The Interplay of Youth and Care Characteristics with a Positive Social Climate in Therapeutic Residential Youth Care

Background: Limited research exists on how therapeutic residential youth care (TRC) achieves treatment outcomes. More specifically, little is known about the association between contextual factors such as treatment organization, youth characteristics, and experienced social climate in TRC. Therefore, this study aims to investigate differences between latent classes of TRC and youth characteristics and their association with a positive perceived social TRC climate. Method: We applied a person-centered approach in a cross-sectional design with a sample of 400 adolescents and 142 staff leaders. We analyzed youth and TRC characteristics in a latent class analysis and established associations with social climate for these two groupings. Results: The two types of TRC settings we found, i.e., larger TRC settings and family-style TRC settings, show small differences in social climate. These settings only differed on youth activities and staff shifts type (more cohabitation and unorganized activities outside TRC in family-style TRC). We identified four adolescent classes: A severe problems group, youth with incidental problems, family problems, and a migrant background group. The migrant background group showed the most positive perceptions of social climate, followed by youth with incidental problems, family problems, and severe problems. Conclusions: TRC staff should acknowledge how perceived social climate is connected to TRC characteristics and the heterogeneity of adolescents in care. As social climate is subjective and dynamic, a continuous dialogue about TRC social climate between staff and youth is recommended. Future research should investigate how these aspects are associated with treatment outcomes to increase our understanding of achieving positive outcomes in TRC.


Introduction
Norwegian Therapeutic Residential Youth Care (TRC) offers treatment and care in an open setting for young people (temporarily) living outside their family environment.TRC "… involves the planful use of a purposefully constructed, multidimensional living environment designed to enhance or provide treatment, education, socialization, support, and protection to children and youth with identified mental health or behavioral needs in partnership with their families and in collaboration with a full spectrum of community based formal and informal helping resources" (Whittaker et al., 2015, p. 24).In most countries, including Norway, TRC is seen as a last resort and considered to be the treatment of choice mainly if outpatient treatment and foster care placement have been unsuccessful (Backe-Hansen et al., 2011;Frensch & Cameron, 2002;Knorth et al., 2008).
Meta-analyses on TRC outcomes (e.g.De Swart et al., 2012;Grietens, 2002;Knorth et al., 2008;Scherrer, 1994) show small to moderate effects in terms of, for example, decreasing externalizing behavior problems of adolescents.There is even evidence that non-residential youth care may be more effective in reducing mental health problems than residential youth care (Gutterswijk et al., 2020;Strijbosch et al., 2015).Furthermore, the longer the follow-up period, the less convincing the effect of the intervention is (Frensch & Cameron, 2002;Harder & Knorth, 2015;Knekt et al., 2016;Scherrer, 1994).
Moreover, there is limited evidence for how TRC actually achieves treatment goals: TRC remains too much of a "black box" (Harder & Knorth, 2015;Knorth, 2003;Libby et al., 2005).In order to gain more durable positive treatment results we need to know more about how results are achieved, rather than investigating the achieved results (Harder & Knorth, 2015).One of the factors associated with this process of change is how the living environment within TRC settings is experienced by young people (e.g., Lanctôt et al., 2016;Leipoldt et al., 2019;Mathys, 2017), hereafter denoted as social climate or perceived social climate.
Social climate can be defined as the discrete, and consistent continuity of events containing collective elements in environmental demands (Murray, 1938;Stern, 1970;Theunissen, 1986;Van der Ploeg, 1982).These 'environmental demands' are shared among individuals in the same environment (Moos, 2003).For example, when staff members always make sure they leave room for young peoples' initiatives including that everyone gets a chance to set-up initiatives, the social environment may be perceived as stimulating the autonomy of young people.This definition describes social climate on the micro level and a more recent definition defines social climate as "… the quality of the social and physical environment in terms of the provision of sufficient and necessary conditions for the physical and mental health, well-being, and personal growth of the residents, with respect for their human dignity and human rights as well as (if not restricted by judicial measures) their personal autonomy, aimed at participation in society" (Stams & Van der Helm, 2017, p. 4).
The definitions clearly illustrate a link to the self-determination theory (Van der Helm et al., 2018), where relatedness, autonomy, and competence are necessary for motivation and personal growth of youths (Ryan & Deci, 2000).Furthermore, social climate is considered a factor that is unrelated to treatment content but which can set the conditions for achievement of successful treatment outcomes (Andrews, 2011;Cantora et al., 2014;Lanctôt et al., 2016).Social climate can also be described as a contextual correlate of TRC.Studies investigating these contextual factors, such as setting characteristics (including type, structure, size, setting, staff, and daily routines) are scarce in contemporary TRC studies (e.g., Attar-Schwartz, 2017;Leipoldt et al., 2019).Previous studies have shown that these contextual factors can explain some 30 up to 50 percent of the variance in treatment outcomes (Attar-Schwartz, 2017;Moos & Lemke, 1996).Therefore, contextual factors, such as social climate, are important and should be considered when evaluating TRC results.
In a recently conducted systematic review of research over the past 28 years (Leipoldt et al., 2019), results indicated that TRCs focusing on strengths of young people, using a small group size, applying evidencebased interventions, having open (vs.closed) units, having staff with more work experience, and providing opportunities for training, and good clinical leadership are associated with a positive social climate.Furthermore, structured routines with varied daily activities are considered prerequisites for a positive social climate in TRC.Youth with a longer care history, more psychological and psychiatric (both internalizing and externalizing) problems, and living in secure units, more often perceive a negative social climate than youth without these characteristics (for a more detailed summary of the available evidence, see Leipoldt et al., 2019).
Although the findings of this review provide insight into which factors underlie a positive perception of social climate in TRC, they tell us little about how and why these factors interplay.For example, small groups are associated with a more positive social climate, but it is unclear whether larger organizations that have multiple small units can also perform as well as small centers.In addition, it is unknown how characteristics of adolescents interact with characteristics of TRC settings in relation to perceived social climate.The present study will address these issues in a cross-sectional perspective and tries to identify associations between youth, staff, and organizational characteristics in relation to perceived social climate in TRCs.

