Introduction
Children and adolescents in out-of-home care often experience cumulative stressors and adversities, including high rates of childhood maltreatment, influencing their long-term health and functioning (Esser & Schmidt, Reference Esser and Schmidt2017; Fischer et al., Reference Fischer, Dölitzsch, Schmeck, Fegert and Schmid2016; Garcia et al., Reference Garcia, Gupta, Greeson, Thompson and DeNard2017; Schmid et al., Reference Schmid, Goldbeck, Nuetzel and Fegert2008). Considering the cumulation of adversities, it is not surprising that higher levels of mental health problems and psychopathology are more common among young people placed out-of-home compared to those in the general population; a pattern often persisting into young adulthood (Beaudry et al., Reference Beaudry, Yu, Langstrom and Fazel2021; Bronsard et al., Reference Bronsard, Alessandrini, Fond, Loundou, Auquier, Tordjman and Boyer2016; Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022). Both, childhood maltreatment and mental health problems can lead to the indication of a placement in the child welfare or juvenile justice systems or might co-occur with them (Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021). Still, few longitudinal studies have investigated the course of psychopathology in such at-risk groups (Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022) and none, to our knowledge, used a longitudinal person-oriented modeling approach. Although previous research has shown that psychopathology in out-of-home placed individuals often persist into young adulthood (Seker, Boonmann, d’Huart, et al., Reference Seker, Boonmann, d’Huart, Bürgin, Schmeck, Jenkel, Steppan, Grob, Forsman, Fegert and Schmid2022; Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022), research investigating shifts in patterns of psychopathological symptoms and factors influencing symptom persistency is needed. Therefore, the aim of the current study is to describe the course of mental health problems among adolescents in youth residential care into adulthood using latent transition analyses.
Childhood maltreatment (e.g., interpersonal trauma, abuse, and neglect) is unfortunately a too common experience children are exposed to, often resulting in long lasting consequences, in particular a heightened risk for psychopathology (Bellis et al., Reference Bellis, Hughes, Ford, Ramos Rodriguez, Sethi and Passmore2019; Copeland et al., Reference Copeland, Shanahan, Hinesley, Chan, Aberg, Fairbank, van den Oord and Costello2018; Hughes et al., Reference Hughes, Bellis, Hardcastle, Sethi, Butchart, Mikton, Jones and Dunne2017; McCrory et al., Reference McCrory, Gerin and Viding2017; McLaughlin et al., Reference McLaughlin, Colich, Rodman and Weissman2020; McLaughlin, Reference McLaughlin2016). Such forms of adversity have been found to be highly prevalent in samples and societies (Bürgin et al., Reference Bürgin, Boonmann, Schmid, Tripp and O’Donovan2020; Bellis et al., Reference Bellis, Hughes, Ford, Ramos Rodriguez, Sethi and Passmore2019; Copeland et al., Reference Copeland, Keeler, Angold and Costello2007; Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards, Koss and Marks1998; Green et al., Reference Green, McLaughlin, Berglund, Gruber, Sampson, Zaslavsky and Kessler2010; Kessler et al., Reference Kessler, Aguilar-Gaxiola, Alonso, Benjet, Bromet, Cardoso, Degenhardt, de Girolamo, Dinolova, Ferry, Florescu, Gureje, Haro, Huang, Karam, Kawakami, Lee, Lepine, Levinson, Navarro-Mateu, Pennell, Piazza, Posada-Villa, Scott, Stein, Ten Have, Torres, Viana, Petukhova, Sampson, Zaslavsky and Koenen2017; Witt et al., Reference Witt, Brown, Plener, Brahler and Fegert2017). This is a major public health concern, as higher risks for various mental disorders, especially internalizing disorders (e.g., affective disorders, anxiety disorders, and post-traumatic stress disorder [PTSD], complex PTSD, and adjustment disorder) are more common among maltreated individuals (Green et al., Reference Green, McLaughlin, Berglund, Gruber, Sampson, Zaslavsky and Kessler2010; Heim & Binder, Reference Heim and Binder2012; Heim et al., Reference Heim, Newport, Mletzko, Miller and Nemeroff2008; Kessler et al., Reference Kessler, McLaughlin, Green, Gruber, Sampson, Zaslavsky, Aguilar-Gaxiola, Alhamzawi, Alonso, Angermeyer, Benjet, Bromet, Chatterji, de Girolamo, Demyttenaere, Fayyad, Florescu, Gal, Gureje and Williams2010; Lewis et al., Reference Lewis, Koenen, Ambler, Arseneault, Caspi, Fisher, Moffitt and Danese2021; Moffitt & Klaus-Grawe Think, Reference Moffitt and Klaus-Grawe Think2013). Maltreated individuals with internalizing disorders show an earlier age of onset of symptoms, greater symptom severity, more comorbidity across disorders, increased risk for suicide, and poorer treatment response than non-maltreated individuals (Cloitre et al., Reference Cloitre, Stolbach, Herman, Kolk, Pynoos, Wang and Petkova2009; Heim & Binder, Reference Heim and Binder2012; Heim et al., Reference Heim, Newport, Mletzko, Miller and Nemeroff2008; Teicher & Samson, Reference Teicher and Samson2013; Widom et al., Reference Widom, DuMont and Czaja2007). Recent developments in the field highlight the broad risks following childhood maltreatment for transdiagnostic psychopathology (Dalgleish et al., Reference Dalgleish, Black, Johnston and Bevan2020; McCrory et al., Reference McCrory, Gerin and Viding2017; McLaughlin et al., Reference McLaughlin, Colich, Rodman and Weissman2020). Taken together, childhood maltreatment is related to adverse functional outcomes, especially psychopathology across descriptive diagnostic categories and, thus, is considered one of the most important transdiagnostic risk factors for the emergence of psychopathology.
Out-of-home placed children and adolescents in particular report high rates of childhood maltreatment and mental health problems (Garcia et al., Reference Garcia, Gupta, Greeson, Thompson and DeNard2017; Gonzalez-Garcia et al., Reference Gonzalez-Garcia, Vassiliadis, Moreno-Manso, Alcantara, Del Valle and Bravo2023; Jaritz et al., Reference Jaritz, Wiesinger and Schmid2008; Jozefiak et al., Reference Jozefiak, Kayed, Rimehaug, Wormdal, Brubakk and Wichstrøm2016), which both have been associated with increased risks for poor functioning and lower quality of life in adolescence that might persist into adulthood (Dorsey et al., Reference Dorsey, Kerns, Trupin, Conover and Berliner2012; Fischer et al., Reference Fischer, Dölitzsch, Schmeck, Fegert and Schmid2016; Greger et al., Reference Greger, Myhre, Lydersen and Jozefiak2016; Jozefiak & Sonnichsen Kayed, Reference Jozefiak and Sonnichsen Kayed2015; Woods et al., Reference Woods, Farineau and McWey2013). Many young people in the child welfare and juvenile justice systems show complex and co-occurring patterns of adversity and maltreatment (Armour et al., Reference Armour, Elklit and Christoffersen2013; Charak et al., Reference Charak, Ford, Modrowski and Kerig2019; Havlicek, Reference Havlicek2014; Rebbe et al., Reference Rebbe, Nurius, Ahrens and Courtney2017; Yampolskaya et al., Reference Yampolskaya, Sharrock, Armstrong, Strozier and Swanke2014). Next to adversity and maltreatment, out-of-home placed young people often face various additional challenges (e.g., housing instability, homelessness, financial difficulties, or unemployment) along the transition from in-care to out-of-care, which further increase the risks of developing new or perpetuate extant mental health problems (Ahrens et al., Reference Ahrens, Garrison and Courtney2014; Courtney et al., Reference Courtney, Okpych and Park2018; Havlicek et al., Reference Havlicek, Garcia and Smith2013; Jozefiak et al., Reference Jozefiak, Kayed, Rimehaug, Wormdal, Brubakk and Wichstrøm2016; Vinnerljung & Sallnäs, Reference Vinnerljung and Sallnäs2008). On top, many care systems, due to structural difficulties resulting from their legal bases and funding structures, are perceived as inflexible and create system inherent pressure to leave the care setting at the age of 18, with often limited preparation and challenges in accessing appropriate aftercare (Atkinson & Hyde, Reference Atkinson and Hyde2019; Haggman-Laitila et al., Reference Haggman-Laitila, Salokekkila and Karki2020; Schmid, Fegert, Schmeck, et al., Reference Schmid, Fegert, Schmeck and Boonmann2022). As such it is not surprising, that previous studies showed that prevalence rates of mental health problems are still higher among adults who were previously placed in out-of-home care when compared with the general adult population (McKenna et al., Reference McKenna, Donnelly, Onyeka, O’Reilly and Maguire2021; Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022). Taken together, out-of-home placed children and adolescents are at risk for the emergence and persistence of psychopathology, especially at the critical period of transition from residential care settings to out-of-care and independent living.
