/2024 10:45 am - 12:00 pm Room: Majestic Complex (Posters 61-120) Poster Session 11: Cultural Neuropsychology \| Education/Training \| Professional Practice Issues Final Abstract #95 Differential Item Functioning in the Geriatric Depression Scale-15 by Ethnoracial group Katie Singsank, University of Nevada Las Vegas, Las Vegas, United States Chad Cross, University of Nevada Las Vegas, Las Vegas, United States Samantha John, University of Nevada Las Vegas, Las Vegas, United States Category: Assessment/Psychometrics/Methods (Adult) Keyword 1: cross-cultural issues Keyword 2: psychometrics Keyword 3: demographic effects on test performance Objective: Assess the influence of ethnoracial identity on differential item functioning (DIF) in a gold-standard patient reported outcome measure, the Geriatric Depression Scale-15 (GDS). Participants and Methods: Data from all versions of the National Alzheimer’s Coordinating Center Uniform Dataset (NACC UDS) were used to investigate DIF across ethnoracial groups (Non-Hispanic white (NHW), Hispanic white (HW), Non-Hispanic Black (NHB), and Hispanic Black (HB)). The sample (n = 41,361) consisted of older adults (M_age=71.51, SD_age=10.331; 42.7% male, 57.3% female) with complete baseline visit data who predominantly identified as NHW (81.9% NHW, 12.5% NHB, 5.3% HW, 0.3% HB). The sample was cognitively heterogeneous (31.3% Dementia, 22.8% Mild Cognitive Impairment (MCI), 4.7% Impaired not MCI, and 41.2% Normal Cognition). To assess scale dimensionality prior to item response theory (IRT) analyses, we conducted Confirmatory Factor Analysis (CFA) and Exploratory Factor Analysis (EFA) and dropped four items with strong loadings on a secondary factor. IRT was performed on the 11 remaining binary (i.e. 0, 1) items and logistic regression was utilized for DIF. We evaluated for uniform DIF (i.e., one group score differs from another in magnitude only and uniformly across the latent variable) and non-uniform DIF (i.e., one group score shows a differential response depending on the location along the uniform variable). Ethnoracial differences in GDS scores were assessed using Welch’s corrected ANOVA with Games-Howell post-hoc comparisons. Results: From the IRT model, item fit statistics were all significant (< 0.001) and discrimination values ranged from 0.883 (moderate) to 14.696 (very high). Seven items showed high discrimination (>2) and 3 items showed very high discrimination (“more problems with memory than most” = 14.696, “feel worthless” = 3.142, and “life is empty” = 2.766). We found evidence of both uniform and non-uniform DIF. Omnibus DIF test results for all items were significant at p<0.008, with 10 of 11 items showing statistically significant uniform DIF and 9 of 11 items showing statistically significant non-uniform DIF. Among the three items with very high discrimination values, two of those: “problems with memory” and “life is empty”, had significant uniform and non-uniform DIF. Ethnoracial groups were significantly different on average GDS score (F(3, 616.039)=62.171, p<.001), with significant post-hoc differences among HW participants (M=3.01, SD=3.15) who endorsed depression at higher levels than NHW (M=2.18, SD=2.62) and NHB (M=2.20, SD=2.65) participants. Ethnoracial groups differentially endorsed GDS items; “problems with memory” was endorsed more frequently by NHW (34.5%) and HW (33.3%) than NHB (23.9%) and HB (26.3%) and “life is empty” was endorsed most by HW (19.1%) and HB (17.4%) relative to NHB (10.1%) and NHW (8.3%). Conclusions: Confirming the presence of DIF on the GDS-15 by ethnoracial group provides insight into how cultural differences influence endorsement of depressive symptoms (or interpretation of language surrounding depressive symptoms). Results can be used to recalibrate existing measures for more accurate scoring and should inform future item development. Future analyses will evaluate the presence of DIF across diagnostic groups, including further analyses of “memory problems” and other disease-specific items within a clinical sample.

