Methods: Multi-group confirmatory factor analysis (CFA) methods are used to test whether the items on the USSIS function differently for older adults with: (1) moderate or lower levels of depression than those with (2) moderate to severe levels of depression as measured by the Geriatric Depression Scale (GDS). The data from 222 older adults are used to fit a hypothesized measurement model testing for full measurement equivalence, hence the absence of DIF, using SPSS AMOS version 24.
Results: The CFA results suggest DIF is exhibited by all four USSIS items in the form of differential factor loadings and intercepts. The fit indices for the final CFA model are consistent with a close fitting model, χ2(2) = 2.23, p > .05; CFI = 1.0, TLI = .99; RMSEA = .02, 90% CI [< .001, .14], PCLOSE = .50; SRMR = .01. These results suggest all four USSIS items function differently for older adults with moderate or higher depression levels than for those with lower depression levels. The factor loadings of all four USSIS items on the construct “suicidal thinking” are higher for the moderate to severely depressed group; and all intercepts are higher for the moderately to severely depressed group. These results suggest the presence of differential test functioning for USSIS scores with respect to depression, specifically that the slope of the relationship between depression and suicidal thinking is steeper for the moderately to severely depressed group, a conjecture consistent with results of a multi-level analysis.
Discussion and Implications: These results suggest scores on the USSIS must be interpreted in context of the magnitude of level of depression in older adults. A depression context-dependent inferential strategy for interpreting USSIS scores when assessing suicidal risk suggested by these results is described. This conditional interpretation of scores on the USSIS may help improve the validity of clinical inferences concerning the risk of suicide in older adult clients. These results also suggest that research is needed investigating the possible DIF of items on other measures of suicidal ideation as a function of level of depression with older adults.