Associations between Secure Residential Care and Positive Behavioral Change in Adolescent Boys and Girls

ABSTRACT

Secure residential youth care facilities try to optimize their help by offering gender-specific treatment, in an attempt to achieve positive behavioral change in adolescents. In this study, we examined behavioral change in a sample of 239 Dutch adolescents (M age = 15.59 years, SD = 1.36 years, 54.9% girls) in secure residential care. Pretest, posttest and follow-up measurements were carried out for behavioral problems, PTSD symptoms, emotion regulation, perceived competence and family problems. Comparisons were made between girls in gender-specific care, and girls and boys in regular care. Missing data analyses revealed the dataset contained many missing values. Analyses were performed at group level, using MANCOVA, ANCOVA’s and bootstrapped planned contrast, and at case level, using the Reliable Change Index. At group level, results revealed higher effectiveness of gender-specific care for girls compared to regular care for girls, only in diminishing externalizing behavioral problems. Overall, there were more similarities than differences in the effectiveness of gender-specific versus regular help. At individual level, 0–58% of the adolescents improved during their stay in secure residential care. However, most adolescents showed no change (25–88%) or even deterioration (0–39%). These results strongly emphasize the need for alternative interventions.

KEYWORDS:

• Behavioral problems
• secure residential youth care
• gender
• PTSD
• parenting

Practical Implications

• Girls in gender-specific care show the largest decrease in parent-reported externalizing problem behavior, underlining the importance of gender-specific care

• Boys and girls develop similarly in regular SRC, which refutes part of the criticism of this type of care

• A significant decrease of behavioral problems is mainly seen in adolescents with the most severe behavioral problems at admission

• No change or even deterioration can be seen especially in adolescents who have no behavioral problems upon admission

The findings underline the necessity of careful screening for treatment needs.

Introduction

Secure residential youth care (SRC) offers treatment to adolescents suffering from multiple problems (e.g., (complex) trauma, insufficient emotion regulation, parents lacking parenting skills, externalizing and internalizing behavioral problems; Eltink et al., 2017; Nijhof, 2011). However, the most common reason for placement in SRC is adolescents exhibiting serious behavioral problems and regularly growing up under adverse family circumstances (Eltink et al., 2017; Martin et al., 2017). Therefore, the primary goal of the treatment in SRC is reducing behavioral problems and improving parenting skills, but also diminishing risk factors and strengthening protective factors (Eltink et al., 2017; Martin et al., 2017).

For various reasons, secure residential treatment of youth has been heavily criticized in recent years. The possible iatrogenic effects of incarceration are an essential element of this criticism (Weis et al., 2005) and are caused by occupational and psychosocial deprivation, coercion, repression (Van Ijzendoorn et al., 2020), deviancy training (Souverein et al., 2013; Weis et al., 2005), or the restriction of autonomy of the adolescents (Ryan & Deci, 2017). In addition, secure residential youth care is the most expensive type of youth care that provides treatment in institutions, instead of within the family (James, 2017).

Since the percentage of girls present in SRC has increased significantly in recent years (Griffith et al., 2009), another point of criticism is that treatment in secure residential youth care is mainly based on knowledge about boys (Nijhof et al., 2012). In an attempt to counter part of the critiques, some residential youth care facilities offer gender-specific treatment to make treatment more attuned to the specific risks and needs of boys and girls. This development is based on the principle that girls have different risks and needs than boys (Handwerk et al., 2006; Nijhof, 2011; Sonderman et al., 2015; Weis et al., 2005), including differences in exposure to risk factors. Factors that mainly put girls at risk for problematic psychological development are, for example, past victimization or traumatic experiences, destructive families and substance abuse (Anderson et al., 2019; Nijhof et al., 2018; Sonderman et al., 2015). Furthermore, girls tend to show more internalizing behavioral problems, such as depression and anxiety, at admission to SRC than boys (Handwerk et al., 2006). In addition, girls show more severe problems in the parent–child relationship and in other social relationships than boys (Nijhof et al., 2018). Lastly, girls referred to SRC suffer more frequently than boys from a (mild) intellectual disability than boys (Nijhof et al., 2018). Boys on the other hand show more severe substance abuse and problems with school and work than girls (Nijhof et al., 2018). Research also shows that girls are more sensitive to trauma and show more behavioral problems and depression afterward than boys do (Nijhof et al., 2018). In addition, girls are also more sensitive to the protective effect of healthy social relationships than boys (Nijhof et al., 2018).

The development of gender-specific treatment is prompted by the risks, needs and responsivity (RNR) principles of Andrews and Bonta (2010). These principles state that to optimize the treatments’ effectiveness (i.e., reducing behavioral problems, also through, for example, reducing trauma problems and the improvement of trauma problems and parenting skills) the treatment should match the risks and needs of the adolescents. Although more and more organizations pay attention to gender-specific treatment, we are still in the early stages of implementation. Several meta-analyses have shown that adolescents can benefit from treatment in a secure residential youth care facility (Strijbosch et al., 2015; De Swart et al., 2012); however, research on the outcomes of gender-specific treatment is scarce. Therefore, whether organizations can improve outcomes for adolescents by offering tailored SRC is still unclear.