Poster

02/17/2024
10:45 am - 12:00 pm
Room: Majestic Complex (Posters 61-120)

Poster Session 11: Cultural Neuropsychology | Education/Training | Professional Practice Issues

Final Abstract #95

Differential Item Functioning in the Geriatric Depression Scale-15 by Ethnoracial group

Katie Singsank, University of Nevada Las Vegas, Las Vegas, United States
Chad Cross, University of Nevada Las Vegas, Las Vegas, United States
Samantha John, University of Nevada Las Vegas, Las Vegas, United States

Category: Assessment/Psychometrics/Methods (Adult)

Keyword 1: cross-cultural issues
Keyword 2: psychometrics
Keyword 3: demographic effects on test performance

Objective:

Assess the influence of ethnoracial identity on differential item functioning (DIF) in a gold-standard patient reported outcome measure, the Geriatric Depression Scale-15 (GDS).

Participants and Methods:

Data from all versions of the National Alzheimer’s Coordinating Center Uniform Dataset (NACC UDS) were used to investigate DIF across ethnoracial groups (Non-Hispanic white (NHW), Hispanic white (HW), Non-Hispanic Black (NHB), and Hispanic Black (HB)). The sample (n = 41,361) consisted of older adults (M_age=71.51, SD_age=10.331; 42.7% male, 57.3% female) with complete baseline visit data who predominantly identified as NHW (81.9% NHW, 12.5% NHB, 5.3% HW, 0.3% HB). The sample was cognitively heterogeneous (31.3% Dementia, 22.8% Mild Cognitive Impairment (MCI), 4.7% Impaired not MCI, and 41.2% Normal Cognition).

To assess scale dimensionality prior to item response theory (IRT) analyses, we conducted Confirmatory Factor Analysis (CFA) and Exploratory Factor Analysis (EFA) and dropped four items with strong loadings on a secondary factor. IRT was performed on the 11 remaining binary (i.e. 0, 1) items and logistic regression was utilized for DIF. We evaluated for uniform DIF (i.e., one group score differs from another in magnitude only and uniformly across the latent variable) and non-uniform DIF (i.e., one group score shows a differential response depending on the location along the uniform variable).

Ethnoracial differences in GDS scores were assessed using Welch’s corrected ANOVA with Games-Howell post-hoc comparisons.

Results:

From the IRT model, item fit statistics were all significant (< 0.001) and discrimination values ranged from 0.883 (moderate) to 14.696 (very high). Seven items showed high discrimination (>2) and 3 items showed very high discrimination (“more problems with memory than most” = 14.696, “feel worthless” = 3.142, and “life is empty” = 2.766). We found evidence of both uniform and non-uniform DIF. Omnibus DIF test results for all items were significant at p<0.008, with 10 of 11 items showing statistically significant uniform DIF and 9 of 11 items showing statistically significant non-uniform DIF. Among the three items with very high discrimination values, two of those: “problems with memory” and “life is empty”, had significant uniform and non-uniform DIF.

Ethnoracial groups were significantly different on average GDS score (F(3, 616.039)=62.171, p<.001), with significant post-hoc differences among HW participants (M=3.01, SD=3.15) who endorsed depression at higher levels than NHW (M=2.18, SD=2.62) and NHB (M=2.20, SD=2.65) participants.

Ethnoracial groups differentially endorsed GDS items; “problems with memory” was endorsed more frequently by NHW (34.5%) and HW (33.3%) than NHB (23.9%) and HB (26.3%) and “life is empty” was endorsed most by HW (19.1%) and HB (17.4%) relative to NHB (10.1%) and NHW (8.3%).

Conclusions:

Confirming the presence of DIF on the GDS-15 by ethnoracial group provides insight into how cultural differences influence endorsement of depressive symptoms (or interpretation of language surrounding depressive symptoms). Results can be used to recalibrate existing measures for more accurate scoring and should inform future item development. Future analyses will evaluate the presence of DIF across diagnostic groups, including further analyses of “memory problems” and other disease-specific items within a clinical sample.