The present study
Given the limited empirical knowledge about determinants of social climate in TRCs and the importance of social climate for positive TRC outcomes, this study aims to explore how TRC and adolescent characteristics interplay in predicting perceived social climate in TRCs as intermediate outcomes.Previous research has mainly assessed associations between TRC and adolescent characteristics on social climate separately (see Leipoldt et al., 2019).To the best of our knowledge, no studies have focused on combining TRC characteristics and young peoples' characteristics in predicting experienced social climate in residential youth care centers.Insight into how organizational and adolescent factors interplay in predicting social climate can help TRCs to choose or tailor characteristics and strategies to their residents and to use this knowledge to guide how the TRCs can be improved to create more positive living environments.In the present study we will address the following research questions: (1) How are TRC characteristics associated with experienced social climate in TRC? (2) How are adolescent characteristics associated with experienced social climate in TRC? (3) How do adolescent characteristics and TRC characteristics interact in their association to experienced social climate in TRC?For the first question, based on previous research (Leipoldt et al., 2019), we expect that TRCs that are small and publicly owned, that have daily routines, staff training routines, staff with relevant education and that are located in more populated areas will show associations with a positive social climate.In contrast, we expect that TRCs that do not have these above-mentioned aspects will show associations with a negative social climate.Regarding the second question, we expect that adolescents with an internalizing psychiatric diagnosis will perceive the social climate as more supportive, compared to adolescents with an externalizing psychiatric diagnosis.In addition, and more explorative, we expect that adolescents with a shorter treatment history, Caucasian background, less school problems, and lower emotional regulation problems will perceive the social climate more positively.Finally, since previous studies have not investigated the interplay (interaction) between TRC and adolescent characteristics on social climate, we will perform an exploratory analysis without formulating a priori hypotheses.

Treatment setting
TRC settings in Norway, approved by the Directorate of Children and Family Affairs, can be commercially, non-commercially, or publicly owned.Each TRC is organized as a group of several small units with on average three-to-five residents and staff presence ratios of 1:2 in an environment similar to a group home (Backe-Hansen et al., 2011;Jozefiak & Sønnichsen Kayed, 2015).Youths aged 10-23 years old can be placed due to abuse, neglect, or behavioral problems.TRC is chosen as first placement for a minority, whereas most of those entering TRC do so after unsuccessful foster care.The primary goal is to provide care and parenting substitution to adolescents, to support them in attending school or work, and to let them participate in leisure activities both inside and outside the center.The broader goal of TRC is reducing social and psychological problems of the residents, while helping to socialize them through relationships with staff members and the resident group.TRC tends to be temporary, because (family) foster care is the preferred long-term placement (Backe-Hansen et al., 2011).
In our sample, placements are based on solely abuse or neglect for 76% of all cases, and for the others (24%) on combinations of adolescents' behavioral problems including criminal behavior, substance use, or conduct disorders.More than two-thirds of staff working with young people within TRC settings (70.87%,SD = 22.47) has bachelor's degree or other degrees relevant for social work.The average staff-patient ratio is 1: 2.49 (SD = 1.53).Staff leaders have an average working experience in TRC of 12.75 years (SD = 8.10) and an average of 6.61 years (SD = 6.43) work experience in other relevant social work.