Some findings regarding characteristics of and outcomes for boys and girls in SRC are available. In 2006, Handwerk et al. showed that girls in SRC demonstrated higher rates of internalizing behavioral problems than boys at admission, but also a significantly greater reduction in those problems during treatment. Regarding SRC in The Netherlands, improvement in behavioral problems was found in only 24% (Gevers et al., 2020) to 46% (Dirkse et al., 2018) of adolescent boys and girls. In both studies, no significant differences in effectiveness were found for boys and girls. Overall, this costly and autonomy limiting (Ryan & Deci, 2017) intervention shows poor results in terms of reducing behavioral problems, as shown by studies of Gevers et al. (2020) and Dirkse et al. (2018).

The Present Study

Despite the criticism about secure residential youth care, for some adolescents the intensity and protective context of SRC is inevitable (e.g., when foster care or less restrictive types of institutional youth care have failed, the immediate safety of the adolescents is at stake or the adolescent withdraws itself from the necessary care; Ainsworth, 2017; Gutterswijk et al., 2020). For this reason, it is important to know if and how SRC can be effective in supporting positive psychological development of adolescents for whom other interventions are not considered suitable. Therefore, the aim of this study is to examine the effectiveness of two SRC facilities.

The first facility (called “Hestia”) offers gender-specific care for girls only. The second facility (called “Midgaard”) offers “regular” (i.e., non-gender-specific) secure residential youth care (for more information, see Methods). To investigate whether gender-specific care is of added value for girls in SRC, we compare the effectiveness of (1) gender-specific care for girls to the effectiveness of “regular” care for girls. In addition, we compare the effectiveness of (2) “regular” care for girls to the effectiveness of “regular” care for boys. The following hypotheses were tested, on both the group and the individual level: (1) gender-specific SRC will result in greater improvements of girls’ externalizing and internalizing problem behavior, symptoms of PTSD, perceived competence, the parent–child relationship and parenting skills than “regular” SRC for girls, since gender-specific care is better tailored to the needs of girls than “regular” care, and (2) “regular” SRC for boys and girls results in greater improvements of externalizing and internalizing problem behavior in boys than in girls, since this type of care is mainly based on knowledge about boys.

Methods

Participants

The study population consisted of 239 adolescents (aged 12–18 years) admitted to two secure residential youth care facilities in 2017, 2018, and 2019 in The Netherlands (called “Hestia” and “Midgaard”). By using convenience sampling, a total of 318 cases were examined for eligibility (see Figure 1). In total, 239 cases (75%) were included in this study (Girls from Hestia, n = 59; Girls from Midgaard, n = 73; Boys from Midgaard, n = 107) (for a brief description of the sample, see Table 1). In our sample, there were no adolescents who self-identified as non-binary or transgender. We therefore consider all participants as cis.

Table 1. Descriptives: Pre- and post-intervention and follow-up scores of all participants and percentages of adolescents with problems within the clinical range at admission.

		Boys reg							Girls reg							Girls gen
	M		SD						M		SD					M		SD
Age at admission (yrs.)	15.59		1.44						15.59		1.41					15.57		1.30
	N	T1	% Problematic	N	T2	N	T3	N	T1	% Problematic	N	T2	N	T3	N	T1	% Problematic	N	T2	N	T3
Behavioral problems
Internalizing behavioral problems^1	83	3.76 (2.84)	38.6	68	4.22 (3.31)	46	3.22 (2.62)	49	4.63 (2.89)	53.1	42	5.02 (4.05)	29	4.45 (2.60)	47	4.34 (3.03)	42.6	25	3.28 (2.91)	13	3.00 (3.29)
Externalizing behavioral problems^1	83	7.34 (3.64)	68.7	68	3.72 (2.90)	46	4.43 (3.38)	49	6.71 (3.29)	57.1	42	4.48 (3.11)	29	5.72 (3.68)	47	6.23 (3.57)	55.3	25	2.44 (2.76)	13	2.69 (1.75)
Internalizing behavioral problems^2	105	2.34 (2.72)	10.5	68	1.46 (1.84)	42	2.21 (2.70)	70	3.73 (3.22)	20.0	52	1.58 (2.44)	19	1.84 (1.61)	57	3.88 (2.71)	19.3	41	3.44 (3.01)	13	2.54 (2.93)
Externalizing behavioral problems^2	105	3.55 (2.52)	11.4	68	3.10 (2.47)	42	3.24 (2.24)	70	4.14 (2.54)	18.6	52	2.40 (2.38)	19	3.21 (2.64)	57	3.65 (2.25)	10.5	41	3.1 (2.44)	13	2.62 (1.85)
Individual problems
PTSD symptoms^2	99	17.38 (13.23)	20.2	64	20.77 (17.04)	41	27.56 (19.38)	64	29.11 (16.76)	50.0	49	18.31 (17.18)	20	23.70 (20.99)	56	28.09 (18.62)	46.4	37	23.76 (19.62)	12	24.83 (17.76)
Adaptive emotion regulation^2	89	128.21 (34.61)	22.5	66	133.18 (30.44)	46	122.55 (28.86)	62	122.73 (31.30)	29.0	49	125.33 (36.69)	17	116.71 (43.21)	49	123.06 (39.72)	30.6	37	120.68 (28.92)	10	107.60 (28.04)
Maladaptive emotion regulation^2	89	69.00 (18.92)	9.0	66	76.23 (18.43)	42	74.45 (20.55)	62	78.66 (23.09)	22.6	49	75.98 (19.40)	17	76.12 (17.81)	49	78.00 (20.68)	26.3	37	82.86 (21.24)	10	81.50 (33.51)
Perceived competence^2	98	30.14 (6.56)	2.0	65	29.26 (4.32)	43	23.81 (5.58)	66	26.56 (6.28)	4.5	50	27.40 (7.06)	21	22.38 (6.31)	50	24.48 (5.92)	12.0	38	27.16 (5.84)	13	26.54 (6.17)
Family context
Low-quality parent–child relationship^1	75	13.79 (4.69)	50.7	60	12.23 (4.59)	46	11.07 (4.02)	46	11.70 (4.41)	30.4	41	11.61 (4.79)	29	13.17 (3.81)	41	10.73 (4.24)	24.4	23	11.87 (5.04)	13	8.77 (2.86)
Insufficient parenting skills^1	75	18.40 (4.41)	58.7	60	15.53 (4.76)	46	15.13 (4.78)	46	15.98 (4.53)	37.0	41	15.80 (4.93)	30	16.73 (4.23)	42	16.33 (4.46)	38.1	23	15.74 (4.98)	13	12.38 (3.78)

