Predictors of outcome after a time-limited psychosocial intervention for adolescent depression

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Predictors of outcome after a time-limited psychosocial intervention for adolescent depression © 2022 Parhiala, Marttunen, Gergov, Torppa and Ranta Published version Parhiala, Pauliina; Marttunen, Mauri; Gergov, Vera; Torppa, Minna; Ranta, Klaus Parhiala, P., Marttunen, M., Gergov, V., Torppa, M., & Ranta, K. (2022). Predictors of outcome after a time-limited psychosocial intervention for adolescent depression. Frontiers in Psychology, 13, Article 955261. https://doi.org/10.3389/fpsyg.2022.955261 2022 fpsyg-13-955261 November 2, 2022 Time: 10:36 # 1 TYPE Original Research PUBLISHED 02 November 2022 DOI 10.3389/fpsyg.2022.955261 OPEN ACCESS EDITED BY Daniel Rodriguez, La Salle University, United States REVIEWED BY Meredith Kneavel, La Salle University, United States Jan Becker, Johannes Gutenberg University Mainz, Germany *CORRESPONDENCE Pauliina Parhiala [email protected] SPECIALTY SECTION This article was submitted to Psychology for Clinical Settings, a section of the journal Frontiers in Psychology RECEIVED 28 May 2022 ACCEPTED 10 October 2022 PUBLISHED 02 November 2022 CITATION Parhiala P, Marttunen M, Gergov V, Torppa M and Ranta K (2022) Predictors of outcome after a time-limited psychosocial intervention for adolescent depression. Front. Psychol. 13:955261. doi: 10.3389/fpsyg.2022.955261 COPYRIGHT © 2022 Parhiala, Marttunen, Gergov, Torppa and Ranta. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Predictors of outcome after a time-limited psychosocial intervention for adolescent depression Pauliina Parhiala1,2*, Mauri Marttunen1,3, Vera Gergov1,4, Minna Torppa5and Klaus Ranta2,4 1Department of Adolescent Psychiatry, Helsinki University Hospital and University of Helsinki, Helsinki, Finland, 2Faculty of Social Sciences, University of Tampere, Tampere, Finland, 3Department of Public Health Solutions, Finnish Institute for Health and Welfare in Finland (THL), Helsinki, Finland, 4Faculty of Medicine, University of Helsinki, Helsinki, Finland, 5Department of Teacher Education, University of Jyväskylä, Jyväskylä, Finland Research on the predictors of outcome for early, community-based, and time-limited interventions targeted for clinical depression in adolescents is still scarce. We examined the role of demographic, psychosocial, and clinical variables as predictors of outcome in a trial conducted in Finnish school health and welfare services to identify factors associating to symptom reduction and remission after a brief depression treatment. A total of 55 12–16-year-olds with mild to moderate depression received six sessions of either interpersonal counseling for adolescents (IPC-A) or brief psychosocial support (BPS). Both interventions resulted in clinical improvement at end of treatment and 3- and 6-month follow-ups. Main outcome measures were self-rated BDI- 21 and clinician-rated Adolescent Depression Rating Scale (ADRSc). Latent change score (LCS) models were used to identify predictors of change in depressive symptom scores and clinical remission at end of treatment and 3- and 6-month follow-ups over the combined brief intervention group. Symptom improvement was predicted by younger age and having a close relationship with parents. Both symptom improvement and clinical remission were predicted by male gender, not having comorbid anxiety disorder, and not having sleep difficulties. Our results add to knowledge on factors associating with good treatment outcome after a brief community intervention for adolescent depression. Brief depression interventions may be useful and feasible especially for treatment of mild and moderate depression among younger adolescents and boys, on the other hand clinicians may need to cautiously examine sleep problems and anxiety comorbidity as markers of the need for longer treatment. KEYWORDS adolescent, depression, brief intervention, school mental health services, symptom improvement, predictors Frontiers in Psychology 01 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 2 Parhiala et al. 10.3389/fpsyg.2022.955261 Introduction Depression is highly prevalent and impairing disorder in adolescents across the globe, yet it often goes unrecognized and undertreated (Haarasilta et al.,2003;Merikangas et al., 2010;Jörg et al.,2016;Zuckerbrot et al.,2018). Left untreated, depression adds to the risk of functional impairment and compromised physical and mental health in adulthood (Zisook et al.,2007;Thapar et al.,2012). Research suggests that intervening early to symptoms of depression is associated with higher effectiveness (Horowitz and Garber,2006,Bertha and Balázs,2013) and cost-effectiveness (Chisholm et al.,2016) of interventions. Access to mental healthcare, implementation of mental health programs and insufficient mental health policies challenges the provision of early interventions for clinical depression in adolescents (Rocha et al.,2015). Due to limited knowledge on outcomes and predictors of outcomes of early interventions research on them seems warranted (Davey and McGorry,2019). Structured psychotherapies are considered first-line interventions for adolescent