Taking a symptom level approach to adolescent depression (Defended May 2019)

Abstract: Many individuals experience their first major depressive episode in adolescence (Merikangas et al., 2010), and even sub-clinical depressive symptoms in adolescents predict maladaptive outcomes (Bertha & Bal√°zs, 2013). Still, sum scoring all of the items on a depression scale to determine severity, can be problematic psychometrically (Fried, van Borkulo, et al., 2016) and conceptually (Borsboom, 2008). Therefore, examining how individual depression symptoms relate to depression relevant outcomes appears to be an appropriate approach (Fried & Nesse, 2015b). However, no systematic evaluation of how individual depression symptoms relate to important outcomes has occurred in an adolescent sample.

In Studies 1 and 2, I systematically evaluated which depressive symptoms may be more important during adolescence using network analysis (Borsboom & Cramer, 2013), a method that quantifies which symptoms are more central to a disorder (Valente, 2012). I found that self-hatred, sadness, loneliness, and pessimism were the most central symptoms in a large (N = 1,409), diverse (63.38% Non-Caucasian) sample of adolescents, and all assessed central symptoms replicated in a separate sample (N = 1,059) using a different measure of depression. Further, the more central a depression symptom was in the network the more variance it shared with life satisfaction (r = 0.59, 95% CI: 0.27, 0.76).

In Study 3, I explicitly evaluated the replicability of strength centrality in cross-sectional adolescent depression networks using KL divergence and a pseudo R2 metric. Compared to a null model (R2 = 0) and a perfect replication (R2 = 1), cross-sectional network strength centrality replicated well (R2 = 0.80) across two samples (Ns = 1,159, 402) of non-selected adolescents. Using a novel application of bootstrapping, I found cross-sectional network strength centrality of adolescent depression in one sample predicts change over time centrality of adolescent depression in another sample (r = 0.81, 95% CI: 0.45, 0.92).

My findings indicate more central adolescent depression symptoms as determined by network analysis, self-hatred, sadness, and loneliness, may be more important for depression related outcomes than more peripheral symptoms. Further investigating these central symptoms may more effectively advance our theoretical and clinical understanding of adolescent depression relative to investigating other symptoms.

Michael Mullarkey
PhD Candidate in Clinical Psychology and Clinical Intern

My research interests include developing brief interventions, predicting treatment response, and symptom level analysis


An example talk using Academic’s Markdown slides feature.