Developmental psychopathology, as an interdisciplinary field of study, attempts to better understand the developmental interplay of bio-psycho-social factors through examining normal functioning and (psycho-)pathology, adaption and maladaption, continuity and persistence, or change in patterns over time (Cicchetti & Toth, Reference Cicchetti and Toth2005, Reference Cicchetti and Toth2009; Rutter & Sroufe, Reference Rutter and Sroufe2000). Person-oriented statistical modeling is increasingly used to describe the complex patterning of types of maltreatment, the co-occurrence of psychopathological symptoms, but also the shifts in symptom patterns over time (Achterhof et al., Reference Achterhof, Huntjens, Meewisse and Kiers2019; Lacey & Minnis, Reference Lacey and Minnis2020; Petersen et al., Reference Petersen, Qualter and Humphrey2019; Rivera et al., Reference Rivera, Fincham and Bray2018; Witt et al., Reference Witt, Munzer, Ganser, Fegert, Goldbeck and Plener2016). Person-oriented models are able to describe heterogeneous populations, multidimensional outcomes, non-linear pathways, and changing relations between processes (Hickendorff et al., Reference Hickendorff, Edelsbrunner, McMullen, Schneider and Trezise2018). Such approaches might help to statistically model developmental psychopathological transitions as they convey a “more person-oriented level of analysis of a differential pathways approach which is vital for achieving a primary mission of developmental psychopathology, implied in its definition as the study of the origins and course of individual patterns of behavioral maladaptation” (Cicchetti & Rogosch, Reference Cicchetti and Rogosch1996, p. 598). Taken together, combining both a developmental psychopathological perspective with recent trends to model co-occurring exposures and outcomes with person-oriented-modelling approaches over time is needed.
To date, few longitudinal studies have investigated the association of childhood maltreatment with the course of psychopathology among youths in residential care into young adulthood. Studying such an at-risk population longitudinally is inherently challenging and time consuming due to the cumulation of psychosocial burdens (Kind et al., Reference Kind, Seker, d’Huart, Bürgin, Jenkel, Boonmann, Habersaat, Urben, Fegert, Clemens, Bürgin and Schmid2023). To our knowledge, no study exists employing longitudinal person-oriented modeling among out-of-home placed individuals across the lifespan. As previous research has shown that psychopathology in such samples often persist into young adulthood (d’Huart, Steppan, et al., Reference d’Huart, Steppan, Seker, Bürgin, Boonmann, Birkhölzer, Jenkel, Fegert, Schmid and Schmeck2022; Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022), research investigating shifts in patterns of mental health problems and the factors influencing persistence and shifts of symptom patterns is needed.
In this longitudinal study, we aimed to describe the course of mental health problems among adolescents in youth residential care into young adulthood after they left care, using latent transition analysis. Furthermore, we aimed to investigate how the severity and latent profiles of self-reported childhood maltreatment are associated with the stability of psychopathology over time from a baseline assessment in adolescence to a 10-year follow-up in young adulthood.
Methods
Study design and procedures
In this study, we employed a longitudinal design. The baseline assessment was conducted between 2008 and 2011 throughout Switzerland as part of the epidemiological “Swiss study for clarification and goal-attainment in youth welfare and juvenile justice institutions” study (German: Modellversuch Abklärung und Zielerreichung in stationären Massnahmen [MAZ.]) (Schmid et al., Reference Schmid, Kölch, Fegert and Schmeck2013; Habersaat et al., Reference Habersaat, Ramain, Mantzouranis, Palix, Boonmann, Fegert, Schmeck, Perler, Schmid and Urben2018). Its follow-up study “Youth Welfare Trajectories: Learning from Experiences” (German: Jugendhilfeverläufe: Aus Erfahrung lernen [JAEL]) reassessed these individuals between 2017 and 2020 (d’Huart, Bürgin, et al., Reference d’Huart, Bürgin, Seker, Birkhölzer, Jenkel, Boonmann, Fegert, Schmid and Schmeck2022; Kind et al., Reference Kind, Seker, d’Huart, Bürgin, Jenkel, Boonmann, Habersaat, Urben, Fegert, Clemens, Bürgin and Schmid2023; Schmid, Fegert, Clemens, et al., Reference Schmid, Fegert, Clemens, Seker, d’Huart, Binder, Schröder, Friden, Boonmann, Jenkel, Schmeck and Bürgin2022; Seker, Bürgin, et al., Reference Seker, Bürgin, d’Huart, Schmid, Schmeck, Jenkel, Fegert, Steppan and Boonmann2022; Urben et al., Reference Urben, Habersaat, Palix, Fegert, Schmeck, Bürgin, Seker, Boonmann and Schmid2022).
The baseline MAZ. study aimed to assess the psychosocial health of young people placed within youth residential care in Switzerland. Computerized assessments were conducted by adolescents together with a professional caregiver within the partaking institution. These assessments included a collaborative part in which adolescents and their primary caregiver defined goals and competencies. Furthermore, adolescents and caregivers answered psychometric questionnaires on their own, caregivers were in the room when adolescents conducted the assessments, but were instructed to only interfere when adolescents had questions. The 64 included youth residential care institutions (35% of eligible institutions) were all accredited by the Swiss Federal Office of Justice and represent the various types of Swiss residential youth care institutions regarding their size, educational opportunities, intervention possibilities, and residing of children, adolescents, and young adults. These institutions include both educational institutions for children and adolescents and work-education institutions for young adults (Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021). At baseline, participants with non-sufficient language skills (of either German, French, or Italian) and / or intelligence impairments were excluded from participation in the MAZ. study (see Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021; Schmid et al., Reference Schmid, Kölch, Fegert and Schmeck2013). Within the MAZ. study, 592 young people took part in the study, representing around one third of all adolescents in these respective institutions (Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021). Representativeness of the MAZ.-study participants against not participating juveniles in the same institutions, was checked by asking institutional caregivers to report on mental health problems of adolescents not participating in the assessments (N = 46) on the Child or Young Adult Behavior Checklist (CBCL/YABCL) (Schmid et al., Reference Schmid, Kölch, Fegert and Schmeck2013). Age and gender matched comparisons showed no significant differences in the frequency of scoring in the clinical range on the internalizing-, externalizing- and total problems scales suggesting the participants to be representative of all residing in the partaking institutions (Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021). Of these 592 participants, 511 participants provided consent to be recontacted for a future follow-up. Further details on the MAZ. study design can be found in the final report and selected publications of the study team (Dölitzsch et al., Reference Dölitzsch, Fegert, Künster, Kölch, Schmeck and Schmid2014; Fischer et al., Reference Fischer, Dölitzsch, Schmeck, Fegert and Schmid2016; Jäggi et al., Reference Jäggi, Schmid, Bürgin, Saladin, Grob and Boonmann2021; Schmid et al., Reference Schmid, Kölch, Fegert and Schmeck2013).