depression (NICE,2019). Short term cognitive behavioral therapy (CBT) and interpersonal psychotherapy for adolescents (IPT-A) have gained most research support of efficacy for adolescent depression (Birmaher et al.,2007;Weisz et al.,2013,2017;Zhou et al.,2015;Pu et al., 2017). Dissemination of such treatments and use in primary care and community services is often called for, because structural and organizational factors in public healthcare systems often prevent delivery of longer treatments, which often require extensive mental health training for the professionals (Flaherty et al.,2021). Preliminary evidence of feasibility and positive pre-to-post treatment effect for depression in adolescents have been found for short forms of interpersonal therapy, i.e., Interpersonal Counseling for Adolescents, IPC-A (Wilkinson et al.,2018; Parhiala et al.,2020), and Brief Interpersonal Therapy for Adolescents, BIPT-A (Mufson et al.,2015) in community settings. National and professional guidelines suggest using supportive counseling either in individual or group format in primary care settings and schools for treatment of mild depression in adolescents (Lewandowski et al.,2013;Cheung et al.,2018;NICE,2019). Even when adequately administered, interventions for depression are effective for only 50–70% of treated adolescents, and about 30–40% achieve remission (e.g., Emslie et al.,2003; March et al.,2004;Thapar et al.,2012;Weisz et al.,2017). Identification of individual and psychosocial factors associated with good treatment outcomes for interventions of varying length is important to policymakers and clinical directors for allocating financial resources and for constructing stepped care treatment models (Berger et al.,2021). Effective use of available resources requires knowledge on for whom brief interventions are likely to be sufficient, and who need longer treatments. In clinical intervention research, pre-treatment factors associated with treatment outcome, independent of the treatment condition given, are called predictor variables (Barber,2009). Predictors can be grouped under broader categories in several ways (e.g., Vousoura et al., 2021) with one categorization presented below. Demographic variables Clinical research shows that age, gender, and family socioeconomic status might influence psychotherapy treatment outcomes for adolescent depression. In a recent meta-analysis of psychotherapies for depression, Cuijpers et al. (2020) found that children under the age 13 showed poorer treatment outcomes than adolescents between ages 13 and 18 regardless of the type of therapy. Contrasting this, an earlier meta-analysis (Weisz et al., 2006) found age was not significantly associated with outcome of psychotherapy of child and adolescent depression. Results from trials of CBT, supportive therapy (Clarke et al.,1995;Brent et al.,1998;Jayson et al.,1998;Curry et al.,2006), and family therapy (Brent et al.,1998;Diamond et al.,2019) among depressed adolescents suggest that younger adolescents may have stronger response to psychotherapy than older adolescents. In contrast, Mufson et al. (2004) found symptomatic improvement greater in older, compared with younger adolescents after IPT-A. Most meta-analyses of adolescent depression trials have not found gender to have an effect on treatment outcome (Clarke et al.,1992;Curry et al.,2006;Weisz et al.,2006;Courtney et al., 2022). However, in an early meta-analysis, Weisz et al. (1995) reported adolescent girls to benefit more from psychotherapy than boys, while among prepubertal children the gender effect was not found. The scoping review by Courtney et al. (2022) suggests that socioeconomic status is not a powerful predictor of outcome in adolescent depression. Psychosocial variables As parental behaviors and family problems commonly associate with depression in the young (Feeny et al.,2009), family factors can be expected to have influence on treatment outcomes. Indeed, living in a single-parent household was found to be a predictor of poorer depression treatment outcome in a study by Brent et al. (1998). Furthermore, maternal report of parent–adolescent conflict (Feeny et al.,2009), high family conflict (Asarnow et al.,2009), impairment in social functioning (Jayson et al.,1998), and high overall social dysfunction within the family (Gunlicks-Stoessel et al.,2010), have been found to Frontiers in Psychology 02 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 3 Parhiala et al. 10.3389/fpsyg.2022.955261 predict poorer outcomes for adolescent depression treatment, irrespective of the type of treatment received. Clinical variables Studies examining the effects of clinical variables as predictors of treatment outcomes with depressed adolescents have found baseline severity of depression to predict poorer treatment outcomes across different types of interventions (Brent et al.,1998;Jayson et al.,1998;Mufson et al.,2004; Asarnow et al.,2009;Wilkinson et al.,2009). According to a review by Nilsen et al. (2013), majority of reviewed studies found high severity of depression at baseline to predict poorer treatment response. In the review by Emslie et al. (2011) depressed adolescents having any comorbid psychiatric disorder had poorer