Screening for and Personalizing Prevention of Adolescent Depression

Abstract

Depression is a prevalent, distressing, often recurrent, disorder. Adolescence represents a vulnerable developmental period when rates of depression surge and many experience their first episode. Some professional agencies now recommend universal screening starting at age 12. This paper advocates for a risk-based approach to screening for adolescent depression that can improve upon current approaches to screening and facilitate more seamless connections to enable personalizing prevention of depression based on risk group classification. Empirical examples are reviewed for screening based on established risk factors that predict later depression. Evidence is provided that risk groups can reliably and validly classify adolescents at risk for future development of depression based on cognitive and interpersonal vulnerabilities. These risk groups inform one approach to personalizing prevention of depression by matching youths’ risk to established, evidence-based prevention programs (cognitive or interpersonal). Promising data from a randomized trial suggest that this personalized depression prevention strategy can reduce depression better than a “one size fits all” approach.

Depression is common, debilitating, and costly. Depression is ranked as the number one most burdensome disease in the world for total disability-adjusted years and the leading cause of disability for ages 15–44 (WHO, 2012). Prevalence rates of depressive episodes surge during adolescence. Longitudinal research to ascertain depression trajectories among community youth (ages 8 to 17) showed that depression rates were low and stable until age 14 and then began to skyrocket in adolescence from 5% to about 20% by age 17 (see Figure 1; Hankin et al., 2015). The gender difference in depression (more girls than boys) emerged in early adolescence (ages 12–13). Depression is highly recurrent, and adolescent-onset depression increases risk 6-fold for recurrence in adulthood (Rutter, Kim-Cohen, & Maughan, 2006).

Figure 1. Developmental Trajectories of Depression from Childhood into Late Adolescence.

Results from an Accelerated Longitudinal Cohort Design, in which youth were repeatedly interviewed with diagnostic assessments to ascertain onset of depressive episodes (see Hankin et al., 2015 for details).

Given such facts, various agencies have called for screening youth for depression, so they can be connected to treatment for those currently depressed or prevention for those vulnerable youth. This paper advocates for a risk-based approach to screening of adolescent depression that can more seamlessly link identification of susceptible youth with interventions emphasizing personalized prevention that can enhance outcomes and reduce prevalence better than present “one size fits all” approaches.

DEPRESSION SCREENING

The United States Preventative Services Task Force (Siu et al., 2016) and the American Academy Pediatrics (Zuckerbrot et al., 2018) both recommend universal screening for depression in youth over age 12. Most screenings use assessments that measure the depression syndrome (e.g., CDI, PHQ-9) or broad internalizing (e.g., CBCL) (Stockings et al., 2015). Others recommend very brief screeners given limited time and competing demands (Lavigne, Feldman, & Meyers, 2016). Two items from the Patient Health Questionnaire (PHQ-2), tapping depressed mood and anhedonia, has been used. Yet, this approach does not fully capture depression’s heterogeneity and complexity, its multifactorial risk, nor directly assess suicidality which is important for screening. Still, screening by solely measuring a few symptoms, the full depression syndrome or broad internalizing and emotional problems has limitations (Wissow et al., 2013).

Another option is devising and evaluating screening measures based on risk factors that predict later depression symptoms and disorder in youth. Screening should provide incremental validity beyond exclusive use of current depression symptom levels. When selecting among many risks for youth depression (Hankin, 2012) that could be used in screening, optimal candidates should meet three criteria. First, the vulnerability can be reliably assessed in youth and predicts emergence of depression (symptoms and disorder) longitudinally after controlling for baseline depression levels. Relatedly, the risk screening should be robust and demonstrate reproducible results in independent samples so that clinically relevant decisions are soundly made regarding risk reliability and predictive validity with incremental forecasting beyond only current depression symptoms. Second, the risk should be relatively trait-like and stable over time by early adolescence, before most youth experience increases in depression, so the risk can forecast future depression; at the same time, the vulnerability should not be immutable, so it is amenable to change in interventions. Last, ideally the risk can be used not only for screening purposes but can also directly assess putative risk mechanisms targeted by interventions.

Cognitive vulnerabilities meet these three criteria (Hankin et al., 2016). These risks show promising psychometric properties (sensitivity and specificity) for screening and demonstrate robust, replicable incremental validity in predicting future depression beyond current depressive symptoms levels. For example, rumination and negative cognitive style predict prospective and recurrent episodes of depression, and rumination discriminates between currently clinically depressed and non-depressed compared to the standard depressive symptoms screening measure (Cohen et al., 2018). Other means to assess risk (e.g., passive tracking via smart phones or other digital information, data mining via archival records, implicit or task-based measures, psychophysiological indicators, biomarkers) can also be examined, but extant research does not yet met all three criteria, especially replication across independent samples with longitudinal forecasting and incremental validity that beats current depression levels.