Within the JAEL study and with an extensive recruitment effort, we were able to reassess a total of 231 care leavers within our follow-up study (for an overview see the flow chart in Supplement Figure 1). Between November 2017 and July 2020, participants were reassessed, amongst others, using web-based psychometric questionnaires, as well as clinical interviews about psychopathology, substance use problems, delinquency, quality of life, and own experiences in the youth residential care institutions (d’Huart, Bürgin, et al., Reference d’Huart, Bürgin, Seker, Birkhölzer, Jenkel, Boonmann, Fegert, Schmid and Schmeck2022; Schmid, Fegert, Clemens, et al., Reference Schmid, Fegert, Clemens, Seker, d’Huart, Binder, Schröder, Friden, Boonmann, Jenkel, Schmeck and Bürgin2022; Schmid, Fegert, Schmeck, et al., Reference Schmid, Fegert, Schmeck and Boonmann2022; Bürgin et al., Reference Bürgin, Clemens, Varghese, Eckert, Huber, Bruttin, Boonmann, Unternährer, OʼDonovan and Schmid2022). Participants’ inclusion or exclusion through the longitudinal cohort study leading to the final analytical sample is displayed in the Supplement (see Supplement Figure 1). In total 419 participants between 12 and 21 years of age filled out mental health screenings at baseline and provided consent for later follow-up; we were able to reassess 166 participants of these 419 eligible participants, thus our specific retention rate was 39.6%. Patterns of longitudinal dropouts were explored in detail, these attrition analyses revealed neither significant differences between included and non-included participants at the baseline assessments regarding sociodemographic characteristics and psychosocial burdens (Supplement Table 1), nor regarding their criminal records over time (up until Dec. 2017; Supplement Table 2). Thus, our final sample for the present study comprised a total of 166 adolescents in youth residential care (106 men, 63.9%; 60 women, 36.1%). At baseline, participants were on average 16.1 years old (SD = 1.9, Range [12–21 years]); at follow-up, they were on average 26.4 years old (SD = 2.3, Range [21–33 years]). Participants at baseline were placed into the youth residential care institution due to either civil law (N = 82; 49.7%) or criminal law jurisdiction (N = 43; 26.1%), or due to various other reasons (N = 40; 24.2%). A mean duration of 10.2 years (SD = 1, Range [7.9–12.5 years]) lay between baseline and follow-up assessments.
Ethical approval
Written informed consent for participation in the study was obtained from children and adolescents and their legal representatives at baseline (MAZ. study) and from young adults at follow-up (JAEL study). We only recontacted participants who provided written consent at baseline to be recontacted in case of a follow-up study. The MAZ. study was reviewed and approved by the Ethics Committees for Research on Human in the cantons (states) of Basel and Vaud (Switzerland), and by the Institutional Review Board at Ulm University (Germany). The responsible ethics committee (Ethics Commission Northwestern Switzerland) approved the JAEL study (Ref.-No.: 2017-00718).
Measurement
Mental health problems were assessed using the “Massachusetts Youth Screening Instrument – second version” (MAYSI-2) at both baseline and follow-up. The MAYSI-2 is a screening questionnaire for specific clinical symptoms to be identified early and be given special attention during a juvenile justice placement (Grisso & Barnum, Reference Grisso and Barnum2000). The questionnaire is used nationwide in many US states to guide decisions for treatment needs and was initially developed for juvenile justice and forensic settings (Grisso et al., Reference Grisso, Fusco, Paiva-Salisbury, Perrauot, Williams and Barnum2012), but is also used in the broader child welfare field and participants over the age of 18 years (Dölitzsch et al., Reference Dölitzsch, Leenarts, Schmeck, Fegert, Grisso and Schmid2017; Reilly et al., Reference Reilly, Johnson and Ferguson2019). Previous research on the psychometric properties of the MAYSI-2 has shown its reliability and validity to be equally well supported in participants over the age of 18 (Colins et al., Reference Colins, Grisso, Vahl, Guy, Mulder, Hornby, Pronk, Markus, Doreleijers and Vermeiren2015). The self-reported MAYSI-2 was previously shown to match clinical diagnoses using data from baseline assessments of the overall study (Leenarts et al., Reference Leenarts, Dolitzsch, Schmeck, Fegert, Grisso and Schmid2016). The MAYSI-2 consists of 52 items that assess symptoms during the last month and includes seven scales: Alcohol/Drug Use (ADU, eight items), Angry/Irritable (AI, nine items), Depressed-Anxious (DA, nine items), Somatic Complaints (SC, six items), Suicidal Ideation (SI, five items), Thought Disturbance (TS, five items), and Traumatic Experiences (TE, five items). All questions are answered on a binary scale (no/yes). The responses to the items yield scores on the seven scales listed above. For each scale, a scale-specific cutoff exists to categorize individuals into three groups: “clinically unsuspicious”, “caution range” or “warning range”. We did not analyze the TS scale as it is only available in boys/men, nor did we use the TE scale as items differ between genders (Leenarts et al., Reference Leenarts, Dolitzsch, Schmeck, Fegert, Grisso and Schmid2016). The cutoff for the caution range indicating problematic behaviors was used to dichotomize participants into low versus high mental health problems on the respective scale. These five binary items indicating highly problematic behavior at baseline and follow-up were used to derive latent mental health problem classes as described below.
Childhood maltreatment was assessed retrospectively by self-report at follow-up using the “Childhood Trauma Questionnaire-Short Form” (CTQ-SF) (Bader et al., Reference Bader, Hänny, Schäfer, Neuckel and Kuhl2009; Häuser et al., Reference Häuser, Schmutzer, Brahler and Glaesmer2011). The CTQ-SF consists of 25 retrospective items assessing childhood maltreatment histories scored on a 5-point Likert scale (i.e., “never true” to “very often true”). The CTQ-SF includes the following five subscales: emotional abuse, physical abuse, sexual abuse, physical neglect, and emotional neglect. Individual items are summed to give subscale scores from 5 to 25 for all five dimensions. We summed up all five subscales to create a total score of the CTQ to assess the overall severity of maltreatment. The CTQ-SF was found to show high reliability and validity, with intraclass correlation coefficients ranging from r = 0.76 to 0.86 (Bernstein et al., Reference Bernstein, Stein, Newcomb, Walker, Pogge, Ahluvalia, Stokes, Handelsman, Medrano, Desmond and Zule2003). The CTQ and its short version CTQ-SF are internationally the most widely used questionnaires to assess childhood maltreatment. Based on the German norm data, severity classifications can be formed, commonly used for descriptive purposes (“none-minimal”, “slight-moderate”, “moderate-severe” to “severe-extreme”) (Häuser et al., Reference Häuser, Schmutzer, Brahler and Glaesmer2011; Klinitzke et al., Reference Klinitzke, Romppel, Hauser, Brahler and Glaesmer2012; Witt et al., Reference Witt, Brown, Plener, Brahler and Fegert2017). For descriptive purposes we report categorical rates of maltreatment based on scoring above the moderate cutoff per subscale. Furthermore, in our models, we used both the CTQ-SF total score as an overall index of the severity of maltreatment exposure before the age of 18, and the five distinct dimensional subscale scores of the CTQ-SF to derive latent exposure profiles.