outcomes in the included depression treatment studies. Of comorbid disorders, anxiety disorders were the most common predictors of poor treatment outcome. It have been found to predict poorer treatment outcome irrespective of severity of depression in CBT and IPT-A trials (Curry et al.,2006;Young et al.,2006; Wilkinson et al.,2009) and in a family therapy trial (Diamond et al.,2019). In two studies, comorbid anxiety disorders did not predict treatment outcome for adolescent depression (Jayson et al.,1998;Rohde et al.,2001). In the Rohde et al. (2001) study, higher depression severity in those with comorbid disorders ruled out the effect of anxiety disorders. Sleep difficulties commonly precede depressive episodes in adolescents and predict onset of depression in longitudinal studies (Lovato and Gradisar,2014). According to Kennard et al. (2006) sleep disturbance was one of the most common residual symptoms in adolescents who were fully or partly remitted from MDD after receiving CBT. Yet, sleep difficulties have relatively rarely been studied as a predictor of adolescent depression treatment outcome. In adolescents treated with fluoxetine, Emslie et al. (2012) found that pretreatment insomnia had a significant negative impact on treatment response and remission. In a continuation study of youth with MDD who had responded to acute treatment with fluoxetine (Kennard et al.,2018), residual insomnia after acute treatment predicted almost sevenfold risk of relapse. In a trial of IPT-A and routine treatment McGlinchey et al. (2017) found sleep disturbance to predict worse outcome in adolescent depression irrespective of treatment type. To sum, research on predictors of outcome from trials of psychosocial treatments for adolescent depression show heterogeneous results. Most previous studies have been conducted in university clinics or specialized mental health clinics, generally treating adolescents with severe disorders. It is not clear whether factors associated with positive treatment outcomes are the same when adolescent depression is treated in community, or in school health and welfare services with timelimited interventions, by professionals not having extensive background training in mental health. To add to the literature, we studied predictors of outcome in a clinical trial comparing interpersonal counseling for adolescents (IPC-A) with brief psychosocial support (BPS), delivered in school health and welfare services (Parhiala et al., 2020). In the trial both treatments were effective with no statistically significant differences between the treatments on outcome. Both treatments were brief (i.e., six sessions) and feasible to implement in community services. Thus, IPC-A and BPS treatment arms were combined in the predictor analyses, and treatment modality (IPC-A or BPS) was used as a covariate in the analyses. The aim of this study was to examine selected baseline demographic, psychosocial and clinical variables, identified from previous research, as possible predictors of outcome. Based on extant literature we expected that younger age would predict better treatment outcome and gender would have no effect on the outcome. We further expected that positive relationship between the adolescent and parents would be associated with a positive outcome. Third, we hypothesized that adolescents with milder depression, no comorbid anxiety disorder and no sleep difficulties at baseline would respond better to a brief intervention. Materials and methods Procedure and recruitment The trial from which our data is drawn compared two brief interventions for adolescent depression, Interpersonal counseling for adolescents (IPC-A) and brief psychosocial support (BPS) in Finnish school health and welfare services (Parhiala et al.,2020). All Finnish secondary schools provide student health and welfare services, their staff consisting of school psychologists, social workers, and nurses. In these services, adolescents are provided psychosocial support on as need basis. Those with need of prolonged support or identified mental health disorders are typically referred to specialized mental health services. A cluster randomization design was used. The participating schools were randomized to provide six sessions of either IPC-A or BPS. Outcome measures were given at baseline (session 1), mid-treatment (session 4), end of treatment (session 6), and follow-up meetings (3- and 6-month follow-up). The recruitment process followed routine practice for adolescents to obtain support from school services. Participants were screened for eligibility to the study using the Finnish modification of the 13-item Beck Depression Inventory, R-BDI (Beck and Beck,1972;Raitasalo,2007). Those who screened positive (R-BDI sum score >5) and consented were referred to Frontiers in Psychology 03 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 4 Parhiala et al. 10.3389/fpsyg.2022.955261 diagnostic interview. For a more detailed description (e.g., flow chart, study design, referral process), see Parhiala et al. (2020). Participants Fifty-five 12–16-year-old students were recruited from the student health and welfare services of the public lower secondary schools of a city of approximately 250,000 