Other research shows that multiple risk factors can be combined together to better predict future depression. Cohen and colleagues (2019) demonstrated that various risks, including cognitive vulnerabilities, stress, adversity exposure, and emotional predispositions, could be joined to create a screening protocol that reliably identifies youth at risk for developing first onsets of depression using two, independent, multi-wave longitudinal data sets (one to “test” and second for “validation”). Rumination, negative affect, and impairment in social and academic domains replicated as risks that predicted first depression onsets over a 2-year follow-up. These measures reliably formed an evidence-based algorithm that provided incremental validity beyond the present screening standard of youth self-report of depressive symptoms. Adolescents who scored highly on all three measures (rumination, negative affect, impairment) were approximately twice as likely to experience a first depressive episode over two years. For example, a 14-year old girl exhibiting high risk on all three measures was calculated to have 18.46% overall probability of a first depression relative to 8.33% for an average 14-year girl.

Ideally, evidence-based forecasting of risk can identify those at higher risk but also provide more direct information to connect youth to particular intervention options and inform which approaches may best reduce risk and propensity for depression. Additionally, surprisingly little research has examined how multiple psychosocial risks are structured and relate to each other. Together, such information could be translated to form evidence-based risk classification groups to enable screening for later depression and inform allocation of prevention efforts based on non-redundant risk information that could connect risk factors to intervention modalities to reduce depression prevalence and burden. Universal depression screening, based presently on current symptom measures, needs to do more than merely identify youth at risk, and even here, better and more incrementally valid approaches exist and supersede current depression screening (e.g., Cohen et al., 2019). Establishing a more seamless and tighter connection between screening and intervention is needed to elucidate who is most susceptible, inform who is most likely to benefit from particular interventions, and assist in allotment of precious healthcare resources (Garber, Korelitz, & Samanez-Larkin, 2012).

Hankin, Young, Gallop and Garber (2018) addressed these points focusing on a suite of theoretically-based, empirically supported cognitive and interpersonal risks to youth depression. Presently, there exist many cognitive (rumination, negative cognitive style, dysfunctional attitudes) and interpersonal (conflict, support, excessive reassurance seeking, corumination, social competence) vulnerabilities that each independently associate with and predict depression (Hankin, 2012; Hankin et al., 2016), yet the degree of overlap among these psychosocial risks, and their optimal organization, had not been evaluated. Cognitive and interpersonal risk theories suggest these risks are not redundant. For screening and risk identification, it is important to know about overlap and organization of risks. If multiple psychosocial measures substantially overlap and form one single latent risk factor, then there is no reason practically to separately assess them, as a shorter screening assessment could be devised to more simply evaluate that one latent psychosocial risk. Factor analyses showed that various cognitive variables loaded onto one factor, and interpersonal risks loaded onto two factors with one characterized by relationship conflict and the other by interpersonal support (Hankin et al., 2018). Thus, cognitive and interpersonal risk measures tap different vulnerabilities that convey nonredundant information that are important translationally for risk screening and intervention targets.

Next, to make screening simpler and clinically useful, a few measures marking these three latent factors were selected (Hankin et al., 2018). Feasible, practical cutoffs were created to classify whether any individual adolescent is at high or low risk on cognitive and interpersonal vulnerabilities, respectively. One goal in establishing these cutoffs was to identify scores which would lead to a relatively balanced number of adolescents in each of four “cells” (see Figure 2) for eventual testing of a personalized prevention approach (see later about the Personalized Depression Prevention—PDP—study). Factor structure and cutoff results replicated in an independent sample (similar aged adolescents for the PDP study). Last, validity of these risk group-based screening was supported: These high and low cognitive and interpersonal groups predicted future onset of depressive episodes (Hankin et al., 2018). The low/low group had the lowest rate (8%) of depressive episode onset over an 18-month follow-up, the high/high group had the highest rates (24%), and the other high/low off diagonal groups were in between (15–20%).

Figure 2. Cognitive and Interpersonal Risk Groups Classification.

(see Hankin et al., 2018 for details). The bolded off-diagonal boxes representing the high/low vulnerability groups are most salient for personalizing depression prevention by matching youths’ highest risk profile to a prevention program targeting that vulnerability to improve depression outcomes for adolescents.

In summary, this risk-based classification work (Hankin et al., 2018) shows that adolescents can be reliably, validly, and relatively easily identified as being at low or high cognitive and interpersonal risk, respectively. While this grouping approach is based on self-reported questionnaires to assess risk, past work highlights self-report as one valuable, valid data source (Samuel, Suzuki, & Griffin, 2016), and this research meets benchmarks proposed for identifying reliable, valid risk markers that could be clinically useful (Kapur et al, 2012). This information can be used in screening to forecast future clinical depression, and these relatively proportional risk classification groups can inform one approach to personalizing prevention of depression. Next is a description and initial results showing how this risk group classification can enable more seamless individualization to reduce future incidence of depression by matching adolescents’ risk profiles to the prevention modality that maximizes targeting of the depression vulnerability.

DEPRESSION PREVENTION

Depression prevention programs for adolescents work with significant, albeit modest effects compared to active controls (Hetrick, Cox, Witt, Bir, & Merry, 2016; Merry at al., 2011). Estimates suggest that 22% to 38% of depressive episodes could be prevented if current depression prevention programs were implemented (Cuijpers, van Straten, Smit, Mihalopoulos, & Beekman, 2008). Universal preventions are offered to all youth regardless of risk. Selective preventions are provided to those at risk, and indicated prevention focuses on those exhibiting subthreshold depression symptoms.