Sociodemographic characteristics were assessed using self-developed items. Migration background was categorized dichotomously as absent or present in case the participant or one of their parents was born outside of Switzerland.
Data analyses
The analyses followed the proposed procedure for conducting a Latent Transition Analysis (LTA) by Nylund et al. (Reference Nylund, Asparouhov and Muthén2007). A multistep procedure was used. First, a Latent Class Analysis (LCA) at each time point (baseline, follow-up) was modeled. Using LCA, models with 1–5 classes were considered at baseline and follow-up separately. To avoid local maxima, we used 2,000 random sets of starting values and 200 final stage optimizations. For LCAs, we used five binary items indicating highly problematic behavior on the respective subscale. In a second step, we examined the levels of measurement invariance between classes at baseline and follow-up. Then longitudinal changes over time were analyzed using LTA.
Latent transition analysis (LTA) is a methodological tool for modeling transitions in co-occurring behavior (Collins & Lanza, Reference Collins and Lanza2009; Lanza & Bray, Reference Lanza and Bray2010). This method has been used to understand how individuals move in and out of symptom profiles, e.g., in profiles of substance abuse (e.g., Bray et al., Reference Bray, Berglund, Evans-Polce and Patrick2021). The method is discussed in detail elsewhere (Collins & Lanza, Reference Collins and Lanza2009; Lanza et al., Reference Lanza, Bray and Collins2013). LTA is a longitudinal extension of LCA/LPA where latent class membership is dynamic and individuals may transition between different time points (Bray et al., Reference Bray, Berglund, Evans-Polce and Patrick2021).
Finally, covariates and/or distal outcomes were added to determine (1) whether class membership at baseline was associated with gender, age, profiles of childhood maltreatment, and overall severity of childhood maltreatment and (2) whether these variables were associated with transitions in class membership over time. As some types of child maltreatment were present in all participants of the present study and different types of childhood maltreatment usually co-occur (Herrenkohl & Herrenkohl, Reference Herrenkohl and Herrenkohl2009; Manly et al., Reference Manly, Kim, Rogosch and Cicchetti2001), we conducted a Latent Profile Analysis (LPA) to identify distinct patterns of childhood maltreatment to more accurately describe individual experiences (Witt et al., Reference Witt, Munzer, Ganser, Fegert, Goldbeck and Plener2016), and to use them in subsequent analyses. For LPA, we used the dimensional scores on the five subscales of the CTQ. To determine the optimal model, four specific fit indices were used: Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC), sample size adjusted Bayesian Information Criteria (SSaBIC), and Entropy. The bootstrap-likelihood ratio test (BLRT), and the Vuong-Lo-Mendell-Rubin test (VLMR-LRT) were used to decide on the best fitting model (Nylund et al., Reference Nylund, Asparouhov and Muthén2007). These indicators are used to compare models with K and K + 1 latent classes. A significant result indicates that a model with K + 1 classes or profiles shows a bette fit. Parsimony and interpretability were also considered. All 166 included participants had complete data; therefore, complete case analyses were conducted assuming missingness at random (MAR). Analyses were conducted using MPlus Version 8.1 (Muthén & Muthén, Reference Muthén and Muthén2012) and SPSS, Version 21. Figures were written in R (Version 4.2.2) via RStudio (Version 2022.12.0, Boston, MA, USA). Respective p-values and confidence intervals for significance tests are indicated.
Results
Descriptive data on sociodemographic characteristics, mental health problems, and childhood maltreatment are presented in Table 1. The study is overall characterized by high rates of mental health problems and childhood maltreatment, with many participants experiencing different types of mental health problems and having been exposed to diverse forms of maltreatment (see Table 1).
Sociodemographic information, psychopathology, and childhood maltreatment
Notes. MAYSI-2: Massachusetts Youth Screening Instrument – second version; *categorical prevalence was defined using the above moderate cutoff per scale.
Latent class analyses (LCAs) of mental health problems at baseline and follow-up
First, we performed LCAs at baseline and follow-up. Model fit information and model selection criteria are displayed in Table 2. Models with 1–5 classes were considered for the LCAs. Based on these results, models with 2–4 classes should be considered for final model selection employing longitudinal LTA as fit indices are reasonably close together. In the LCAs, the AIC and aBIC were minimized for the 3-class model at follow-up, whereas the BIC was minimized for the 2-class model at baseline. However, at baseline, and follow-up AIC, BIC, and aBIC showed only little differences between 2-class, 3-class, and 4-class solution. We selected the 3-class model as optimal for interpretation and additional analysis, based on aBIC, VLMR-LRT and interpretability.
Model fit information and model selection criteria for latent class analyses (LCAs). Each LCA included five domains of mental health problems at baseline and follow-up assessment both testing two- to five-class solutions
a 0.70 is used as cutoff (Fonseca & Cardoso, Reference Fonseca and Cardoso2007), values closer to 1 indicate better fit.
Lower AIC, BIC and sample size adjusted BIC (SSaBIC) values indicate better fit. Significant VLLMR-LRT and BLRT indicate better fit of the model. *p < .05, **p < .01, ***p < .001.
The plausibility of measurement invariance across the two follow-ups was examined using a likelihood ratio test, consideration of model selection criteria, and physical inspection of the LCAs. The likelihood ratio test was significant (−2DLL = 36.77, Ddf = 15, p < .001); the AIC and aBIC favored invariance and the BIC favored variance. For interpretation measurement invariance was deemed plausible and was assumed for all models.
Latent transition analyses (LTA)
Second, we performed LTA across both measurement timepoints. Parameter estimates for the selected 3-class LTA are shown in Table 3. Class 1 (51.8% prevalence at baseline, 33.7% at follow-up) was defined to be the “multiproblem”-class, as these young people reported problems across all scales of the MAYSI-2. Class 2 (39.2% at baseline and 60.2% at follow-up) was characterized by low symptoms and was therefore defined as “low symptoms” class. Class 3 (9.0% at baseline and 6.0% at follow-up) was characterized by high externalizing symptoms and was therefore labeled as “externalizing.” Item response probabilities across the five domains of mental health problems at baseline and follow-up are displayed in Figure 1.
Latent classes of mental health problems at baseline and follow-up. Exact values of item response probabilities are reported in Table 3. Item response probabilities are displayed on the binary scale of the MAYSI-2 ranging between zero and one.
Parameter estimates for the selected 3-class model
Note. Bold numbers highlight the proportion of participants staying in the same class over time, and thus stability parameters per class.
Transition rates for movement between classes across the observational period are presented in Table 3. Transition probabilities represent transitions between classes from baseline to follow-up. The diagonals in Table 3 (section latent transition probabilities, parameters in bold) represent stability parameters (Bray et al., Reference Bray, Berglund, Evans-Polce and Patrick2021). These stability parameters indicate the proportion of participants that stay in the same class at both time points. For example, 72.3% of those in the “low symptoms” class at baseline were also in the “low symptoms” class at follow-up, whereas 18.5% of them transitioned to the “multiproblem” class, while only 9.0% transitioned to the “externalizing” class. The “multiproblem” and “externalizing” class showed more group transitions over time and thus a more dynamic pattern. Those with externalizing symptoms at baseline were more likely to transition to the “low symptoms,” but 20% transitioned to the “multiproblem” class. Individuals in the “multiproblem” class at baseline were also likely to transition to the “low symptom” class at follow-up and only few transitioned to the “externalizing” class (see Table 3: latent transition probabilities; and Figure 2).