inhabitants in Southern Finland. Of the participants, 43 were girls and 12 were boys. Their mean age (SD) was 14.53 (0.78) years. Measures Symptom measures The Beck Depression Inventory (BDI-21; Beck et al., 1961) was used as a measure of self-reported depressive symptoms. It has demonstrated good psychometric properties in previous studies in adolescents (Brooks and Kutcher,2001; Myers and Winters,2002). In the present study, internal consistency of BDI-21, measured by Cronbach’s alpha (α) was 0.89. The Adolescent Depression Rating Scale (ADRSc; Revah- Levy et al.,2007) was used as the measure of clinician-rated depression symptoms. The ADRSc was administered by the school professionals delivering the interventions. According to a previous study, ADRSc has good convergent, discriminant, and factorial validity to assess depression in adolescents (Revah-Levy et al.,2007). In the present study, internal consistency of ADRSc was 0.80. Both measures were administered at baseline (session 1), in mid-treatment (session 4), and at the end of treatment (session 6), and in both follow-up meetings (3- and 6-months after the end of treatment). Diagnostic interview The Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS-PL; Kaufman et al.,1997), version for DSM-5 (K-SADS-5), was administered by a clinical psychologist to assess present and lifetime mental health disorders according to Diagnostic and Statistical Manual of Mental Disorders (DSM-5). Diagnostic evaluation was administered during baseline and again at 3- and 6- month follow-ups. The psychometric properties of the instrument (DSM-IV) have been good (Kaufman et al.,1997). All adolescents receiving a diagnosis of mild or moderate MDD, dysthymia, or depressive disorder not otherwise specified, were offered to be included in the study. Exclusion criteria were severe psychiatric disorder, ongoing psychiatric treatment, and an acute need for child protection services. Four adolescents met the exclusion criteria (one severe major depression, one acute need for child protection services, one with primary and severe anxiety disorder, one with a psychotic disorder) and were referred to a service they needed. Treatments Interpersonal counseling is a brief individual treatment focusing on current symptoms of depression in an interpersonal context (Weissman and Verdeli,2013). In this study, IPC was delivered in six 45-min sessions over a 6–12-week period. The treatment was administered according to the procedures specified in the IPC treatment manual (Weissman and Verdeli, 2013), and its adaptation for adolescents (IPC-A; Wilkinson and Cestaro,2015). A 3-day IPC-A training was given to all school health and welfare professionals at schools randomized to give IPC-A prior to onset of study. In addition, professionals got ongoing clinical method-specific supervision every second week for the duration of the trial. Brief psychosocial support was based on the methods and techniques used by the school health and welfare professionals in their routine work. At BPS sites, the professionals delivered BPS without specific methodological training. However, they were instructed to target the intervention to symptoms of depression, and to assess depressive symptoms repeatedly. To ensure comparability across treatments, BPS was delivered with the same frequency and session duration as IPC-A. Before the trial, all participating school health and welfare professionals were given 1-day training course in the identification and assessment of depression and the use of assessment measures included. Professionals in both treatment arms were instructed to assess and monitor symptoms of depression in their adolescent patients systematically and repeatedly. Predictors of treatment outcome Due to the relatively small sample size, we limited potential predictors to the most relevant based on previous literature. We ended up to seven putative predictor variables in three categories: demographic variables, psychosocial variables, and clinical variables. These baseline predictor variables are described in Table 1. Age and gender were used as demographic variables. Of psychosocial variables, we included family constellation (living with one or both biological parents) and the closeness of adolescent’s relationship with parents. Adolescents were asked about their relationship with parents by a question “how do you perceive your relationship with your mother and father at the moment.” When the adolescent reported both relationships were close (e.g., warm, easy to talk with parent), or close enough (e.g., no problems but not talking about everything with parent) this variable was coded as “close.” If the adolescent reported that the relationship was close with one parent, but not with the other (e.g., frequent conflicts, not feeling good about sharing feelings with parent), or if the adolescent felt the relationship was not close with either parent, the variable was coded as “not close.” Frontiers in Psychology 04 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 5 Parhiala et al. 10.3389/fpsyg.2022.955261 TABLE 1 Predictor variables: Descriptive characteristics at baseline, post-treatment, and follow-ups. Variable Baseline N= 55 Post-treatment N= 