To measure depressive symptoms, we utilized the Center
for Epidemiologic Studies Depression Scale (CES-D),
a 20-item self-rating scale assessing frequency of symptoms.
A child with a score of 16 or higher was categorized
as displaying depressive symptoms (Radloff 1977).
The Multidimensional Scale of Perceived Support
(MSPSS) is a 12-item scale measuring perceived social
support from family, friends and a significant other
(Chronbach α = 0.93) (Zimet et al. 1990; Canty-Mitchell
and Zimet 2000; Bruwer et al. 2008).
To assess the extent to which students feel connected
to their school, we used selected items from the Healthy
Kids Resilience Measure of School Connectedness that
measure students’ perceived connectedness with adults
at their school and the strength of these relationships
(Constantine et al. 1999). All items in this scale showed
strong internal consistency (Cronbach α = 0.87).
Age, gender and ethnicity, potential confounders identified
from prior research, were also collected from students
at baseline (Davis and Siegel 2000).
Data analysis
For the implementation evaluation, we performed descriptive
analysis to describe the ease/difficulty and perceived
helpfulness of LPC steps and materials. For open-ended
questions about barriers, we used content analysis to identify
and create categories of themes.
For our outcome evaluation, we analyzed changes in
psychological symptoms, social support and school connectedness
over time. To control for the correlation
among longitudinal responses collected on the same student
and among responses collected within the same
school, we first fit hierarchical mixed effects linear regression
models that included random effects to induce
clustering at both the student level and the school level.
Compared with standard repeated measures ANOVA, the
hierarchical mixed effects model is a more flexible approach
to account for irregular time measurement points,
missing observations and time-dependency (Gueorguieva
and Krystal 2004). We fit our initial models with an autoregressive
correlation structure at the student level to
allow for the magnitude of the correlation between two
measurements to depend on the time period between the
measurements. (For example, observations taken at 2-
and 4- weeks follow-up are assumed to be more
highly correlated than measurements taken at 2- and
8-weeks follow-up.) At the school level, we employed
an exchangeable correlation structure. However, based on
the variance component estimates for these models, there
was no evidence of school-level clustering. Therefore, we
used a simpler mixed effects model that accounted only
for the correlation among longitudinal responses collected
on the same student. Age, gender, ethnicity and types of
trauma were included in the model as covariates and