Boys reg, boys in regular RC; Girls reg, girls in regular RC; Girls gen, girls in gender-specific RC.

This table shows the percentages of boys and girls with problems within the clinical range, based on the cutoff scores of the instruments.

¹Parent-report.

²Adolescent-report.

Figure 1. Flowchart showing how participants were selected.

A minimum length of stay of 6 weeks was used as an inclusion criterium since the first 6 weeks of placement are considered as the stabilization and observation phase. After 6 weeks, in general, the more tailored treatment is started (Figure 1 and Table 1)

Settings

Hestia is a gender-specific (“girls-only”) facility offering treatment to vulnerable girls who are, for example, victims of commercial sexual exploitation, excluding girls exhibiting severe externalizing problem behavior or girls who are being known for recruiting girls for commercial sexual exploitation. A trauma-sensitive approach is used. Screening for PTSD-symptoms is part of any treatment pathway and trauma therapy is deemed necessary in almost all cases. Midgaard offers “regular” help to the other girls (and boys) who are referred to SRC. The term “regular” is only used to indicate that the help at Midgaard is non-gender-specific help. In Table 2, the main characteristics of both SRC facilities are presented. Both facilities exclude adolescents who show a (mild) intellectual disability or (serious) psychiatric problems that are so severe that treatment in SRC impossible. In order to receive treatment, the adolescents live in secure residential 24-h care in a living group with a highly structured daily routine. Within 6 weeks after admission, a treatment plan is established, under supervision of a behavioral scientist, in collaboration between the adolescent and his professional and social network. Furthermore, a family counselor is appointed when problems were identified in the family context and in both settings, sociotherapists try to achieve a positive living group climate to optimize treatment results (Van der Helm et al., 2018). Moreover, additional individual therapy (e.g., trauma therapy of family therapy) is offered when indicated by a behavioral scientist of psychiatrist. Lastly, for some of the adolescents, pharmacotherapy is used for the treatment of, for example, ADHD, depression or sleep problems (Table 2).

Table 2. Key criteria of the two settings.

	Regular SRC (“Midgaard”)	“Gender-specific” SRC (“Hestia”)
Facility
Capacity	60	30
Gender	Boys and girls	Girls
Living group
Number of adolescents per living group	8–10	8–10
Number of sociotherapists per living group	2–3	2–3
Treatment
Main treatment goals	Reducing behavioral problems, improving parenting skills, improving emotion regulation	Reducing behavioral problems, improving parenting skills, improving emotion regulation, preventing revictimization, improving empowerment, reducing PTSD symptoms
Treatment approach	Solution-oriented approach, system-oriented approach, cognitive behavioral approach, presence approach, social competence model, positive working alliance, shared decision making, informal mentoring, motivational interviewing, positive living group climate	As regular SRC and in addition: trauma-sensitive approach, social network analysis, family therapist involved in every case, psychomotor group therapy, workshops on healthy relationships, sexuality and intimacy.
Individual therapy (e.g., trauma therapy, family therapy, psychomotor therapy)	56% of adolescents received individual therapy	90% of adolescents received individual therapy
Average length of stay	203 days	202 days

Procedure

Within 3 weeks after admission, research assistants provided parents and adolescents with a paper version of the questionnaires, either at home or in the institution. Furthermore, a written informed consent was obtained from the adolescents and their parents (or legal guardians). The questionnaires were filled out at admission (T1), at discharge (T2) (range = 43–636 days after admission), and at follow-up (T3) (6 months after discharge) by a biological parent (in some cases substituted by a legal guardian) and the adolescents themselves. The response rate for the measurement at admission was 89% for adolescents and 68% for parents; at discharge 61% for adolescents and 53% for parents; and at follow-up 31% and 35%, respectively. During the filling-out of the questionnaires, the researcher remained within reach of the informant to offer help when/if necessary. The data was pseudonymized before processing, according to the guidelines of the medical ethical review committee. After submitting the research proposal to the medical ethical review committee, we were exempt from the reviewing process (TWOR – 2018-24).