51 3-month follow-up N= 49 6-month follow-up N= 49 Demographic variables Age: mean (SD) 14.53 (0.78) 14.52 (0.73) 14.48 (0.71) 14.48 (0.71) Female gender, n(%) 43 (78.2%) 40 (78.4%) 39 (79.6%) 39 (79.6%) Psychosocial variables Family constellation (living with both biological parents), n(%) 28 (50.9%) 27 (52.9%) 26 (53.1%) 26 (53.1%) Close relationship with parents, n(%) 41 (74.5%) 40 (78.4%) 39 (79.6%) 39 (79.6%) Clinical variables Comorbid anxiety disorder, n(%) 16 (29.1%) 15 (29.4%) 15 (30.6%) 15 (30.6%) Sleep difficulties, n(%) 10 (18.2%) 10 (19.6%) 9 (18.4%) 9 (18.4%) Self-reported depressive symptoms, BDI-21, mean (SD) 17.47 (9.12) 10.54a(9.00) 11.21a(11.29) 7.56a(8.40) Clinician-reported depressive symptoms, ADRSc, mean (SD) 16.31a(7.78) 9.47 (8.16) 10.80 (8.58) 7.57 (7.05) BDI-21, Beck Depression Inventory; ADRSc, Adolescent Depression Rating Scale, clinician version. aData missing in one case. Of clinical variables, we included baseline severity of depressive symptoms, comorbid anxiety disorders, and sleep difficulties. Severity of baseline depressive symptoms was defined using both the continuous BDI-21 score and the ADRSc score. The BDI-21 scores were categorized into three groups according to symptom severity: (1) no/minimal depressive symptoms (0–9 points), (2) mild depressive symptoms (10–18 points), (3) moderate depressive symptoms (19 points or more) (Beck et al.,1988). The ADRSc baseline scores were classified into three severity levels: 1. no clinical depression (0–14 points), clinical depression (15–19 points), severe depression (20 points or more), as defined by Revah-Levy et al. (2007). The presence of comorbid anxiety disorders and sleep difficulties were drawn from the K-SADS-5 interview. The presence of sleep difficulties was defined as either initial, middle, or terminal phase difficulty in getting to sleep or staying asleep, or hypersomnia. The participant was classified as suffering from sleep difficulties if he/she reported symptoms nearly every night (i.e., five to seven nights per week), including any type of sleep symptoms. Statistical analysis The relationship between each predictor variable and observed change in depressive symptoms were tested with Latent Change Score (LCS) models. In the LCS models (Figure 1) two assessments of the observed outcome variables (BDI-21 and ADRSc), were included in each model. Therefore there are three models for BDI-21 and ADRSc. In the LCS models 1, the change between baseline and post-treatment (post) was modeled. In the LCS models 2, the change between post-treatment and 3-month follow-up (3-mo) was modeled, and in the LCS models 3, the change between 3- and 6- month follow-up (6-mo) was modeled. In the LCS models, the change between the two timepoints is modeled as latent variable. Predictors of the latent change were baseline variables including all tested variables in a certain model (intervention group, age, gender, family constellation, relationship with parents, anxiety disorders, sleep difficulties). The autoregressive parameters and factor loadings were fixed to one (marked as * in Figure 1). The standardized beta values for the models are reported in Table 2 and in Supplementary Table 1. The LCS models were run for both BDI-21 and ADRSc scores to explore the person-to-person variability in the change of depressive symptoms during the intervention and follow-up points. In the initial LCS model, all predictor variables were included in the model: the demographic variables (age, gender), the psychosocial variables (family constellation, relationship with parents), and the clinical variables (baseline severity of depression symptoms, comorbid anxiety disorder, sleep difficulties). Intervention type (IPC-A or BPS) was used as a covariant in the analyses. In the second, final LCS model, only the statistically significant predictor variables were included. FIGURE 1 Latent change score (LCS) model specification. Frontiers in Psychology 05 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 6 Parhiala et al. 10.3389/fpsyg.2022.955261 TABLE 2 The standardized estimates of the final latent change score (LCS) model for Beck Depression Inventory (BDI)-21 and Adolescent Depression Rating Scale (ADRSc). Post BDI-21 β(s.e.) R2 3-mo BDI-21 β (s.e.) R2 Post ADRSc β(s.e.) R2 3-mo ADRSc β (s.e.) R2 Depressive symptoms –0.46 (0.08)*** 0.21 –0.66 (0.08)*** 0.44 –0.57 (0.09)*** 0.32 –0.76 (0.09)*** 0.58 Intervention group 0.07 (0.11) 0.00 0.08 (0.10) 0.01 0.16 (0.10) 0.03 0.00 (0.10) 0 Age 0.29 (0.13)* 0.08 Gender 0.36 (0.10)** 0.13 0.25 (0.09)** 0.06 Close relationship with parents –0.22 (0.10)** 0.05 –0.17 (0.10) 0.03 Comorbid anxiety disorder –0.24 (12)* 0.06 Sleep difficulties 0.39 (0.11)*** 0.15 R2 0.37 0.50 0.49 0.60 Model fit χ2(3) = 4.27, p>0.05, RMSEA = 0.09, CFI = 0.96, SRMR = 0.10 χ2(2) = 1.80, p>0.05, RMSEA = 0.00, CCFI = 1.00, SRMR = 0.09 χ2(4) = 8.68, p>0.05, RMSEA = 0.15, CFI = 0.90, SRMR = 0.16 χ2(2) = 0.36, p>0.05, RMSEA = 0.00, CFI = 1.00, SRMR = 0.04 Post, change in depression score from baseline to post-treatment; 6-mo, change in depression score from 3- to 6 month follow-up; BDI-21, Beck Depression Inventory; ADRSc, Adolescent Depression Rating Scale clinician version; β, standardized