Participants
We obtained data for the present study from a large-scale Norwegian project on mental health in adolescents living in TRCs (Jozefiak et al., 2016).All TRC settings for adolescents aged 12-23 years were invited to participate in the study (see Fig. 1), however none of the participants were older than 20 years of age.
Unaccompanied minors without asylum in Norway and adolescents living in emergency care units are excluded from participation.They are considered to be in such a high state of crisis that data collection should not be a priority (Jozefiak & Sønnichsen Kayed, 2015).In addition, some centers specialized in conduct problems are excluded because of their high level of research activity (Andreassen, 2015).Furthermore, adolescents with insufficient proficiency in Norwegian language are excluded.Of the 98 eligible centers and 601 eligible adolescents, 86 centers and 400 adolescents consented to participate in the study, resulting in a 67% response rate (Jozefiak et al., 2016).
More than half of the 400 participating adolescents in the present study (57.5%) is female.The average age is 16.7 years (SD = 1.37), ranging from 12 to 20 years old.Adolescents have on average experienced 3.34 former placements (SD = 2.43) and have been placed for the first time in care when they were on average 12.5 years old (SD = 3.88).Most of the adolescents attend school (69.2%) or have jobs (11.3%).A majority (76%) meets criteria for at least one psychiatric diagnosis.Of the 86 participating TRC units (with a sample of 142 staff leaders), 18% have routines for visits by health-care workers.For more details about the sample characteristics, see Jozefiak et al. (2016).

TRC characteristics
TRC characteristics are measured with a questionnaire for TRC leaders that has been specifically designed for the present study.The questionnaire includes work experience of TRC leaders, work position of staff members and routines, division of free time for adolescents, school organization during care, and routines for daily activities.The items contain multiple answering formats including open-ended items, yes/no items, and Likert-scale items.An example item from the center's leader part is: "How many years have you been a leader in this institution?"A sample item from the routines part is: "Are there regular household meetings for the youngsters and staff members?".Information from the above-mentioned sections is used in the latent class analysis (see below) of TRC characteristics.

Semi-structured interview
The aim of the semi-structured interview is to gather information regarding the youths' life before and during stay in TRC.The interview consists of questions around themes relating to care history and number of placement (10 questions, e.g."How old were you at the first placement?");organized family care and information about the family ( 16questions, e.g."Does your father have a chronic disease?"); own and parental substance problems (nine questions, e.g."Have you used drugs in this institution?";eight questions, e.g."Does your mother often use much medicine?");school history (10 questions, e.g."Do you go to school every day?"); and physical measurements (seven questions, e.g."What is your length in cm.?").Some questions are open-ended, others use yes/no or Likert scale response alternatives.Information from the above-mentioned themes is used in the latent classes of adolescent characteristics.

Psychiatric interview
We used the Child and Adolescent Psychiatric Assessment (CAPA; Angold & Costello, 2000) to gather information on psychiatric problems reported by the youths.The CAPA uses a computer-based scoring algorithm for diagnostic evaluation, which results in DSM-IV-TR diagnoses, except for autism spectrum disorder; reactive attachment disorder, and attention deficit hyperactivity disorder, which are based on information from TRC staff, also evaluated with DSM-IV algorithms.Inter-rater reliabilities between interviewer/rater pairs as estimated by Gwet's AC 1 .Agreement rates are: Conduct disorder (0.78, 86%), reactive attachment disorder (0.82, 88%), major depressive disorder (0.89, 93%), and generalized anxiety disorder (0.93, 95%) (Jozefiak et al., 2016).Seven diagnostic group categories (any anxiety disorder, any depressive disorder, any substance abuse disorder, ADHD diagnosis, reactive attachment disorder, autism spectrum disorder, and 'any DSM diagnosis') are used in the latent classes of adolescent characteristics.

Emotion regulation
To measure emotion regulation, we used the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004).The 36 items were translated into Norwegian for the present study.An example item is: "I pay attention to how I feel".The DERS consists of six subscales and a combined total scale score.Participants answered on a five-point Likert Scale ranging from 1 = almost never, to 5 = almost always.Higher scores indicate more difficulties in emotion regulation.To compensate for missing values, we calculated an average mean score and multiplied this by 36 to obtain a total score.Internal consistency for the total score was high (α = 0.93).The total score for the DERS is included in the latent classes of adolescent characteristics.