Measures

Externalizing and internalizing problem behavior. The Dutch versions of the Brief Problem Monitor Parent-version (BPM-P) and the Brief problem Monitor Youth-version (BPM-Y; Verhulst et al., 1997) were used to identify psychosocial problems, and were filled out by parents or substitute caregivers and by the adolescents themselves. Two scales of the BPM’s (19 items) are used in this study: ”externalizing problem behavior” (7 items), and “internalizing problem behavior” (6 items). Answers are given on a three-point scale (0 = not true, 1 = sometimes true and 2 = very true).

Symptoms of post-traumatic stress disorder. The Children’s Revised Impact of Event Scale (CRIES-13) (Verlinden & Lindauer, 2015) screens for signs of post-traumatic stress disorder (PTSD). The instrument (13 items) is used to ask the adolescent what impact a certain stressful event has had on his well-being the past 7 days. The answers are given on a four-point scale (0 = never, 1 = rarely, 3 = sometimes and 5 = often) (Verlinden & Lindauer, 2015).

Perceived competence. Adolescents rated the subscale “intrapersonal empowerment” (8 items) of the Empowerment questionnaire (EMPO 3.1; Damen et al., 2017), to measure perceived competence. The answers are given on a five-point Likert-scale (1 = totally disagree, 2 = disagree, 3 = do not agree/do not disagree, 4 = agree, and 5 = totally agree).

Parent–child relationship and parenting problems. The Parenting Stress Questionnaire (OBVL; Veerman et al., 2014) is a self-report questionnaire (34 items), filled out by parents, and measures the stress parents can experience in the upbringing of their children. The questions can be answered on a 4-point-scale (1 = doesn’t apply, 2 = applies a little, 3 = applies fairly and 4 = applies completely). In the present study we used the subscales “parent–child relationship” (6 items) and “parenting problems” (7 items).

Emotion regulation. The FEEL-KJ (FEEL-children and adolescents; Grob & Smolenski, 2013; 90 items) is an instrument to measure emotion regulation and is filled out by the adolescents themselves. The possible answers are “Almost never,” “rarely,” “occasionally,” “often,” and “almost always.” We used two subscales of the instrument: “adaptive strategies” (42 items) and “maladaptive strategies” (30 items).

For further information about the measures, see the online appendix.

Data Analysis

The Statistical Package for the Social Sciences (SPSS version 25, IBM, Armonk, NY, USA) was used to test for overall differences over time, between girls in gender-specific care and girls in regular care and between boys and girls in regular SRC, using both an overall multivariate analysis of covariance (MANCOVA) on all outcome measures (with age as a covariate), as well as separate bootstrapped ANCOVA’s and bootstrapped planned contrasts for individual outcome measures (with age as a covariate). The independent variable was the treatment group (“Girls Hestia,” “Girls Midgaard” and “Boys Midgaard”) and the dependent variables were problem behavior, symptoms of PTSD, adaptive and maladaptive emotion regulation, perceived competence, the parent–child relationship and parenting skills. We first tested for a general multivariate effect on T2-T1 difference scores for the subset of cases with complete T1-T2 data on all outcome variables, using listwise deletion. Then, conditional upon this multivariate test, we performed separate univariate tests on the larger set of all available T2-T1 difference scores for all outcomes. Given the amount and distribution of missing data, we opted for a dual approach of reporting both a follow-up analysis on all available T1-T2 data, as well as an analysis using smaller restricted sample consistent with our initial multivariate analysis. For any significant outcome, we then tested bootstrapped simple contrasts on T2-T1 difference scores for girls in regular vs gender-specific care, and for boys vs girls in regular care. Finally, for any consistently significant pattern of group differences on T2-T1 difference scores, we then performed bootstrapped simple contrasts on T3-T1 difference scores to explore whether the group differences were still present at T3.

To test whether the complete vs missing data both revealed a significant T1-T2 decrease which are comparable in magnitude, we split our dataset into two non-overlapping sets: a set of cases that was complete (i.e., both T1 and T2 were present), and a set of cases that contained missing data (i.e., either T1 or T2 were present but not both).

In addition, significant change at the individual level in the different outcome measures was examined using the Jacobsen–Truax Reliable Change Index (RCI; Jacobsen et al., 1999). The RCI allows to determine whether change from pretreatment to discharge, and from pretreatment to follow-up is not the result of random measurement error. Accordingly, difference scores that were possibly due to random measurement error were classified as ‘no change.’ Difference scores that were not the possible result of random measurement error were classified as ‘improvement’ or ‘deterioration.’

Results

Missing Data Analysis and Group Comparisons at Baseline (T1)

For an overview of our missing data analysis and group comparisons at baseline, see the supplementary online appendix.

Given the percentages of missing data, their distribution across conditions and time, and the observed differences at baseline, we decided not to impute missing values. Instead, we decided to set up a conservative conditional sequence of statistical tests on the available cases – given the restrictions imposed by the pattern of missing data – that aimed to preclude Type-I errors while still allowing to test planned contrasts between the intervention conditions.

Differences in Psychosocial Development between Girls in Gender-specific Care and Girls and Boys in Regular Care

Group Comparisons over Time (T1-T2)

A MANCOVA with age and all outcomes at baseline (T1) as covariates (boys regular care n = 20, girls in regular care n = 11, and girls in gender-specific care n = 15), indicated that the three treatment groups differed in their T2-T1 difference scores (F(20,46) = 1.85, p < .05, Wilk’s Λ = 0.307, partial η2 = .46).