Latent transition frequencies displayed as an alluvial plot. Frequencies and latent transition probabilities are reported in Table 3.
Latent profile analysis (LPA) of childhood maltreatment
Model fit information and model selection criteria for LPA for patterns of childhood maltreatment are shown in Table 4. Models with 1–5 classes were considered for the LPA. The results favored a 3-class solution, indicated by a significant VLMR-LRT and interpretability of the results, that was used for the following analyses. Class 1 (10.2% prevalence) was defined as “sexual abuse,” as these individuals reported high levels of sexual abuse in addition to other types of childhood maltreatment. Class 2 (73.3%) was characterized by multiple maltreatment but overall lower severity and was therefore defined as “low multiple.” Class 3 (16.5%) was characterized by high levels of multiple maltreatment (with less extent of sexual abuse as compared with Class 1) and was therefore defined as “high multiple” (see Figure 3).
Latent profiles of childhood maltreatment. Item response probabilities are displayed on the dimensional CTQ subscales ranging between a minimum severity score of 5 and a maximum of 25.
Model fit information and model selection criteria for latent profile analysis (LPA). LPA included five domains of childhood maltreatment tested for two- to five-class solutions
a 0.70 is often used as cutoff (Fonseca & Cardoso, Reference Fonseca and Cardoso2007), values closer to 1 indicate better fit.
Lower AIC, BIC, and sample size adjusted BIC (SSaBIC) values indicate better fit. Significant VLLMR-LRT and BLRT indicate better fit of the model. *p < .05, **p < .01, ***p < .001.
Association between childhood maltreatment and latent classes and transitions of mental health problem
To address whether experiences of childhood maltreatment were associated with class membership at baseline, we included covariates in these models. Parameter estimates for the baseline-category, multinomial, logistic regression predicting latent class membership are shown in Tables 5 and 6. The “low symptoms” class was the reference latent class (see Table 5), all predictors were entered into the model simultaneously. Only gender, but not maltreatment profile membership was significantly associated with the “multiproblem” class at baseline (see Table 5), with women showing higher odds of being in the “multiproblem class” than men. A higher overall maltreatment severity indexed by the CTQ-total score was significantly associated with being in the “multiproblem” class at baseline (see Table 6). Due to the small number of participants in the “externalizing” class, multinomial logistic regression did not provide interpretable information regarding maltreatment profiles. However, age was significantly associated with “externalizing” class at baseline, with older age increasing the odds of being in the “externalizing” class (see Table 5).
Multinomial logistic regression effects of predictors on latent class membership at baseline including the latent profiles of maltreatment
Note. HM = high maltreatment, SA = sexual abuse & other maltreatment, LM = low multiple.
a Reference category is the “low internalizing class.”
Multinomial logistic regression effects of predictors on latent class membership at baseline including severity of overall childhood maltreatment
Note.
a Reference category for mental health problems classes is the “low internalizing” class.
To further investigate the associations of childhood maltreatment and gender with transitions between classes over time, we examined odds ratios for “movers” between- versus “stayers” within-classes. These analyses were only conducted for all classes in which more than n = 10 participants transitioned to another class, as predicting transition probabilities is only reasonable to perform when these occur. Due to the small number of participants in the externalizing class, this class was omitted from these analyses. Generally, the results showed that the “low symptom” class was the most stable and transitions from “multiproblem” and “externalizing” to “low symptoms” were the most likely classes (see Table 3). The results showed that a lower CTQ-total score and a higher age were associated with a transition from “multiproblem” to the “low symptoms” class (see Tables 7 and 8). Both CTQ-total score and maltreatment profiles were not significantly related to transitions between “low symptom” stayers versus movers to “externalizing” and “multiproblem” (see Supplement Tables 3.1 & 3.2).
Binary logistic regression effects of predictors on transitions between “Multiproblem” stayers versus movers to “low symptom” including the latent profiles of maltreatment
Note. HM = high maltreatment, SA = sexual abuse & other maltreatment, LM = low multiple.
Binary logistic regression effects of predictors on transitions between “Multiproblem” stayers versus movers to “low symptom” including severity of overall childhood maltreatment
Discussion
To the best of our knowledge, this is the first study to employ latent transition analysis to model shifts in mental health problem patterns over time in out-of-home placed adolescents. In this 10-year follow-up study of youth in residential care, we found three latent classes of psychopathology over time: a “multiproblem” class (51.8% at baseline, 33.7% at follow-up), a “low symptom” class (39.2% at baseline and 60.2% at follow-up), and an “externalizing” class (9% at baseline and 6% at follow-up). Overall, we found high stability for the “low symptom” class (72.3% of the “low symptom” class), with only 18.5% of participants transitioning from the “low symptom” class to the “multiproblem” class over the course of 10 years. Results for the “multiproblem” and “externalizing” class showed less stable and more dynamic patterns of symptom shifts. Individuals in the “multiproblem” class were likely to transition to the “low symptom” class (50.0% of the “multiproblem” class), showing less-complex patterns of psychopathology at follow-up, whereas only few transitioned to the “externalizing” class (2.3% of the “multiproblem” class). Individuals with externalizing symptoms at baseline were also more likely to transition to the “low symptom” class (66.7% of the “externalizing” class) and fewer transitioned to the “multiproblem” class (20.0% of the “externalizing” class). Taken together, our study shows transitions from more complex, co-occurring, and externalizing symptom patterns towards less complex and less severe mental health problems from adolescence into young adulthood. Additionally, cumulative childhood maltreatment among individuals in residential out-of-home placement was associated with persistence of more complex mental health problem patterns.
Previous research using person-oriented modeling in particular LTA to model shifts in classes of psychopathology over time is sparse and rather heterogenous regarding the investigated populations (mostly community- or population-based studies), time intervals between measurement points, the numbers of follow-ups, and their measure of psychopathology (Blok et al., Reference Blok, de Mol, van der Ende, Hillegers, Althoff, Shaw and White2022; Göbel & Cohrdes, Reference Göbel and Cohrdes2021; Healy et al., Reference Healy, Brannigan, Dooley, Staines, Keeley, Whelan, Clarke, Zammit, Kelleher and Cannon2022; Isdahl-Troye et al. Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017; Moore et al., Reference Moore, Dowdy, Nylund-Gibson and Furlong2019; Petersen et al., Reference Petersen, Humphrey and Qualter2022). None of these studies investigated out-of-home placed individuals, nor used the same well-validated measures for mental health problems as we did. Despite these heterogenous study designs of other previous studies, most of the studies (those merely focusing on mental health symptoms) found a four-class model as best fitting solution with a normative, a primarily internalizing, a primary externalizing, and a comorbid/multiproblem/co-occurring group (Blok et al., Reference Blok, de Mol, van der Ende, Hillegers, Althoff, Shaw and White2022; Healy et al., Reference Healy, Brannigan, Dooley, Staines, Keeley, Whelan, Clarke, Zammit, Kelleher and Cannon2022; Isdahl-Troye et al., Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017). Only one study found a three-class solution as best fit with a normative class, an emotional problem class, and a multiple problem class (Göbel et al., Reference Göbel, Ortelbach, Cohrdes, Baumgarten, Meyrose, Ravens-Sieberer and Scheithauer2022). Both studies investigating dual-factor mental health concepts, which explore psychopathology next to life-satisfaction or well-being simultaneously (Antaramian et al., Reference Antaramian, Scott Huebner, Hills and Valois2010; Suldo & Shaffer, Reference Suldo and Shaffer2008), found a higher number of classes (Moore et al., Reference Moore, Dowdy, Nylund-Gibson and Furlong2019; Petersen et al., Reference Petersen, Humphrey and Qualter2022). Within our sample, we did not find a clearly identifiable purely “internalizing” or “emotional problem” class, which is often found within community- and population-based samples (Blok et al., Reference Blok, de Mol, van der Ende, Hillegers, Althoff, Shaw and White2022; Healy et al., Reference Healy, Brannigan, Dooley, Staines, Keeley, Whelan, Clarke, Zammit, Kelleher and Cannon2022; Isdahl-Troye et al., Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017). Instead, we found a “low symptom” class with low levels of internalizing and externalizing symptoms, a small merely externalizing class, and a large class showing both mental health problems across the internalizing and externalizing spectra. In summary, our findings revealed that at-risk samples might present with different patterns of mental health problems compared to the general population, with a larger proportion exhibiting co-occurring internalizing and externalizing mental health problems. Replication in larger out-of-home placed samples using different measures of mental health problems and psychopathology is needed.