Social climate
We measured the young person's perception of social climate with the Norwegian version of the Community Oriented Programs Environment Scale (CPES; Moos, 2009).We used a short, refined version (Leipoldt et al., 2018) consisting of ten subscales divided over three dimensions (see Table 1).The scale consists of 40 true/false statements and higher scores are indicative for a higher endorsement of a social climate aspect.Scores range between 0 (low endorsement) and 4 (high endorsement).Since all subscales consist of four dichotomous items, the scale scores are directly comparable between subscales.An example The extent to which staff use measures to keep members under necessary controls item of the involvement scale is: "The members are proud of this program."The reliability and construct validity of the original CPES have been well documented (Moos, 2009).In the present sample, the refined version shows good internal consistency values (α = 0.81).In the current study, internal consistency values for the subscales varied between α = 0.50 and 0.66, which is, especially for short scales, acceptable for group comparisons.In addition, composite reliability (Raykov, 1997) scores indicated acceptable to good reliability for the short version of the CPES in the refinement study of the instrument.The psychometric quality is further strengthened with a good construct validity where the short scales showed a good representation of the social climate construct (for more information, see Leipoldt et al., 2018).

Procedure
All eligible Norwegian TRC centers listed in the 2010 national database of the Directorate for Child and Family Affairs were contacted in random order.Centers were provided with information about the research project and its goals.Data collection was carried out between 2011 and 2014 by four research assistants, all of whom holding master's degrees in psychology or social work, and had extensive experience working with adolescents and families.They administered multiple questionnaires and interviews to adolescents, teachers, primary contacts (staff members), and TRC leaders.
The CPES was administered as the second instrument in the datacollection process, immediately after the CAPA interview, given its importance to the main study.Thereafter, questionnaires concerning mental health, emotional regulation, social support, and quality of life were completed.The CAPA interview took on average 2.5 h and different questionnaires took approximately half-an-hour to fill out.Youths completed the questionnaires in the presence of the research assistant, and they could ask clarification questions regarding the content of the items.If the participants had difficulty reading the questions, the research assistant would read it aloud for them.During the entire data-collection period, a team of child and adolescent psychiatrists and psychologists was available in case of emergencies.Each adolescent received a gift certificate of 500 NOK (approximately USD 90) for participation.

Ethics
Participants were recruited in accordance with procedures approved by the Norwegian Regional Committee for Medical and Health Research Ethics (reference number 2016/1169/REC Central).Consent was given by the adolescent, and for those younger than 16 years of age, informed consent was also obtained from a significant caregiver.At the start of the data-collection process, the details of the research project were discussed with the adolescent once more to ensure informed consent.In addition, informed consent was also provided by staff leaders and main contact persons of the adolescents.All participants could withdraw from the study at any time.

Data-analysis
This study used latent class analysis (LCA) as an exploratory strategy to obtain different classes of TRC and adolescent characteristics (Von Eye & Bergman, 2003).First, we calculated descriptive statistics for the relevant variables that were to be included in the LCA.Second, we performed two separate LCA's to obtain different classes for TRC and adolescent characteristics.We did not a priori specify the number of classes that would emerge, but rather examined fit indices, class percentages, and interpretability of the data to determine the most appropriate number of classes.To help determine the optimal number of classes, we used the − 2 Log Likelihood Statistic, (Sample size adjusted) Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), Entropy Values, Vuong-Lo-Mendell-Rubin test (VLMR; Lo, 2001), average posterior class probabilities (AvePP), and bootstrapped likelihood ratio tests.The first three criteria should be as small as possible, bootstrapped likelihood ratio tests should be statistically significant (Nylund et al., 2007), and entropy values should be >0.80 (Clark & Muthén, 2009).To ensure sufficient statistical power, each class also had to consist of at least 30 participants (Lanctôt et al., 2016).For the classes, item-probability scores of <0.30 and >0.70 indicate high class homogeneity membership, and AvePP values of >0.70 indicate wellseparated classes (Nylund-Gibson & Choi, 2018).
After establishing classes, and to answer the research questions, we applied three Multivariate Analysis of Variance (MANOVA).In all MANOVA's, the independent variables are the class variables of TRC and adolescent characteristics.In the first MANOVA, the dependent variables are the three social climate subscales (involvement, support, and spontaneity) of the relationship dimension.In the second MANOVA, the dependent variables are the four social climate subscales (autonomy, practical orientation, personal problem orientation, and anger and aggression) of the personal growth dimension.In the third MANOVA, the three final subscales (order and organization, program clarity, and staff control) belonging to the system maintenance dimension are entered as dependent variables.In all analyses, multivariate interactions between TRC and adolescent characteristics were examined first.If the interaction did not show significance, main effects were investigated.We performed the LCA in Mplus version 7.11 (Muthén & Muthén, 2008) and the MANOVA analyses in IBM SPSS Statistics version 26.For all analyses, alpha levels of <0.05 were considered statistically significant, but we applied a Bonferroni correction for pairwise comparisons and a Tukey correction for multiple comparisons in the MANOVA analyses.In addition, partial eta-squared (η 2 ) is used as a measure of effect size, where values of 0.01, 0.06, and 0.14 are considered small, medium, and large respectively (Cohen, 1988).