Separate univariate tests on all available data indicated that the multivariate effect was driven by the externalizing behavioral problems reported by parents: an ANCOVA with age and baseline (T1) as covariates (boys regular care n = 54, girls in regular care n = 26, and girls in gender-specific care n = 24), indicated that the three treatment groups differed in their T2-T1 slopes (F(2,100) = 4.36, p < .05, partial η2 = .08. This effect is mathematically equivalent to an ANCOVA with age and baseline outcome score (T1) as covariates and T2 scores as the outcome). Given that the overall multivariate test included a subset of participants included in the univariate follow-ups (due to missing data), we additionally ran the univariate tests using the identical subset included in the initial MANCOVA (boys regular care n = 20, girls in regular care n = 11, and girls in gender-specific care n = 15). Here, we also only observed a significant effect for parent-reported externalizing behavioral problems (see Note 1). No other univariate tests were significant.

Corroborating the univariate test, follow-up bootstrapped simple contrasts revealed that the T2-T1 difference scores for externalizing behavioral problems reported by parents were significantly larger for girls in gender-specific care vs regular care (b = – 2.36, SE_b = 0.81, BCa 95% CI [−3.86, −0.97]). The possibility of limited statistical power precludes strong conclusions; however, the findings seem to indicate a significantly larger decrease of externalizing problem behavior in girls in gender-specific care vs regular care. No significant difference was observed for girls vs boys in regular care (b = – 0.64, SE_b = 0.73, BCa 95% CI [−2.01, 0.64]). Bootstrapped tests on the estimated marginal means indicated that all three treatment groups demonstrated significant decreases in externalizing behavioral problems reported by parents from T1 to T2 (girls in gender-specific care: M_T2-T1 = – 4.34, SE_MT2-T1 = 0.64, BCa 95% CI [−5.58, −3.06]; girls in regular care: M_T2-T1 = – 2.62, SE_MT2-T1 = 0.55, BCa 95% CI [−3.68, −1.59]; boys in regular care: MT2-T1 = – 1.98, SEMT2-T1 = 0.73, BCa 95% CI [−3.52, −0.21]).

Given the consistent differences in T1-T2 slopes between treatment groups, combined with interpretable non-zero T1-T2 slopes within each treatment group, we decided to explore whether the pattern of results would still be observable at T3, 6 months after T2. Bootstrapped simple contrasts indicated that the T3-T1 decreases for externalizing behavioral problems reported by parents (boys regular care n = 32, girls in regular care n = 22, and girls in gender-specific care n = 13), were relatively larger for girls in gender-specific care vs regular care (b = – 2.43, SE_b = 0.96, BCa 95% CI [−4.65, −0.40]). Here, we again observed the same pattern using the subset of subjects included in the initial MANCOVA (see Note 2). No significant difference was observed for girls vs boys in regular care (b = – 0.40, SE_b = 0.99, BCa 95% CI [−2.32, 1.47]). Bootstrapped tests on the estimated marginal means indicated that the girls in gender-specific care and girls in regular care, but not boys in regular care, showed significant decreases in externalizing behavioral problems reported by parents from T1 to T3 (girls in gender-specific care: M_T2-T1 = – 3.65, SE_MT2-T1 = 0.72, BCa 95% CI [−5.08, −2.21]; girls in regular care: M_T2-T1 = – 1.62, SE_MT2-T1 = 0.73, BCa 95% CI [−3.02, −0.23]; boys in regular care: M_T2-T1 = – 1.22, SE_MT2-T1 = 0.81, BCa 95% CI [−2.66, 0.24]). In sum, we observed a larger decrease in parent-reported externalizing behavior from T1 to T2 for girls in gender-specific care, which persisted to T3 (see Figure 2). Unfortunately, due to the combination of both large drop-out rates and late-inclusions, there were insufficient cases available to analyze treatment effects on externalizing behavior per time point, as a function of missing status (Figure 2).

Figure 2. Estimated marginal means of externalizing behavioral problems reported by parents by Condition and Time. Error bars represent standard errors of the means. (T1-T2: boys regular care n = 54, girls in regular care n = 26, and girls in gender-specific care n = 24, T3: boys regular care n = 32, girls in regular care n = 22, and girls in gender-specific care n = 13).

Although we observed consistent results over different subsets of the available data (multivariate and univariate analyses on T1-T2 with n = 46, univariate analyses on T1-T2 with n = 104, univariate analyses on T1-T3 with n = 67), an important question is whether our results may be spuriously due to selection biases (see Note 3). To reiterate, we observed that treatment conditions were not associated with missing status on T1-T2 difference scores in externalizing behavior (see Section 3.1), precluding a selection bias relating to the presence vs absence of data. However, we additionally tested whether the complete vs missing data both showed a significant T1-T2 decrease which are comparable in magnitude. As expected, a bootstrapped within-subject T-test on the complete set (n = 104) indicated a significant T1-T2 decrease in parent-reported externalizing behavior (M_T2-T1 = – 2.86, SE_MT2-T1 = 0.47, BCa 95% CI [−3.81, −1.91]). More importantly, a bootstrapped between-subjects T-test on the set containing missing data (only T1 present with n = 75, only T2 present with n = 31) also revealed a significant T1-T2 decrease which was comparable in magnitude (M_T2-T1 = – 3.70, SE_MT2-T1 = 0.47, BCa 95% CI [−4.96, −2.42]), suggesting that missing status did not have a strong influence on the T1-T2 decrease observed in parent-reported externalizing behavior (see Figure 3).