estimate for regression; s.e., standard error; R2, amount of explained variance. *p<0.05, **p<0.01, ***p<0.001. Due to sample size, the LCS models were run separately for the three separate time periods: Post (change in depression score from first treatment session to last session), 3-mo (change in depression score from last session to 3-month follow-up), and 6-mo (change in depression score from 3- to 6-month followup). One-way analyses of variance (ANOVA) were conducted to compare gain-scores (i.e., decrease or increase of the depression score during Change period) in groups categorized according to baseline BDI-21 and ADRSc throughout the treatment and follow-ups. Direction of change was analyzed using Dunnett’s correction separately for BDI-21 and ADRSc. Last, we analyzed whether the chosen baseline variables predicted clinical remission according to BDI-21 and ADRSc at post-treatment, at 3- and at 6-month follow-up, using Chisquare tests for nominal variables and t-tests for continuous data (i.e., age). We also checked whether depression symptom scores were different between the two comparison groups within each predictor variable already at baseline, using t-tests. In these analyses, nominal variables were compared, and continuous variable (i.e., age) was divided into two groups using the mean value. Clinical remission was defined as the absence of clinically significant depressive symptoms score <10 in BDI (Beck et al., 1988) and score <15 in ADRSc (Revah-Levy et al.,2007). Missing data were imputed by carrying the last observation forward until the sixth session if the adolescent had at least one completed BDI-21 or ADRSc after baseline. Data analyses were carried out using SPSS (version 22.0) and Mplus (version 8; Muthén and Muthén,1998-2017) programs. Results Descriptive data on baseline predictor variables are presented in Table 1. Examination of depressive symptoms at baseline showed that participants suffered from moderate depressive symptoms according to BDI-21 scores (M= 17.47, SD = 9.12). According to ADRSc sores their symptoms were in the clinical depression range (M= 16.10, SD = 7.78). As shown by standard deviations, variation in scores was high. Depression scores decreased during the intervention, but variation between participants remained large through the duration of intervention and follow-ups (see Table 1). Four participants dropped out from the treatment after the third session, and one participant did not answer the BDI- 21 questionnaire after the baseline. Therefore, the number of participants included in the BDI-21 analyses was 50 at the end of treatment, for ADRSc it was 51. Two participants dropped out after the fifth session and one participant did not answer the BDI-21 questionnaire at the 3-month follow-up. Thus, the number of adolescents included in the 3-mo analyses of BDI-21 is 48 and 49 in the ADRSc analyses. Predicting change in depressive symptoms during the intervention The initial LCS models including all predictor variables is presented in Supplementary Table 1. Separate models were run for BDI-21 and ADRSc scores. Table 2 presents the final LCS models including only the statistically significant predictors and the intervention type as a covariant, separately for BDI-21 and ADRSc scores. The amount of explained variance in the change factor is reported in the above tables. Note that the amount of unique variance explained by each predictor equals to the squared standardized path estimates (betas). Change in self-reported depressive symptoms (Beck Depression Inventory-21) The results from LCS models (Table 2 and Supplementary Table 1) showed that the previous BDI-21 score significantly predicted the subsequent BDI-21 score at each of the three time periods examined (i.e., post, 3-mo, 6-mo). A larger decrease in BDI-21 was found for adolescents with higher BDI-21 baseline scores (Figure 2). In the post model, Frontiers in Psychology 06 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 7 Parhiala et al. 10.3389/fpsyg.2022.955261 a larger decrease in BDI total score was predicted by younger age and male gender. The initial model for 3-mo resulted in no significant predictors of BDI change. The model for 6-mo indicated that a larger decrease in BDI-21 total score between the 3- and 6-month follow-up assessments was predicted by having close relationships with parents. The initial model for post and 6-mo had insufficient model fit, as there were too many variables in comparison to the small number of adolescents, but the final models for post and 6-mo fitted the data well (Table 2 and Supplementary Table 1). The initial model for 3-mo did not fit the data well, as none of the examined predictors explained the change. Thus, final model for 3-mo was not run. Change in clinician-rated depressive symptoms (Adolescent Depression Rating Scale) The results from the LCS models (Table 2 and Supplementary Table 1) showed that the previous ADRSc score significantly predicted the subsequent ADRSc score in post and 6-mo models. A larger decrease in ADRSc was found for adolescents with higher ADRSc baseline scores (Figure 3). In the post model, a larger decrease in ADRSc total score was FIGURE 2 Change in clinical severity groups