TRC characteristics
Table 2 shows the descriptive statistics and Table 3 presents conditional item probabilities for the two TRC classes.The results are based on responses of 142 staff leaders and indicated that a model with three classes showed the best fit.However, we chose the model with two classes due to convergence problems with the three-class solution and better interpretability.
The first class is labeled family-style TRC (54% of the sample).The defining characteristic of this class of TRC organizations is that staff members cohabit with the adolescents, resulting in the label "Familystyle TRC".The second class is labeled larger TRC settings (46% of the sample).TRC organizations in the second class are defined by staff that adhere to a work schedule of shifts.These organizations typically consist of multiple units and adolescents meet on average 13 staff members a week.

Differences between the two classes
Both TRC classes have a mix of ownership and location, although family-style TRCs have a higher rate of private ownership (non-profit as well as commercial) and tend to be more often located outside cities, in rural areas.The family-style TRC class also tends to have leaders who are younger and fewer with female gender, without any class difference in leader competence or experience.Both classes have a mix or placement length, with more acute placements in family-style TRCs.Youths in family-style TRCs tend to participate more in organized and unorganized activities outside TRC, more often receive visits from friends in the TRC, and tend more to make friendships in care.
Family-style TRCs tend to be somewhat less organized with regards to meetings and regular tasks for adolescents, but more organized regarding detection and handling of substance use.Coercive measures were rarely used in both classes.They were to a lesser degree connected to behavioral problems and in general occurred less often in family-style TRC.
Although there seemed to be differences in mode averages, mean average differences were not significant regarding staff: adolescent ratio, daily/weekly staff contacts, staff competence, numbers of adolescents or staff per unit.This may be due to wide variation in both classes on these characteristics.

Adolescent characteristics
Descriptive statistics for adolescent characteristics are presented in Table 4. Model fit indices indicate that four different adolescent classes can be identified.Although a model with five classes also shows acceptable model fit, we prefer a model with four classes due to interpretability aspects and higher BIC values.Conditional item probabilities are presented in Table 5.

Severe problems class (38% of the sample)
The largest class is defined by having at least one psychiatric disorder based on our diagnostic interview with the highest prevalence in all major categories of psychiatric disorders.In addition, they have the highest rate of having received a clinical psychiatric diagnosis at any timepoint before the study.Their emotion dysregulation scores are the highest and above average of all classes.Furthermore, they have the lowest probability for attending school, and highest rate of school problems and receiving help for those problems.These adolescents tend to be non-voluntarily placed females.Their parents have increased prevalence of several problems (not somatic illness), but not as high as in the family problems class.Out of twelve individual and social problem categories, these adolescents had at least four and in average more than sixnearly twice as much as the average for the other adolescents.

Incidental problems class (36% of the sample)
This second largest class is defined by low rates of all categories of psychiatric disorders, the lowest rate of previous psychiatric diagnoses, and low rates of most parental problems.The class mainly consists of voluntarily placed males.They tend to have experienced their first placement as difficult and have had a low number of placements.Most of these adolescents are currently in school and show below average school problems.Therefore, this class is labeled as showing incidental problems (not chronic problems).They tend to have been placed due to behavioral problems or drug problems.However, our variable selection does not clearly show their main problems resulting in residential care.

Youth with family problems (13% of the sample)
The smaller third class is mainly defined by a high rate of parental problems, a first placement at a younger age, and the highest number of former placements.The adolescents have a high rate of present (and previous) psychiatric disorders, although lower than the severe problems class.

Adolescents with a migrant background (13% of the sample)
The final small class is mainly defined by their non-Norwegian origin or native language.These adolescents have experienced violence or other problems with their parents and have somewhat less contact with their family during placement.They tend to be males, have been placed voluntary as adolescents, and had the lowest number of former placements.They have experienced their first placement as more positive and show the lowest prevalence of previous and present psychiatric problems, including drug problems.They show below average emotional dysregulation problems, less school problems, and high school attendance.Their parents are reported to have low rates of all problems except chronic (somatic) illness.