Figure 3. Estimated marginal means of externalizing behavioral problems reported by parents for Cases without missings vs cases with missings and Time. Error bars represent standard errors of the means. (within-subjects n = 104, between-subjects T1: n = 75, T2: n = 31).

Results at the Individual Level

To examine to what extent significant psychosocial development over time occurred

at the individual level, the RCI was calculated for all outcome measures using the scores on T1. The development of the adolescents could be classified as an improvement, no change or a deterioration (see Table 3).

Table 3. Results: outcomes of all participants based on RCI.

	Boys reg T1-T2				Boys reg T1-T3				Girls reg T1-T2				Girls reg T1-T3				Girls gen T1-T2				Girls gen T1-T3
	N	Improved	Unchanged	Deteriorated	N	Improved	Unchanged	Deteriorated	N	Improved	Unchanged	Deteriorated	N	Improved	Unchanged	Deteriorated	N	Improved	Unchanged	Deteriorated	N	Improved	Unchanged	Deteriorated
Behavioral problems
Internalizing (parent-report)^1	54	12 (22%)	21 (39%)	21 (39%)	41	41 (27%)	11 (49%)	10 (24%)	26	9 (35%)	9 (35%)	8 (30%)	18	7 (30%)	10 (43%)	6 (27%)	24	9 (37%)	10 (42%)	5 (21%)	11	4 (36%)	5 (46%)	2 (18%)
Externalizing (parent-report)^1	54	27 (50%)	22 (41%)	5 (9%)	32	12 (38%)	12 (38%)	8 (24%)	26	10 (39%)	12 (46%)	4 (15%)	21	8 (38%)	9 (43%)	4 (19%)	24	14 (58%)	9 (38%)	1 (4%)	13	5 (38%)	7 (54%)	1 (8%)
Internalizing (adolescent-report)^2	66	21 (32%)	32 (48%)	13 (20%)	41	11 (27%)	20 (49%)	10 (24%)	49	28 (57%)	16 (33%)	5 (10%)	17	12 (70%)	3 (18%)	2 (12%)	40	14 (35%)	15 (37%)	11 (28%)	14	6 (43%)	7 (50%)	1 (7%)
Externalizing (adolescent-report)^2	66	23 (35%)	24 (37%)	19 (29%)	41	15 (37%)	11 (26%)	15 (37%)	49	28 (57%)	12 (25%)	9 (18%)	18	9 (50%)	3 (17%)	6 (33%)	40	9 (23%)	23 (57%)	8 (20%)	12	5 (42%)	4 (33%)	3 (25%)
Individual problems
PTSD symptoms^2	58	12 (21%)	25 (43%)	21 (36%)	38	4 (11%)	16 (42%)	18 (47%)	41	18 (44%)	15 (37%)	8 (19%)	16	7 (44%)	5 (31%)	4 (25%)	37	7 (19%)	24 (65%)	6 (16%)	12	3 (25%)	6 (50%)	3 (25%)
Adaptive emotion regulation^2	56	13 (23%)	28 (50%)	15 (27%)	32	6 (19%)	14 (43%)	12 (38%)	40	12 (30%)	18 (45%)	10 (25%)	15	4 (27%)	6 (40%)	5 (33%)	34	5 (15%)	23 (67%)	6 (18%)	9	3 (33%)	2 (22%)	4 (45%)
Maladaptive emotion regulation^2	56	6 (11%)	36 (64%)	14 (25%)	36	5 (14%)	23 (64%)	8 (22%)	40	8 (20%)	23 (57%)	9 (23%)	13	5 (38%)	7 (54%)	1 (8%)	34	0 (0%)	30 (88%)	4 (12%)	10	0 (0%)	6 (60%)	4 (40%)
Perceived competence^2	59	6 (10%)	46 (78%)	7 (12%)	38	2 (5%)	11 (29%)	25 (66%)	44	7 (16%)	32 (73%)	5 (11%)	18	3 (17%)	8 (44%)	7 (39%)	37	8 (22%)	29 (78%)	0 (0%)	13	2 (15%)	9 (69%)	2 (15%)
Family context
Low quality parent–child relationship^1	42	13 (31%)	22 (52%)	7 (17%)	30	10 (33%)	16 (53%)	4 (14%)	22	3 (13%)	15 (68%)	4 (18%)	18	5 (28%)	8 (44%)	5 (28%)	20	3 (15%)	13 (65%)	4 (20%)	13	1 (8%)	12 (92%)	0 (0%)
Insufficient parenting skills^1	42	13 (31%)	21 (50%)	8 (19%)	30	7 (23%)	19 (63%)	4 (14%)	22	3 (13%)	15 (68%)	4 (18%)	20	3 (15%)	14 (70%)	3 (15%)	20	4 (20%)	14 (70%)	2 (10%)	13	4 (31%)	9 (69%)	0 (0%)

Boys reg, boys in regular RC; Girls reg, girls in regular RC; Girls gen, girls in gender-specific RC.