Focusing on the stability of symptom classes over time, findings in the literature are heterogenous, with some studies reporting high stability of latent class membership (Göbel et al., Reference Göbel, Ortelbach, Cohrdes, Baumgarten, Meyrose, Ravens-Sieberer and Scheithauer2022; Isdahl-Troye et al., Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017), but shifts being more common among those in the multiproblem or externalizing groups (Isdahl-Troye et al., Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017). While some previous studies suggest more dynamic shifts and transient phases of psychopathology along different developmental periods across childhood, these studies also highlight a subgroup of individuals with persisting complex psychopathological symptoms (Blok et al., Reference Blok, de Mol, van der Ende, Hillegers, Althoff, Shaw and White2022; Healy et al., Reference Healy, Brannigan, Dooley, Staines, Keeley, Whelan, Clarke, Zammit, Kelleher and Cannon2022). Studies investigating dual-factor mental health concepts showed more dynamic patterns with more than half of the sample shifting between class membership (Moore et al., Reference Moore, Dowdy, Nylund-Gibson and Furlong2019; Petersen et al., Reference Petersen, Humphrey and Qualter2022). In our study, we found relatively high numbers of transitions from the “multiproblem” and “externalizing” classes towards other classes similar to what was observed in previous studies (Blok et al., Reference Blok, de Mol, van der Ende, Hillegers, Althoff, Shaw and White2022; Healy et al., Reference Healy, Brannigan, Dooley, Staines, Keeley, Whelan, Clarke, Zammit, Kelleher and Cannon2022; Isdahl-Troye et al., Reference Isdahl-Troye, Villar, Dominguez-Alvarez, Romero and Deater-Deckard2022; McElroy et al., Reference McElroy, Shevlin and Murphy2017). This might be due to the overall higher proportions in the “multiproblem” class, which leads to comparably high proportions of transitions when looking at the entire sample. As such, findings from LTA studies show that mental health problems are mostly stable for all those in the less burdened classes, however, there are more transitions within burdened classes. Understanding factors contributing to transitions towards less troubled classes over time is of high interest. Our findings, in particular, underline that more longitudinal studies examining out-of-home placed samples are needed to better understand the persistence and stability of mental health problems over time and to guide intervention and social policy efforts.
Next to other community- and population-based studies employing LTA, meta-analyses of young adults with a residential care history have shown that the prevalence of mental disorders in young adults is still higher compared to the general population (Seker, Boonmann, Gerger, et al., Reference Seker, Boonmann, Gerger, Jaggi, d’Huart, Schmeck and Schmid2022), but lower compared to estimates from meta-analyses within out-of-home care (Beaudry et al., Reference Beaudry, Yu, Langstrom and Fazel2021; Bronsard et al., Reference Bronsard, Alessandrini, Fond, Loundou, Auquier, Tordjman and Boyer2016). Within our previous work in our cohort study, we have found that general, internalizing, and externalizing mental health problems (assessed dimensionally with the Achenbach Scales) decreased from adolescence to adulthood, with larger decreases in externalizing compared to internalizing symptoms (Seker, Bürgin, et al., Reference Seker, Bürgin, d’Huart, Schmid, Schmeck, Jenkel, Fegert, Steppan and Boonmann2022), whereas latent general psychopathology resulting from hierarchal-clustered psychologist-rated mental disorders seemed to be stable (Seker, Boonmann, d’Huart, et al., Reference Seker, Boonmann, d’Huart, Bürgin, Schmeck, Jenkel, Steppan, Grob, Forsman, Fegert and Schmid2022). Together, our findings show that self-reported mental health overall improved in our cohort of adolescents placed in youth residential care into young adulthood, which is in line with the meta-analytic evidence within the field. Nevertheless, a subgroup of participants of our study still shows persistently high levels of mental health problems.
Our study further indicated that cumulative childhood maltreatment is associated with mental health problems in adolescence and is associated with persisting complex symptom patterns from adolescence to young adulthood in young people placed out-of-home. A large body of previous research has well established that childhood maltreatment and other forms of early adversity are associated with higher levels of psychopathology and mental disorders (Green et al., Reference Green, McLaughlin, Berglund, Gruber, Sampson, Zaslavsky and Kessler2010; Hughes et al., Reference Hughes, Bellis, Hardcastle, Sethi, Butchart, Mikton, Jones and Dunne2017; Kessler et al., Reference Kessler, McLaughlin, Green, Gruber, Sampson, Zaslavsky, Aguilar-Gaxiola, Alhamzawi, Alonso, Angermeyer, Benjet, Bromet, Chatterji, de Girolamo, Demyttenaere, Fayyad, Florescu, Gal, Gureje and Williams2010; Lewis et al., Reference Lewis, Koenen, Ambler, Arseneault, Caspi, Fisher, Moffitt and Danese2021; McCrory et al., Reference McCrory, Gerin and Viding2017; McLaughlin et al., Reference McLaughlin, Colich, Rodman and Weissman2020; McLaughlin, Reference McLaughlin2016; Moffitt & Klaus-Grawe Think, Reference Moffitt and Klaus-Grawe Think2013). Next to these associations, childhood maltreatment exposure has repeatedly been shown to be associated to different characteristics of specific mental disorders, for example an earlier age of onset, higher severity of symptoms, comorbidity, increased risk for suicide, and poorer treatment response (Cloitre et al., Reference Cloitre, Stolbach, Herman, Kolk, Pynoos, Wang and Petkova2009; Heim & Binder, Reference Heim and Binder2012; Teicher & Samson, Reference Teicher and Samson2013; Widom et al., Reference Widom, DuMont and Czaja2007). Interestingly, we did not find a significant association between latent maltreatment profiles on baseline mental health classes, nor on transitions between classes over time. This might be explained by the overall high burden of maltreatment in the studied sample and the small sample size, as such the reference group (“low multiple”) also displays a considerable amount of neglect and comprised over 70% of the total sample, with the two other profiles being rather small and thus corresponding errors of estimates being rather large. Moreover, using a broader assessment of maltreatment with newer and more comprehensive measures might have resulted in different and more specific maltreatment patterns (Jarczok et al., Reference Jarczok, Lange, Meinck, Witt, Clemens, Fegert and Jud2023; Teicher & Parigger, Reference Teicher and Parigger2015), which is important to consider in future research having large sample sizes. In sum, by using a person-oriented longitudinal modeling approach, we were able to show that maltreatment severity, but not latent profiles, is associated to more complex mental health problems in adolescence and to more perpetuating multiproblem psychopathology in a 10-year follow-up across the transition from within residential care to after residential care.
