Objective

To compare the psychometric properties of different short versions of the Geriatric Depression Scale (GDS) among community-dwelling widowed older people.

Methods

A cross-sectional design was conducted between February 2018 and August 2019 in southern Taiwan (n = 330). The GDS short versions included Hoyl et al’s 5-item version, Molloy et al’s 5-item version and van Marwijk et al’s 4-item version.

Results

All GDS short versions had satisfactory internal consistency (α = 0.83-0.90), with strong item loading embedded in the same construct of depression (0.43-0.72). The activity of daily living (ADL) scores had strong associations with van Marwijk et al’s 4-item version (absolute standardised coefficient [|β|] = 0.15-0.59), Hoyl et al’s 5-item version (|β| = 0.16-0.45) and Molloy et al’s 4-item version (|β| = 0.09-0.40).

Conclusions

Hoyl et al’s 5-item version was found to be more suitable than the other two versions of the GDS to assess the precise construct of depression. Meanwhile, van Marwijk et al’s 4-item version can be used when the focus is on ADL performance. However, the conclusion should be interpreted with caution since the results were derived only from a widowed population in Taiwan. More research on this topic among other populations is thus needed to corroborate our conclusion.