A Differential Item Functioning (DIF) Study of the Geriatric Depression Scale-Short Form

Background and Purpose: Depression is common in older adults, and depressive symptoms differ as a function of age (e.g., Qualls & Knight, 2006).  While “older adult” is typically defined as someone 65 years and over, subjects as young as 50 years are commonly used in studies of “older” adults, especially in studies of substance abuse, severe mental illness, and HIV/Aids, which tend to age individuals physically and socially at an earlier point in life.

The Geriatric Depression Scale-Short Form (GDS-SF) is commonly used with adults 50 and older in both social work practice and research.  Differential item functioning (DIF) concerns the possibility items function differently for different groups and is a threat to construct validity.  The differences in depression as a function of age suggest items on the GDS-SF may work differently for adults at the lower part of the “older adult” range than for persons in the upper portion.  DIF could propagate through to GDS-SF total scores in such a way as to require interpretation of scores differentially as a function of age, complicating both clinical and research efforts. 

Relatively few DIF studies of GDS-SF scores have been conducted.  Few studies have investigated DIF as a function of age, and none have included adults younger than 60.  The purpose of this study was to (1) investigate possible DIF in GDS-SF items as a function of age, and (2) investigate if any DIF propagated to total scores so as to require differential score interpretation, a circumstance referred to as differential test functioning (DTF).

Methods:  Data from 647 adults 50 and older involved in gerontological research were used in this study. Item Response Theory (IRT) methods were used to test for DIF, with the grouping variable membership in age groups 50 to 64, and 65 and older.  Test information curves, and conditional standard error of measurement curves, for GDS-SF item scores for the two groups were compared to investigate DIF effects on measurement error in GDS-SF total scores.  Test characteristic curves were also compared to investigate possible DIF induced differences in true score metrics.

Results:  Five GDS-SF items showed uniform DIF and one showed non-uniform DIF.  The test information curves, and conditional standard error of measurement curves, for the two groups nearly overlapped, suggesting the DIF does not propagate to produce differential measurement error.  Comparisons of test characteristic curves suggested similar true score metrics for the two age categories.

Conclusions and Implications:  The results of the current study suggest the presence of DIF as a function of age in GDS-SF item scores that may propagate to create DTF, potentially complicating GDS-SF score interpretation.  Further research is needed investigating DIF, and in particular DTF, of GDS-SF scores as a function of age.  Investigations of sources of the DIF are of particular importance.  One important area of investigation is differences in meaning attributed to the items showing DIF by persons in the different age categories.

References

Qualls, S., & Knight, B.  (2006).  Psychotherapy for depression in older adults.  New York: Wiley.