Associations of TRC and adolescent characteristics with social climate 4.3.1. Relationship dimension
We performed a two-way MANOVA with two independent variables (TRC and adolescent characteristics) and three dependent variables (involvement (M = 2.28, SD = 1.31), support (M = 2.74, SD = 1.30), and spontaneity (M = 2.51, SD = 1.15)).There is a linear relationship between the three dependent variables, as assessed by a scatterplot, and no evidence of multicollinearity as assessed by Pearson's correlation (r < 0.90).There are some univariate outliers, but no multivariate outliers as assessed by Mahalanobis distance (p > .001).There is homogeneity of covariance matrices, as assessed by Box's M test, F(42, 51583) = 0.90, p = .663,Box's M = 39.12 and homogeneity of variances, as assessed by Levene's test of homogeneity of variance based on median values (p > .05).
To further explore the differences in mean involvement scores between adolescent characteristics in both the larger TRC settings and the family-style TRC settings we ran simple comparisons with a Bonferroni adjustment for multiple comparisons (see Fig. 2).

Personal growth dimension
A two-way MANOVA with two independent variables (TRC and adolescent characteristics) and four dependent variables (autonomy (M = 2.26, SD = 1.23), practical orientation (M = 2.21, SD = 1.32), personal problem orientation (M = 1.59,SD = 1.18), and anger and aggression (M = 2.30, SD = 1.35)) shows a linear relationship between the four dependent variables, as assessed by a scatterplot and no evidence of multicollinearity, as assessed by Pearson's correlation (r < 0.90).There are some univariate outliers, but no multivariate outliers as assessed by Mahalanobis distance (p > .001).There is homogeneity of covariance matrices, as assessed by Box's M test, F(70, 44894) = 1.06, p = .339,Box's M = 78.30and homogeneity of variances, as assessed by Levene's test of homogeneity of variance based on median values (p > .05).

System maintenance dimension
Finally, we performed a two-way MANOVA for this dimension with two independent variables (TRC and adolescent characteristics) and three dependent variables (order and organization (M = 2.62, SD = 1.21), program clarity (M = 2.42, SD = 1.24), and staff control (M = 2.91, SD = 1.12)).There is a linear relationship between the three dependent variables, as assessed by a scatterplot and no evidence of multicollinearity, as assessed by Pearson's correlation (r < 0.90).There are some univariate outliers, but no multivariate outliers as assessed by Mahalanobis distance (p > .001).There is homogeneity of covariance matrices, as assessed by Box's M test, F( 42  1.20], p = .009and between the severe problems class and the incidental problems class (M = 2.65, SE = 0.10), − 0.52, 95% CI [− 0.90, − 0.14], p = .003.All other comparisons, including the youth with family problems (M = 2.30, SE = 0.18) are not statistically significant (see Fig. 5).