Our findings, along previous research, inform current debates on transdiagnostic approaches to conceptualize mental disorders and underscore the need to investigate childhood maltreatment as a transdiagnostic risk factor. The descriptive classification of mental disorders has been challenged since its introduction by the question of how to best conceptualize the impact of childhood maltreatment on mental disorders (Schmid et al., Reference Schmid, Petermann and Fegert2013; Teicher & Samson, Reference Teicher and Samson2013; van der Kolk et al., Reference van der Kolk, Pynoos, Cicchetti, Cloitre, D’Andrea, Ford, Lieberman, Putnam, Saxe and Spinazzola2009). Continuing debate revolves around how the latent vulnerability due to maltreatment can be related to distinct subtypes of specific mental disorders, debate also continues on how such complex psychopathology is developmentally appropriately assessed within the disorder category of PTSD and the newly introduced complex PTSD diagnoses (Ford et al., Reference Ford, Charak, Karatzias, Shevlin and Spinazzola2022; Maercker et al., Reference Maercker, Cloitre, Bachem, Schlumpf, Khoury, Hitchcock and Bohus2022; Spinazzola et al., Reference Spinazzola, van der Kolk and Ford2021; Teicher et al., Reference Teicher, Gordon and Nemeroff2022). More recently transdiagnostic approaches to mental health, as the general p-factor theories, the hierarchical taxonomy of psychopathology (HiToP), or the Research Domain Criteria (RDoC), have gained prominence mainly in research (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel, Meier, Ramrakha, Shalev, Poulton and Moffitt2014; Cuthbert & Insel, Reference Cuthbert and Insel2013; Dalgleish et al., Reference Dalgleish, Black, Johnston and Bevan2020; Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon and Zimmerman2017). Such approaches highlight important core dimensions of functioning and conceptualize mental disorder under higher-order dimensions or along core domains of functioning (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel, Meier, Ramrakha, Shalev, Poulton and Moffitt2014; Cuthbert & Insel, Reference Cuthbert and Insel2013; Dalgleish et al., Reference Dalgleish, Black, Johnston and Bevan2020; Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon and Zimmerman2017). Following such an approach, childhood maltreatment (or adversity and trauma) can and should be conceptualized as transdiagnostic risk factors for the emergence of general psychopathology (McLaughlin et al., Reference McLaughlin, Colich, Rodman and Weissman2020; Nolen-Hoeksema & Watkins, Reference Nolen-Hoeksema and Watkins2011). Recent research also is increasingly invested in filling the gap between childhood risks and (psycho-)pathology by investigating mediators and mechanistic pathways transferring these risks (McLaughlin, Reference McLaughlin2016; Moffitt et al., Reference Moffitt, Belsky, Danese, Poulton and Caspi2017). Important mechanisms include emotion regulation, social information processing, and accelerated aging processes, which shape latent liabilities and vulnerability (Keyes et al., Reference Keyes, Eaton, Krueger, McLaughlin, Wall, Grant and Hasin2012; McCrory et al., Reference McCrory, Gerin and Viding2017; McLaughlin et al., Reference McLaughlin, Colich, Rodman and Weissman2020). Next to transdiagnostic outcomes and mechanisms conveying risks, recent debates further foster an understanding of different core dimensions of adverse environments in childhood (Ellis & Del Giudice, Reference Ellis and Del Giudice2019; Humphreys & Zeanah, Reference Humphreys and Zeanah2015; Lange et al., Reference Lange, Goossens, Bakker, Michielse, van Winkel, Lissek, Leibold, Marcelis, Wichers, van Os, van Amelsvoort and Schruers2019; Lewis et al., Reference Lewis, Koenen, Ambler, Arseneault, Caspi, Fisher, Moffitt and Danese2021; McLaughlin, Reference McLaughlin2016). From a developmental perspective, such early deviations from “expectable” environments interfere with different important developmental tasks including the development of emotion regulation, of close attachment bonds and peer relationships, with the development of an autonomous self, and different dimensions of personality functioning (Cicchetti & Doyle, Reference Cicchetti and Doyle2016; d’Huart, Hutsebaut, et al., Reference d’Huart, Hutsebaut, Seker, Schmid, Schmeck, Bürgin and Boonmann2022; Humphreys & Zeanah, Reference Humphreys and Zeanah2015). All these intra- and interpersonal abilities are important protective factors for the emergence of psychopathology and underline core features of resilience.
Along these debates, our findings highlight that many individuals in at-risk populations – specifically young people placed out-of-home – present with comorbid mental health problems and psychopathological symptoms across the internalizing and externalizing spectra. As such, a transdiagnostic understanding of mental disorders and mental health problems is essential within out-of-home placed populations, who seldomly present with diagnostically well-fitting symptom presentations. Our findings also highlight, that almost half of participants in the multiproblem group shifted to a less troubling pattern of psychopathology over a 10-year follow-up and as such intervention by child welfare system might pay-off (at least for some). At the same time a quarter of all participants, almost half of all those in the multiproblem group at baseline show a stable pattern of co-occurring psychopathology over time. This subgroup might specifically benefit from trauma-focused interventions, including both evidence-based and trauma-focused psychotherapy, but also from concepts of trauma-informed care (Bath, Reference Bath2008; Forkey et al., Reference Forkey, Szilagyi, Kelly and Duffee2021; Hopper et al., Reference Hopper, Bassuk and Olivet2010). The implementation of trauma-informed pedagogical concepts in different youth welfare contexts hold promise as they also aim to build capacity and to foster resilience by not merely treating the individual but shaping the socio-ecological context for young people to be able to develop and thrive (Bunting et al., Reference Bunting, Montgomery, Mooney, MacDonald, Coulter, Hayes and Davidson2019; Lang et al., Reference Lang, Schröder and Schmid2021; Schmid et al., Reference Schmid, Lüdtke, Dolitzsch, Fischer, Eckert and Fegert2020; Zhang et al., Reference Zhang, Conner, Lim and Lefmann2021). These trauma-informed approaches not only account for trauma-related threat-responses of young people during daily life but are intended to enable young people from deprived environments to cope with their different developmental tasks at hand, to regulate their emotions with the help of a caregiver, to form close attachment and relationships, and to develop an autonomous self.
