Examination of the construct validity of the Geriatric Depression Scale in older persons living in long-term care...

To conduct a comprehensive and simultaneous examination of the unidimensionality of all Geriatric Depression Scale (GDS) versions to determine their construct validity as depression-screening tools in older adult populations living in long-term care institutions.

Two hundred residents completed all assessments. The mean age was 80.6 ± 9.9 years, and 89% were male. The average length of institutional stay was 5.8 ± 5.5 years. Rasch analysis was used to evaluate the unidimensionality of the GDS-30 and its shorter version. Key indicators such as infit and outfit mean-square (MnSq) values and eigenvalues from the residual principal component analysis were assessed to identify misfit items and verify the construct validity of each version.

Six GDS-30 items were misfit (outfit MnSq=0.4∼2.94). Removing these resulted in 24 items (the Rasch-derived GDS-24), which better aligned with unidimensionality (infit MnSq = 0.74-1.31, outfit MnSq = 0.52-1.48). The GDS-15 had three misfitting items (infit MnSq = 0.70-1.33; outfit MnSq = 0.49-1.56), the GDS-10 had one misfitting item (outfit MnSq = 0.45), the GDS-5 had two misfitting items (outfit MnSq = 3.27 and 4.67), and the GDS-4 had three misfitting items (outfit MnSq = 1.75-9.90).

The Rasch-derived GDS-24 emerged as a balanced alternative, offering comprehensive screening with improved psychometric performance compared to the GDS-30 and its shorter versions.