Discussion
The aim of this study was to investigate how classes of TRC and adolescent characteristics are associated with experienced social climate in TRC.Using LCA, results showed two types of TRC and four classes of youth, which differ in their relation to experienced social climate reported by young people living in TRC.
Regarding TRC characteristics, we expected that small and publicly owned TRCs located in more populated areas and having (staff training) routines and staff with relevant education were associated with a positive social climate.However, the results show little variation between the TRC settings and the LCA only indicated two TRC clusters consisting of family-style TRC and larger TRC settings.Both these settings are mostly publicly owned, have routines for youth, staff training programs, and have staff members with mostly a social work degree.They mainly differ in size, as the family-style TRC is smaller, and family-style TRCs have more staff cohabitation compared to the larger TRC settings.
Contrary to our expectation, social climate in terms of the perception of anger and aggression (i.e., the extent to which young people argue, become openly angry, and display other aggressive behavior between each other and staff) was higher in family-style TRC settings than in larger TRC settings.An explanation for this unexpected finding may be that these emotional expressions are more tolerated in smaller familyoriented settings than in larger TRC settings.For example, youth in family-style TRC seem to have a more positive emotional development (Leloux-Opmeer et al., 2018) compared to children in residential care or foster care.Furthermore, there are indications for more attachment related problems in youth admitted to family-style TRC compared to larger TRC settings (Leloux-Opmeer et al., 2017), which may explain why more anger is perceived and expressed.The cohabitation schedule may also provoke more conflict and turbulence.In addition, larger TRC settings may focus more on adaptive behavior of youth during care and, consequently, allow less angry expressions or behavior (Harder, 2018).Overall, our findings do not show evidence that small family-style TRC settings have a more positive social climate compared to larger settings, which is not in line with previous research (Anglin, 2002;Caldwell & Rejino, 1993;Chipenda-Dansokho & Bullock, 2003).A potential problem in explaining this finding is that the variation and differences in TRC characteristics are generally small, within as well between the two identified TRC classes.Future studies on more heterogenous and different TRC types may be necessary to clarify the associations between TRC characteristics and perceived social climate.
Regarding youth characteristics, we identified a migrant background group, youth with family problems, incidental problems, and youth with severe/comorbid problems.The severe problems group showed, as expected, the least positive perception of social climate in terms of involvement, support, spontaneity, autonomy, practical orientation, order and organization, and program clarity.In addition, they experienced the highest level of anger and aggression in the TRC living environmentespecially in family-style TRC.This finding is in line with previous research, which shows that youth with more emotional and behavioral problems perceive social climate more negatively than youth with fewer problems (e.g., Bastiaanssen et al., 2012;Lanctôt et al., 2016;Leipoldt et al., 2019;Pellerin et al., 2020;Robinson et al., 2018;Southwell & Fraser, 2010;Van der Helm et al., 2014).
The results did not support our hypotheses that adolescents with an internalizing psychiatric diagnosis perceive the social climate as more supportive than adolescents with an externalizing psychiatric diagnosis.Previous research shows that staff members use more autonomy granting for youth with internalizing problems, compared to more controlling interventions for externalizing problems in TRC (Bastiaanssen et al., 2012).Therefore, there seems to be a discrepancy in what staff members do and what adolescents perceive.It is important to investigate these discrepancies in future research as previous findings have indicated that autonomy is positively associated with treatment motivation, less aggression ( Van der Helm et al., 2013;Van der Helm et al., 2014), and a focus on youth's strengths in treatment programs (Barton et al., 2008;Barton & Mackin, 2012).Furthermore, our results indicate that youth with family problems perceive the lowest amount of autonomy in the environment, regardless of TRC characteristics.An explanation might be that youth with family problems in residential care have experienced negative parenting at home, e.g., more emotional and physical abuse and neglect (see review of Leloux-Opmeer et al., 2016), which may lead to higher sensitivity toward perceived autonomy restrictions.It is important that staff members are sensitive towards this possible sensitivity and potential helplessness feelings of youths and allow them more autonomy and support development of autonomy skills during treatment.
Although we expected that having a Caucasian background was associated with a more positive perception of social climate, our results show that the migrant background class perceived social climate most positively on all measured aspects.This is not in line with findings from Dutch TRC where youth with a Turkish and Moroccan background perceived lower levels of support (less trust, respect and taken seriously) than native Dutch youth (Sevilir et al., 2020) and evidence where Dutch (Knorth & Eldering, 1998) and Norwegian (Drange & Telle, 2021) children with an immigrant background have less opportunities for adequate care.A possible explanation is that the migrant background youth (and this also accounts for the incidental problems group) have less mental health problems and the shortest treatment history.That might make them less burdened by problems and more optimistic about their environment, influencing their experience of TRC social climate.In other words, the migrant background group (and incidental problems group) might have a less pessimistic attitude and be more adaptive during placement with subsequently more positive perceptions of their environment (Southwell & Fraser, 2010).
Finally, regarding the interplay of TRC and youth characteristics and associations with social climate, we found only one significant interaction.Only youth with family problems had a more positive perception of involvement in smaller family-style TRC settings, than in larger TRC settings.Opposite to this, the incidental problems class perceived involvement as higher in larger TRC settings than in smaller family-style TRCs settings.A potential explanation for this is that youth with familyrelated problems benefit more from a smaller family-style setting, because of the nature of their problems (e.g., mental, or chronic illness of mother) and mental health problems (e.g., in terms of youth DSM-IV diagnoses) than youth with incidental problems.Youth with familyrelated problems also show the most attachment-related problems and previous research shows that youths with these problems seem to be placed more often in family-style TRC than in larger TRC settings (Lee & Thompson, 2008;Leloux-Opmeer et al., 2017).Consequently, a familystyle setting might fit better with their needs, resulting in a higher perceived social climate in terms of involvement.Youth with incidental problems seem to have different needs, because they report few family problems and might not have a need for gaining positive experiences in a new family setting.Consequently, a family-style setting might fit poorly with their needs and results in a lower perceived social climate in terms of involvement.However, they also show mainly behavioral problems making them more likely to be placed in non-family style TRC.

