Exploring the Construct Validity of Two IRT-Derived Scales of the PAI-A

The Personality Assessment Inventory – Adolescent (PAI-A) is a 264-item self-report instrument designed to assess various facets of psychopathology and of personality in adolescents. Three studies were used to examine the performance of the Anxiety (ANX) and Depression (DEP) scales of the PAI-A, the first two studies using item response theory (IRT) methods to revise the ANX and DEP scales of the PAI-A using the graded response model (Samejima, 1996), and the third study investigating the construct validity of these IRT-derived scales. In Study 1, item performance was examined separately for each scale in a community sample (N = 707) of adolescents; Study 2 examined the PAI-A ANX and DEP scales in a clinical sample of adolescents (N = 1160). On the basis of these IRT analyses, some items were eliminated from the ANX and DEP scales, resulting in 13-item sets for both scales. In Study 3, the construct validity of the IRT-derived item sets as well as the original full-length ANX and DEP scale scores was investigated using a sample of adolescents admitted to the adolescent program of a private tertiary care inpatient treatment facility (N = 169). Correlations were computed between the PAI-A full-length and the IRT-derived ANX and DEP scales and each continuously scored validating instrument. These instruments included the anxiety and depression scale scores of the Youth Self-Report (YSR), Child Behavior Checklist (CBCL/6-18), Multidimensional Anxiety Scale for Children (MASC), and the Beck Depression Inventory-II (BDI-II). Correlations were strongest for the two most narrowly focused validating measures (i.e., MASC and BDI-II), and were notably smaller for parent-report measures (i.e., P-DISC and CBCL/6-18). For the categorically scored Youth and Parent Diagnostic Interview Schedule for Children (Y-DISC and P-DISC), t-tests of PAI-A scores between DISC diagnostic groups were conducted. For the full-length PAI-A scale scores, those with a positive diagnosis on the youth and parent DISC had higher scores compared to those with no/intermediate diagnosis. For the IRT-revised scores, mean differences were found between the youth DISC categories for ANX and DEP; however, differences between the parent DISC categories were found only for DEP. Those with a diagnosis had higher scores compared to those with no/intermediate diagnosis.