¹Parent-report.

²Adolescent-report.

We found an improvement of behavioral problems ranging from 22% (Internalizing behavioral problems – parent report – boys in regular care – T1-T2) to 70% (Internalizing behavioral problems – adolescent report – girls in regular care – T1-T3). Additionally, depending on the intervention, the time of measurement and the informant, up to 74% of adolescents who displayed behavioral problems within the clinical range at admission failed to improve during their stay and after discharge. The parent reports seem to indicate that more girls in gender-specific care experience a decrease in their externalizing problem behavior at discharge than girls in regular care do (58 vs 38%). However, the adolescent reports seem to indicate the opposite (improvement in 57% of girls in regular care vs 23% of girls in gender-specific care). No difference in the decreases of internalizing problems was found based on parent report, however, based on the self-reports, 57% of girls in regular care seem to make progression with regard to their internalizing problem behavior, against 35% of girls in gender-specific care. Overall, a decrease of problem behavior was mainly seen in adolescents who had clinical scores at the time of intake, and an increase was particularly visible in adolescents who had non-clinical scores at admission.

Furthermore, PTSD-symptoms decreased mainly in girls of “Midgaard.” According to the self-reports on T1 and T2, progress was achieved in 44% of the trajectories, against 21% (boys “Midgaard”) and 19% (girls “Hestia”). No progress or deterioration was seen in 17% (girls “Midgaard”) to 62% (girls “Hestia”) of the adolescents who displayed PTSD-symptoms at admission.

Regarding emotion regulation, improvement rates ranged from 0% (Maladaptive emotion regulation – girls in gender-specific care – T1-T2 and T1-T3) to 38% (Maladaptive emotion regulation – girls in regular care – T1-T3). For the adolescents who reported problems in their emotion regulation, the percentages of improvement range from 0% (Maladaptive emotion-regulation – girls “Hestia” – T1-T2) to 80% (Adaptive emotion-regulation – boys “Midgaard” – T1-T2).

Progression rates are relatively low regarding perceived competence, with percentages ranging from 5% (boys in regular care – T1-T3) to 22% (girls in gender-specific care – T1-T2). Adolescents who displayed alarming levels of perceived competence at the time of admission failed to make progress in 17% (girls “Midgaard” – T1-T2) to 82% (boys “Midgaard” – T1-T2) of the cases.

Lastly, regarding the quality of the relationship between parents and adolescents, 13–31% of the families experienced progress, between admission and discharge. In more detail, for parents who reported a problematic parent–child relationship at admission, progress was seen in 27% (girls in regular care – T1-T2) to 63% (girls in regular care – T1-T3) of the families. These findings indicate that girls in regular care seem to continue to make progress after discharge in the perception of their parent(s). Improvement of the parenting skills between admission and discharge varies over the different settings and ranges from 13% of the cases for girls in “Midgaard” to 31% of the cases of boys in “Midgaard.” Moreover, progression was seen in 23% (girls in regular care – T1-T2) to 67% (girls in gender-specific care – T1-T3) of the parents who reported insufficient parenting skills at the time of admission.

Discussion

The present study compared the effectiveness of two types of SRC for three groups of adolescents: girls in gender-specific care, girls in regular care, and boys in regular care. We first hypothesized that gender-specific SRC, which is adjusted to the specific needs of girls, would result in greater improvements in girls’ externalizing and internalizing problem behavior, symptoms of PTSD, perceived competence, parent–child relationships, and parenting skills than regular SRC for girls. However, we only found a statistically significant larger decrease of parent-reported externalizing problem behavior for girls in gender-specific care than for girls in regular care, which might indicate that gender-specific care is indeed better suited to the needs of girls. Another possible explanation is the risk of adolescents reinforcing one another’s deviant behaviors (Souverein et al., 2013; Weis et al., 2005). Girls in regular care live together with boys, showing more externalizing problem behavior than girls, making it likely that the risk of “deviancy training” reinforcing externalizing problem behavior is greater for girls in regular care than for girls in gender-specific care. For all other outcomes the intervention effects of both SRC types are similar. A possible explanation for this finding is that adolescent characteristics (e.g., previous care trajectories) or treatment characteristics (e.g., working alliance and living group climate; Ayotte et al., 2016; Sonderman et al., 2015) outside the scope of this study may also be (partially) predictive of the outcomes. Furthermore, the parents of the girls in gender-specific care were likelier to be involved in treatment than the parents in regular care, possibly influencing their views of their children’s problems. Lastly, due to the possibility of limited statistical power of our analyses, we used conservative statistical methods. This can also serve as a possible explanation for the absence of other statistically differences in the treatment outcomes of boys and girls.

