Data Analysis

All analyses were conducted using Statistical Package for Social Sciences (SPSS) Version 27.0 (Corp IBM, 2017) and R Studio (R Studio Team, 2020).  To assess for missing Top Problem data, we performed Little’s MCAR test and found that data was missing completely at random (χ2(9151) = 7941.39, p = 1.00).  Since data were missing completely at random, we used pairwise deletion to account for data missingness, as it is less biased, preserves more information, and is an optimal solution when there is a low percentage of missing data (Kang, 2013).  We also assessed for missingness for the RCADS measure and performed Little’s MCAR test. We found that data was missing completely at random (χ2 (51) = 36.809, p = 0.932), with pre-treatment RCADS missing data ranging from 6% to 8.3% and post-treatment data ranging from 32.3% to 39.8%.  Here, we also performed a pairwise deletion to account for data missingness on the symptom measures.

For each treatment session, we averaged the TPA severity ratings collected to obtain a total average “Top Problems” score for each session.  This procedure was completed separately for parent- and youth-reported TPA ratings such that we generated a parent- and youth-reported average Top Problems score. This approach has demonstrated acceptable test-retest reliability and convergent validity while also allowing for the inclusion of the clients that had more than three problems (Milgram et al., 2021; Weisz et al., 2011).  

To investigate Aim 1, change in TPA severity over the course of treatment, we conducted a paired samples t-test to examine mean differences in the average ratings of the first and last session for both parent- and youth-reported TPAs.

To address Aim 2, we investigated clinical significance by computing the reliable change index (RCI) for each participant and determined the percentage of participants with meaningful change in TPA rating for youth and caregiver (Jacobson & Traux, 1991; Wise, 2004).  The RCI is calculated as a proportion of the difference in top problem ratings and the standard error of the difference score in our sample (Guhn et al., 2014).

To investigate Aim 3, we examined which demographic and clinical variables predicted this improvement in TPA ratings. To do so, we computed a TPA change score for each participant (computed as mean first session rating – mean last session rating).  Using linear regression, we regressed said TPA change score on demographic variables such as age, ethnicity, and gender.  We also examined if clinical variables, such as change in depression and anxiety (measured by the RCADS), early treatment response (ETR), defined as a change in severity rating of at least one standard deviation below the mean within the first four sessions, and mid-treatment response (MTR), defined as a change in severity rating of at least one standard deviation below the mean at session 8, occurring 4 sessions after the ETR (roughly 8 weeks), predicted TPA change for both caregiver and child using linear regression.  All statistical assumptions relevant to the aforementioned analyses were met.