according to Beck Depression Inventory (BDI)-21-scores from baseline to 6-month follow-up. FIGURE 3 Change in clinical severity groups according to Adolescent Depression Rating Scale (ADRSc)-scores from baseline to 6-month follow-up. Frontiers in Psychology 07 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 8 Parhiala et al. 10.3389/fpsyg.2022.955261 predicted by male gender and not having sleep difficulties. Not having comorbid anxiety disorder was almost significant (p= 0.059) in the initial LCS model including all predictor variables. It was therefore also included in the final model and predicted change of ADRSc sum score in the final model. In the initial LCS model for 6-mo close relationship with parents was almost significant (p= 0.054) and was included in the final model. In the final model it was not, however, a significant predictor. The model for 3-mo resulted in no significant predictors of ADRSc change. The initial models for post and 6-mo had insufficient model fit, as there were too many variables in comparison to the small number of adolescents, but the final models for post and 6-mo fitted the data well (Table 2 and Supplementary Table 1). The initial model for 3-mo did not fit the data well, as none of the predictors explained the change. Thus, the final model for 3-mo was not run. Change according to clinical severity The LCS models suggested that baseline depression scores predicted changes during the intervention and follow-up. To examine the effect more closely, we analyzed how the baseline scores categorized according to clinical severity of symptoms BDI-21: no/minimal symptoms, mild symptoms, moderate symptoms (Beck et al.,1988) ADRSc: no clinical depression, clinical depression, severe depression (Revah-Levy et al.,2007) affected the outcome at the end of treatment and follow-up. As can be seen from Figures 2,3, the depression scores of participants in the moderate symptoms (BDI-21) or severe depression (ADRSc) groups decreased rapidly. However, these adolescents still ended up having higher scores during the follow-up in comparison to those with lower scores at baseline. We compared the three baseline depression symptom and depression severity groups (see Section “Predictors of treatment outcome”) according to baseline BDI-21 and ADRSc using gain-scores in three analyses: (1) between baseline and posttreatment, (2) between post-treatment and 3-month follow-up, and (3) between 3- and 6-month follow-up, using one-way ANOVAs. The first gain-score analysis of BDI-21 scores showed a significant group difference between the symptom severity groups [F(2,47) = 5.51, p= 0.007]. Pairwise comparisons showed that BDI-21 scores decreased more in the moderate symptom severity group during the intervention (n= 22, mean gain = – 10.50, SD = 8.52) compared to both the no/minimal (n= 11, mean gain = –4.64, SD = 2.91) and mild (n= 17, mean gain = –3.23, SD = 7.14) symptom severity groups. In the second gain-score analysis of BDI-21 scores, a decrease or no change was observed in the mild (n= 17, mean gain = –3.88, SD = 5.44) and moderate (n= 21, mean gain = 0.67, SD = 5.69) symptom severity groups, compared with the no/minimal symptom severity group (n= 11, mean gain = 8.36, SD = 15.89), these differences being statistically significant [F(2,45) = 6.39, p= 0.004]. We found no symptom severity group differences in the third gain-score comparison of BDI-21 scores. Examining the change in symptoms assessed with ADRSc, significant differences emerged in the first gain-score comparison from baseline to post-treatment [F(2,48) = 6.98, p= 0.002]. Pairwise comparisons using Dunnett’s correction showed that ADRSc scores decreased more in the severe depression group during this period (n= 17, mean gain = – 11.19, SD = 7.58) when compared to no clinical depression (n= 11, mean gain = –4.82, SD = 1.67) and clinical depression groups (n= 23, mean gain = –4.61, SD = 5.65). No depression severity group differences were observed for the other two gain-score analyses. Predictors of remission from depression To identify predictors of remission from depression, we analyzed whether the selected baseline variables predicted remission according to BDI-21 (sum score <10; Beck et al., 1988) and ADRSc (sum score <15; Revah-Levy et al.,2007) at post-treatment, at 3- and at 6-month follow-up, and whether differences were already apparent at baseline. At post-treatment, 58% (29/50), at 3-month follow-up 56% (27/48), and at 6-month follow-up 67% (32/48) of participants achieved remission as defined by the BDI-21 total score. The respective rates of remission as defined by the ADRSc score were at post-treatment 78% (40/51), at 3-month follow-up 76% (37/49), and 86% (42/49) at 6-month follow-up. Examining baseline level of self-reported depressive symptoms associated with the predictor variables, we found that among participants with comorbid anxiety disorder, baseline BDI-21 scores were already higher than among those without comorbid anxiety disorder. Three baseline variables predicted belonging to the remission group according to BDI-21 on at least one of the examined time points. These variables were gender, comorbid anxiety disorder, and sleep difficulties (Table 