Strengths and limitations
Our findings within the present study need to be interpreted considering some important limitations. First, we used a retrospective self-report measure of childhood maltreatment which are debated considering certain biases (e.g., recall biases) with new studies reporting only poor overlap between maltreatment exposure assessed in prospective versus retrospective research designs (Baldwin et al., Reference Baldwin, Reuben, Newbury and Danese2019; Bürgin et al., Reference Bürgin, Boonmann, Schmid, Tripp and O’Donovan2020; Danese & Widom, Reference Danese and Widom2020; Danese, Reference Danese2020; Hardt & Rutter, Reference Hardt and Rutter2004). Still, the CTQ(-SF) is one of the most used measures of childhood maltreatment and childhood “trauma” worldwide, but also within the German-speaking spheres (Bader et al., Reference Bader, Hänny, Schäfer, Neuckel and Kuhl2009; Häuser et al., Reference Häuser, Schmutzer, Brahler and Glaesmer2011; Klinitzke et al., Reference Klinitzke, Romppel, Hauser, Brahler and Glaesmer2012; Witt et al., Reference Witt, Brown, Plener, Brahler and Fegert2017). Importantly, the CTQ does not differentiate when or in which specific context maltreatment occurred, however, around half of the items target the specific family context, whereas other items are broader-framed and might also have assessed peer- or caregiver-conducted maltreatment during the time or in between placements, within or outside of the institutions. Second, a control group of young people without residential care placements is lacking, thus our findings are not suited to address efficacy of youth residential care placement in general. Furthermore, the individual residential care trajectories and received specific pedagogical and therapeutic interventions vary greatly, and there is only sparse quantitative self-report data available about early childhood years and different forms of interventions. All of these are intertwined with specific behavioral and emotional problems, as such we did not test any placement related characteristics as predictors of mental health trajectories, nor do we believe that our study design is suited to test the efficacy or non-efficacy of out-of-home care or of specific psychotherapeutic interventions. Third, even though the analyses provided stable solutions, the LTA approach generally favors larger sample sizes (Muthén, Reference Muthén2021). Fourth, generalizability needs to be made regarding the Swiss residential care setting and the broad age range (12–21) at baseline. As outlined, our sample draws from adolescents that resided in various kinds of youth residential care institutions, both the population of study and the youth residential care background are highly heterogenous. Participating institutions, and participants in these institutions seem to be representative for the overall Swiss youth residential care population. Moreover, our attrition analyses have shown no systematic drop-out. Still, generalizability and inference should be made with caution and inference needs to consider the Swiss-specific and diverse population of study. Fifth, we used the same mental health measure at baseline and follow-up, which comes at the tradeoff of using the measure in an age range and setting for which it was not initially developed for. Despite these shortcomings this study has important strengths. We have investigated a rare and well-characterized high-risk sample of young adults with previous residential care placements over a follow-up period of 10 years, such data is rather rare and Swiss-unique. Moreover, our study incorporated psychometric assessment on child maltreatment and two measurement time-points on mental health within residential youth care and afterwards, combining such data from a rare sample with person-oriented longitudinal modeling approach is a great asset of our work.
Implications
Our findings have important implications for (child-)welfare systems and social policy. First, we need to continue to invest into the prevention of transdiagnostic risk factors for the emergence and perpetuation of psychopathology – foremost childhood maltreatment. Second, we need to promote accessible and suitable interventions and trauma-informed care concepts in different at-risk settings (Bunting et al., Reference Bunting, Montgomery, Mooney, MacDonald, Coulter, Hayes and Davidson2019; Forkey et al., Reference Forkey, Szilagyi, Kelly and Duffee2021; Lang et al., Reference Lang, Schröder and Schmid2021; Schmid et al., Reference Schmid, Lüdtke, Dolitzsch, Fischer, Eckert and Fegert2020; Zhang et al., Reference Zhang, Conner, Lim and Lefmann2021). Third, next to treating maltreatment-related psychopathology, welfare systems should foremost aim to build capacity within young people and help them to cope with their developmental tasks at hand to regulate their emotions, to form close attachments and supportive relationships, and to develop an autonomous self to be able and enabled to be well and partake within society (Bürgin et al., Reference Bürgin, Clemens, Witt, Sachser, Jud, Brähler, Strauß, Petrowski, Schmid and Fegert2023; Schmid, Fegert, Clemens, et al., Reference Schmid, Fegert, Clemens, Seker, d’Huart, Binder, Schröder, Friden, Boonmann, Jenkel, Schmeck and Bürgin2022). Last, transitioning from within residential or psychiatric care to after care and an independent life is a critical phase of vulnerability for which well-fitting and accessible help is needed as development is far more dynamic than a linear increase in age (Banaschewski et al., Reference Banaschewski, Fegert, Freyberger, Karow, Reif, Schaff, Voderholzer and Driessen2019; Danese, Reference Danese2022; Haggman-Laitila et al., Reference Haggman-Laitila, Salokekkila and Karki2020).
Future research should aim to study populations with high levels of adversities over time within developmentally informed designs implementing person-oriented modeling approaches. Such research is time and cost consuming, but longitudinal studies following such populations are needed to elucidate risk and protective factors and mechanisms transferring these effects (Kind et al., Reference Kind, Seker, d’Huart, Bürgin, Jenkel, Boonmann, Habersaat, Urben, Fegert, Clemens, Bürgin and Schmid2023; Panter-Brick & Leckman, Kind et al., Reference Panter-Brick and Leckman2013; Schmid, Fegert, Clemens, et al., Reference Schmid, Fegert, Clemens, Seker, d’Huart, Binder, Schröder, Friden, Boonmann, Jenkel, Schmeck and Bürgin2022; Schmid, Fegert, Schmeck, et al., Reference Schmid, Fegert, Schmeck and Boonmann2022). Moreover, future research should incorporate new measures of childhood maltreatment and lifetime stressors (e.g. the MACE, the ICAST-R, or the STRAIN), which are more inclusive of different forms of adversity and stressors and allow the assessment and the statistical modeling of frequency, timing, and duration of different exposures (Isele et al., Reference Isele, Teicher, Ruf-Leuschner, Elbert, Kolassa, Schury and Schauer2014; Jarczok et al., Reference Jarczok, Lange, Meinck, Witt, Clemens, Fegert and Jud2023; Schalinski et al., Reference Schalinski, Teicher, Nischk, Hinderer, Müller and Rockstroh2016; Sturmbauer et al., Reference Sturmbauer, Shields, Hetzel, Rohleder and Slavich2019; Teicher & Parigger, Reference Teicher and Parigger2015). Future research might also focus on latent profiles of adjustment regarding broader outcomes beyond merely studying mental health problems. For instance, “dual-factor mental health” perspectives enable to investigate psychopathology together with functioning, in order to better understand which young people still report a high quality of life despite presenting with psychopathology (Antaramian et al., Reference Antaramian, Scott Huebner, Hills and Valois2010; Moore et al., Reference Moore, Dowdy, Nylund-Gibson and Furlong2019; Petersen et al., Reference Petersen, Humphrey and Qualter2022). Such perspectives might help to build capacity, foster well-being and quality of life and help young people coming from deprived and maltreating backgrounds to partake in society and to be well (Bürgin et al., Reference Bürgin, Clemens, Witt, Sachser, Jud, Brähler, Strauß, Petrowski, Schmid and Fegert2023; Schmid, Fegert, Clemens, et al., Reference Schmid, Fegert, Clemens, Seker, d’Huart, Binder, Schröder, Friden, Boonmann, Jenkel, Schmeck and Bürgin2022).
Conclusion
The findings of this study shed light on the stability and shifts in classes of mental health problems over the course of 10 years from adolescence into young adulthood in adolescents placed within youth residential care in Switzerland. Our findings underscore the importance to understand childhood maltreatment as transdiagnostic risk factor for the emergence, but also for the perpetuation of psychopathology across the internalizing and externalizing spectra. Findings also show that half of young adults within the multiproblem class shifted to less troubling patterns of psychopathology, while the other half showed persisting internalizing and externalizing symptoms, highlighting the need for trauma-informed and -focused interventions within youth residential care, as well as interdisciplinary collaboration among residential care institutions, child welfare, youth mental health services, and child- and adolescent psychiatric care centers. More longitudinal research in at-risk populations is needed to better understand trajectories of mental health problems and risk and protective factors for adjustment over time.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0954579423001426
Acknowledgements
This work was supported by the Swiss Federal Office of Justice with two grants for the MAZ. and JAEL study awarded to MS, JMF, & KS. DB was supported by the Research Fund for Excellent Junior Researcher of the University of Basel for the study “CT-Quest.”
Competing interests
None.
Open access
