Depression was measured with clinical assessment scales. In NSCAW, depression was
measured with the Children’s Depression Inventory long form, consisting of 27 self-report
items32 (Cronbach’s a =0.88 in this study). In
Add Health, depression was measured with the
19-item modified version of the Center for
Epidemiologic Studies Depression Scale (CESD).33 The full CESD consists of 20 items that
assess depression symptoms for the previous
week. Add Health used 16 of the 20 items as
originally worded, plus 2 items whose wording
was slightly altered from the original CES-D, and
1 item added by Garrison et al.34 for an
adolescent adaptation. The modifications do
not meaningfully affect the internal structure of
the measure35 (Cronbach’s a =0.87 in this
study). Both the Children’s Depression Inventory
and the CES-D are psychometrically sound
instruments for the measurement of adolescent
depressive symptoms.32,36,37 However,
because the measures were not identical,
we determined that the best comparison across
the 2 samples would be based on clinical
cutoffs for depression. The clinical cutoff
score for depression on the CES-D was 1636;
the clinical cutoff for depression on the
Children’s Depression Inventory was 66.32
Thus, we considered depression a dichotomous variable indicating not depressed or depressed.