Differential item functioning in the K-SADS using diagnosis of major depressive disorder as a grouping variable

Abstract

The goal of the current study was to test the assumption of measurement invariance between populations with Major Depressive Disorder (MDD) and those without MDD for the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children (K-SADS). When measurement invariance is violated, analyses comparing the results would be uninterpretable as they may mask or exaggerate true treatment effects. Due to the unscalability of the MDD group, Differential Item Functioning (DIF) analyses were done comparing the total sample to the nonMDD group. A 2-parameter logistic IRT model was run with nine depression symptoms assessed by the K-SADS using 3,386 participants. Then the same model was run using only the nonMDD participants’ data, which amounted to 2,246 participants from the total sample. Raju’s Z(H) and Lord’s ÷^2 were used for analyses. Lord’s ÷^2 identified Depression/Irritability and Psychomotor Disturbance as having DIF. A new model replacing these two symptoms with their component symptoms revealed that Irritability and Psychomotor Agitation were significantly easier and more discriminating for the nonMDD group than the total sample. These results indicate that the symptoms Irritability and Psychomotor Agitation maybe introducing multidimensionality to the measured latent MDD construct. Except for these two symptoms, the K-SADS showed strong measurement invariance across these groups. Limitations of this study were that the grouping variable was strongly related to the latent continuum being measured, a direct comparison between the nonMDD and MDD groups could not be conducted, and the DIF detecting statistics assumed independent samples.