Strengths and limitations
A strength of this study is the usage of a person-centered approach over a variable-centered approach (Von Eye & Bergman, 2003).Many studies in TRC concentrate on effectiveness, but rather overlook how these effects are obtained.By adhering to a person-centered approach, we were able (1) to include much contextual information to gain a better understanding of TRC types and its users, and (2) to identify what works for whom (Harder & Knorth, 2015) in terms of perceived social climate.Furthermore, practical implications can be generated from this approach by tailoring the emphasis on different social climate aspects to adolescent characteristics in order to improve their perception of social climate.
This study also has some limitations.First, the obtained results are cross-sectional and therefore make it difficult to generalize how social climate might change (the group dynamic) over time and provide causational interpretations.Longitudinal studies monitoring change and development of social climate to increase our understanding of this dynamic interplay are underway (Strijbosch et al., 2019).Second, most data came from youth self-report measures, including information regarding behavioral problems of youth and parental problems.This limits the objectivity of the number of problems that may exist according to caregivers and could potentially have underestimated the number of problems in the present study.Caution is therefore necessary when considering the parental problems in this study.
Third, not all items discriminated well between groups, which indicate that the groups are to some extent homogeneous.Although caution is therefore necessary with the interpretation of the description of classes, the present overlap also provides an indication of the low variation between TRC settings in Norway.Therefore, our results may not be valid beyond this relatively narrow variation between TRC's.A final limitation is that information regarding TRC characteristics was provided only by staff leaders and not by the (daily) staff members that work with the adolescents.This can potentially limit the accurateness of the perception of daily tasks carried out by staff members and how they are experienced by youth.

Implications and future directions
Despite the limitations, this study provides several implications for clinical practice and future research.First, staff members should consider the heterogeneity of adolescents within TRC and acknowledge that they have different environmental needs.For example, a continuous conversation about the perception of the residential environment and how this can be improved can be beneficial for both adolescents and staff members (Leipoldt & Strijbosch, 2020;Levrouw et al., 2020;Knorth et al., 2004).Expectations from and towards the resident youth should be synchronized at the start of treatment during an intake by discussing with adolescents how they expect and prefer the social climate to be in care.This can be done by using the revised CPES questionnaire (Leipoldt et al., 2018) with a reformulation of the questions to an expected social climate style (see Moos, 2003 for a description).By using this questionnaire at intake, expectation management can already be performed before treatment starts, and this subsequently can increase positive and more realistic expectations.This study shows that youth with the most problems perceive social climate the least positive.Therefore, youth care workers should pay extra attention to these youth and not only focus on the many present problems, but also on how they can optimize the environment for these youth based on their needs.
Future research should focus on longitudinal studies to determine how the perception of social climate interplays with youth development and organizational changes.Having a single snapshot of social climate limits our understanding of how social climate perception changes during treatment and whether adjustments are necessary.In addition, future research should investigate the discrepancies between staff and youth perceptions of autonomy and other aspects of the environment to align experience, expectations, and action to both perspectives.Finally, the findings from this study should be extended to include treatment outcome indicators, such as quality of life, treatment satisfaction, and changes in emotional and behavioral problems.A study that includes youth and contextual factors, treatment outcomes, and social climate as covariate can provide a better view of what works for whom in TRC.

Conclusion
This study has contributed to a clearer picture of how social climate aspects could advance the existing knowledge on "what works for whom" in TRC, including both TRC and youth characteristics.This study indicated that adolescent characteristics interact with the practical organization of staff-youth relationships in producing different social climate experiences.A close staff-youth relationship may counteract the differences in adolescent characteristics, but that this is more challenging for some groups of adolescents.In addition, this study indicated that the perception of social climate is not homogenous and that staff members should consider the heterogeneity of their adolescents when changing aspects of the social climate aspects within TRC.This clearly requires more differentiation within and between TRCs and to move beyond the 'one size fits all' approach.

Fig. 2 .
Fig. 2. Interaction effects of TRC and adolescent characteristics on the involvement subscale.

Fig. 4 .Fig. 3 .
Fig. 4. Estimated marginal means of the autonomy, practical orientation, and anger and aggression subscales for adolescent characteristics.¤ Significant difference between the migrant and family class.& Significant difference between the incidental and family class.* Significant difference between the severe and incidental class.$ Significant difference between the incidental class and the severe and family class.^ Significant difference between the severe and migrant class.¥ Significant difference between the severe and incidental class.

Fig. 5 .
Fig. 5.Estimated marginal means of the order and organization and program clarity subscales for adolescent characteristics.# Significant difference between the severe and incidental class.* Significant difference between the severe class and the incidental and migrant class.
(continued on next page) J.D.Leipoldt et al.

Table 3
Conditional Mean and Item Probabilities for TRC Characteristics.

Table 5
Conditional Mean and Item Probabilities for Adolescent Characteristics.