Cognitive-behavioral (CB) and interpersonal preventions have been developed and evaluated for youth; both show modest effects (Brunwasser & Garber, 2016). Coping with Stress (CWS; Garber et al., 2009) is a CB prevention that targets cognitive distortions, problem-solving difficulties, and lack of engagement in pleasurable activities. IPT-AST is an interpersonally-oriented prevention program (Young, Mufson, & Schueler, 2016) that teaches adolescents communication strategies and interpersonal problem-solving skills to reduce peer and family conflict and enhance social support.

The modest empirical support based on randomized trials for depression prevention have adopted a “one size fits all” approach that compare youth who are randomized to receive an intervention versus control. The present evidence-base for depression prevention has not empirically examined personalization of interventions in a systematic manner via a randomized trial using evidence-based risk classifications to match youth to a particular prevention program linked to their specific form of vulnerability (cognitive or interpersonal). This matching approach was taken in the recently completed Personalized Depression Prevention (PDP) trial. PDP is consistent with recent calls highlighting the potential of evidence-based precision medicine in both physical and mental health (Hamburg & Collins, 2010).

It may be possible to improve the relatively modest impact of depression prevention programs, which are currently based on a “one size fits all” model. One option to enhance the effects of existing interventions is personalizing who gets which prevention program. Existing evidence-based preventions might be particularly effective if there is a match between individuals’ risk (e.g. cognitive) and the techniques taught and targeted in the intervention (e.g., cognitive restructuring in CWS). The PDP study was designed to evaluate this possibility in a randomized trial of adolescents. Among its primary aims, PDP sought to evaluate: 1) bending longitudinal depression trajectories for adolescents who receive an evidence-based prevention (IPT-AST or CWS) relative to a natural history control group, and 2) determining whether matching youth on their particular risk to the preferred prevention group (e.g., high cognitive/low interpersonal risk receive CWS) do better (less depression over time) relative to those receiving a mismatch (e.g., high cognitive risk/low interpersonal risk receiving IPT-AST). In PDP, the bolded high/low off-diagonal groups in Figure 2 are most salient for testing this approach to personalizing prevention based on extant empirically supported programs. PDP focused on risk factors specifically targeted in either CWS (reducing negative thinking patterns) or IPT-AST (decrease conflict in close relationships and increase support).

PDP, a prospective randomized controlled trial of community adolescents, has recently been completed. Youth were screened at baseline using cognitive and interpersonal risk measures based on prior research establishing risk group classifications (Figure 2). Youth were randomized to receive either CWS or IPT-AST. This design enabled a rigorous test of the personalization aim and ensured that half of the adolescents with a high/low risk group classification received an empirically supported program that matched their elevated risk type (e.g, high cognitive/low interpersonal for CWS), while the other half received a mismatch (e.g., high cognitive/low interpersonal randomized to IPT-AST). Depressive symptoms and diagnostic episodes were assessed every six months after completion of CWS and IPT-AST preventions.

Results through the 18-months follow-up are promising (Young et al., 2019). Youth receiving prevention (either CWS or IPT-AST) had significantly reduced rates of prospectively assessed depressive episodes compared to a natural history control group of same-aged adolescents who received no intervention. As expected and consistent with past work (Horowitz et al., 2007), there was no difference between CWS and IPT-AST for all youth (i.e., risk groupings not considered). Importantly for testing personalization, when analyses were conducted based on youths’ risk group (see Figure 2, especially the off-diagonals that are critical to evaluate personalization) results showed that youth who received a match between their high risk (e.g., cognitive) reported significantly less depressive symptoms over follow-up relative to those who received a mismatch. No significant difference between CWS and IPT-AST was found for the low-low or high-high risk classification groups; it was only the high-low risk groups that showed a difference between CWS and IPT-AST when intervention modality (e.g., IPT-AST) matched their high risk (e.g., interpersonal).

SUMMARY

Adolescent depression is prevalent and burdensome. Risk-based screening protocols offer promise to identify those vulnerable for depression and perform better (incremental validity) than the current recommended standard using current depression symptom measures. Moreover, risk-based screening offers clearer translational benefit by efficiently guiding youth to which evidence-based cognitive or interpersonal interventions is optimal and more effective for them. The PDP study suggests that personalizing prevention by matching youths’ risk profile to a particular program modality enhances outcomes. This work also illustrates proof-of-concept that theoretically-based and empirically supported risk research can inform personalized interventions beyond a “one size fits all” approach to intervention for the average youth. This work demonstrates how linking together screening results, based on risk-based classification, with particular prevention programs can reduce depression and serves as an example for risk screening and personalizing interventions for other psychopathologies. Still, considerable future research is needed to evaluate risk-based screening and personalized interventions given broad cultural and systematic factors affecting depression risk, prevalence, and service utilization. Demonstrating that risk-based forecasting and personalization via matching apply equally across genders, races, ethnicities, and other cultural groups will be important, as will investigations across ecological systems (e.g., SES and communities).