3). Boys achieved remission more often than girls at post-treatment and at 3-month follow-up. The probability of remission was higher among participants with no comorbid anxiety disorder than among those with anxiety disorder at post-treatment, 3-month follow-up, and at 6-month follow-up. Adolescents without baseline sleep difficulties were more likely to achieve remission than those with sleep difficulties at post-treatment and at 3-month follow-up. Examining baseline level of depressive symptoms as defined by ADRSc, as associated with each of the predictor variables, we observed that boys’ ADRSc scores were lower than those of girls. In addition, among participants with comorbid anxiety disorder, baseline ADRSc scores were already higher than scores among those without comorbid anxiety disorder. One baseline variable predicted belonging to remission group according to ADRSc at least on one of the examined time points. Not having sleep difficulties predicted remission as defined by ADRSc score at Frontiers in Psychology 08 frontiersin.org fpsyg-13-955261 November 2, 2022 Time: 10:36 # 15 Parhiala et al. 10.3389/fpsyg.2022.955261 Rohde, P., Seeley, J. R., Kaufman, N. K., Clarke, G. N., and Stice, E. (2006). Predicting time to recovery among depressed adolescents treated in two psychosocial group interventions. J. Consult. Clin. Psychol. 74, 80–88. doi: 10.1037/ 0022-006X.74.1.80 Sun, M., Rith-Najarian, L. R., Williamson, T. J., and Chorpita, B. F. (2019). Treatment features ssociated with youth cognitive behavioral therapy follow-up effects for internalizing disorders: A meta-analysis. J. Clin. Child Adolesc. Psychol. 48, 269–S283. Thapar, A., Collishaw, S., Pine, D. S., and Thapar, A. K. (2012). Depression in adolescence. Lancet 379, 1056–1067. doi: 10.1016/S0140-6736(11)60871-4 Vousoura, E., Gergov, V., Tulbure, B. T., Camilleri, N., Saliba, A., Garcia-Lopez, L. J., et al. (2021). Predictors and moderators of outcome of psychotherapeutic interventions for mental disorders in adolescents and young adults: Protocol for systematic reviews. Syst. Rev. 10, 1–14. doi: 10.1186/s13643-021-01788-1 Weersing, V. R., Iyengar, S., Kolko, D. J., Birmaher, B., and Brent, D. A. (2006). Effectiveness of cognitive-behavioral therapy for adolescent depression: A benchmarking investigation. Behav. Ther. 37, 36–48. doi: 10.1016/j.beth.2005.03. 003 Weissman, M. M., and Verdeli, H. (2013). Interpersonal counseling. New York, NY: University of Colorado. Weisz, J. R., Kuppens, S., Eckshtain, D., Ugueto, A. M., Hawley, K. M., and Jensen-Doss, A. (2013). Performance of evidence-based youth psychotherapies compared with usual clinical care: A multilevel meta-analysis. JAMA Psychiatry 70, 750–761. doi: 10.1001/jamapsychiatry.2013.1176 Weisz, J. R., Kuppens, S., Ng, M. Y., Eckshtain, D., Ugueto, A. M., Vaughn- Coaxum, R., et al. (2017). What five decades of research tells us about the effects of youth psychological therapy: A multilevel meta-analysis and implications for science and practice. Am. Psychol. 72, 79–117. doi: 10.1037/a0040360 Weisz, J. R., McCarty, C. A., and Valeri, S. M. (2006). Effects of psychotherapy for depression in children and adolescents: A meta-analysis. Psychol. Bull. 132, 132–149. doi: 10.1037/0033-2909.132.1.132 Weisz, J. R., Weiss, B., Han, S. S., Granger, D. A., and Morton, T. (1995). Effects of psychotherapy with children and adolescents revisited: A meta-analysis of treatment outcome studies. Psychol. Bull. 117, 450–468. doi: 10.1037/0033-2909. 117.3.450 Wilkinson, P., and Cestaro, V. (2015). Interpersonal Counselling (IPC) for Adolescents with Depression Symptoms. Unpublished manual. Wilkinson, P. O., Cestaro, V., and Pinchen, I. (2018). Pilot mixed-methods evaluation of interpersonal counselling for young people with depressive symptoms in non-specialist services. Evid. Based Ment. Health 21, 134–138. doi: 10.1136/ebmental-2018-300028 Wilkinson, P., Dubicka, B., Kelvin, R., Roberts, C., and Goodyer, I. (2009). Treated depression in adolescents: Predictors of outcome at 28 weeks. Br. J. Psychiatry 194, 334–341. doi: 10.1192/bjp.bp.108.052381 Young, J. F., Mufson, L., and Davies, M. (2006). Impact of comorbid anxiety in an effectiveness study of interpersonal psychotherapy for depressed adolescents. J. Am. Acad. Child Adolesc. Psychiatry 45, 904–912. doi: 10.1097/01. chi.0000222791.23927.5f Zhou, X., Hetrick, S. E., Cuijpers, P., Qin, B., Barth, J., Whittington, C. J., et al. (2015). Comparative efficacy and acceptability of psychotherapies for depression in children and adolescents: A systematic review and network meta-analysis. World Psychiatry 14, 207–222. 10.1002/wps. 20217 Zisook, S., Lesser, I., Stewart, J. W., Wisniewski, S. R., Balasubramani, G. K., Fava, M., et al. (2007). Effect of age at onset on the course of major depressive disorder. Am. J. Psychiatry 164, 1539–1546. doi: 10.1176/appi.ajp.2007.0610 1757 Zuckerbrot, R. A., Cheung, A., Jensen, P. S., Stein, R., Laraque, D., and GLAD-PC STEERING GROUP (2018). Guidelines for Adolescent Depression in Primary Care (GLAD-PC): Part I. Practice Preparation, Identification, Assessment, and Initial Management. Pediatrics 141:e20174081. doi: 10.1542/peds. 2017-4081 Frontiers in Psychology 15 frontiersin.org