Parents also reported that, on an individual level, girls in gender-specific care demonstrated the largest decrease in behavioral problems at discharge. Nevertheless, there was no difference 6 months after discharge. In contrast, girls in regular care achieved the most progress in overcoming externalizing behavioral problems from admission to discharge, according to their self-reports. This difference was much less apparent 6 months after discharge. We believe there are several possible explanations for the results being different at discharge than at follow-up. First, the number of subjects that could be reached during the follow-up was significantly lower than the number at the time of discharge. This finding increases the risk of selection bias if one assumes that respondents with successful outcomes are more easily reached. Another possible explanation is adolescents showing relapse. The fact that Hestia’s intervention invests more in aftercare can partly explain the self-reporting of more positive results in terms of externalizing problem behavior in Hestia girls at follow-up than at discharge. Lastly, the treatment adolescents receive after discharge from SRC was not taken into account in this study. The differing treatments among adolescents can explain individual differences at follow-up. The results found are similar to relatively positive compared to the findings by Gevers et al. (2020) and Dirkse et al. (2018), who found a decrease of internalizing and externalizing problems in 22–46% of the adolescents studied in SRC. However, although a significant part of the population shows improvement in their behavioral problems, an even larger part shows no significant change or even deterioration. These results are alarming, since these adolescents, to make the treatment possible, are placed out of their homes, into a very intensive, restrictive and expensive type of care. These findings confirm that part of the population may suffer from the iatrogenic effects critics warn about. Some of the possible causes for iatrogenic effect mentioned by Van Ijzendoorn et al. (2020) and Souverein et al. (2013) are coercion and repression exhibited by care professionals, violence amongst the adolescents themselves or re-traumatization.

A clear difference between the groups in our study can also be seen with regard to family problems at the time of follow-up. Here, too, the girls in gender-specific care in most cases improved. In Nijhof’s (2011), parents reported no improvement in family functioning; however, they did report a significant decrease in parenting stress. Second, we hypothesized that regular SRC for boys and girls would result in greater improvement in externalizing and internalizing problem behavior in boys than in girls. However, in contrast to our expectations, we found no significant differences in improvements between boys and girls in regular care at the group level. This finding might imply that the interventional effects of SRC are similar for boys and girls, which could refute the common belief that regular SRC is only effective for treating boys. It is in line with the observation by Griffith et al. (2009) that boys and girls show similar outcomes when leaving residential care, for example, similar behavioral and familial outcomes, and with the findings by Handwerk et al. (2006) noting improvements in the problem behavior of boys and girls during residential treatment. In contrast to our findings, these researchers found greater improvement in internalizing problem behavior in girls than in boys. A possible explanation for not finding a significant difference in effectiveness for boys vs girls can be found in the perception of the living group climate. A positive living group climate, especially the safety experienced by the adolescents, enhances the interventions’ effectiveness (Eltink, 2020). Since girls perceive the living group climate more negative than boys do (Sonderman et al., 2015), this may limit the intervention effectiveness for girls. On an individual level, girls self-reported a larger decrease in externalizing problem behavior than boys as well as in internalizing problems. Parents reported fewer differences in improvement. Boys showed a greater decrease in externalizing problem behavior than girls, while girls showed a greater decrease in internalizing problem behavior than boys. It was striking that progress in externalizing as well as internalizing problem behavior was mainly made by adolescents who had a clinical score at the time of intake and that an increase was particularly visible in adolescents who scored in the normal range at the time of intake. This finding, however, can be caused by what is known as ‘regression toward the mean,’ the phenomenon that if one sample of a variable is extreme, the next sampling of the same variable is likely to be closer to its mean. Accordingly, this finding must be interpreted with caution. We believe a possible explanation for the increase in behavioral problems among these adolescents is deviancy training (Souverein et al., 2013) and the restriction of autonomy (Ryan & Deci, 2017; Van Ijzendoorn et al., 2020).

Regarding PTSD symptoms, no significant differences in improvement were found on the group level. However, on the individual-level PTSD symptoms seemed to decrease mainly in girls of Midgaard. This is noteworthy since it contradicts expectations, as a trauma sensitive approach is used in Hestia. Possible explanations for these findings are that, although the level of PTSD symptoms was similar for girls in both Midgaard and Hestia, trauma issues in girls of Hestia may have been more complex. In addition, although re-traumatization in Hestia is prevented as much as possible, treatment of trauma may initially exacerbate the severity of the symptoms, before they eventually decrease.

Another important finding of our study is that although 22–58% of the adolescents showed a decrease in behavioral problems during their stay in SRC, most showed no significant change between admission and discharge or follow-up for PTSD symptoms, perceived competence, emotion regulation, and family problems. This finding can partly be explained by the surprising fact that most adolescents had a nonclinical score for PTSD symptoms (69%) and perceived competence (95%) at admission. We also found a non-clinical score at admission in the majority of the adolescents for parent–child relationships (62%), parenting skills (63%), adaptive emotion regulation (73%), and maladaptive emotion regulation (84%) (see Table 2). Since these results are based on self-reports, it is possible that both adolescents and parents answered the questionnaires to make their situation at admission appear more positive than it actually was. However, as these adolescents show deterioration of their problems or no significant improvement, it seems that for a large proportion of adolescents SRC is not appropriate and even harmful. Improving knowledge about for whom SRC is (not) appropriate is necessary. Furthermore, extensive screening before referral is crucial.

With regard to the possible iatrogenic effects of SRC, deterioration of externalizing problem behavior was observed in 4–15% of the cases (parent-reported), and in 18–20% of the cases (self-reported). Gevers et al. (2020) study in open and secure residential care found deterioration of externalizing problems ranging from 8% (self-reports) to 13% (parent reports) among adolescents. Furthermore, we found a deterioration of internalizing problems among 10–31% of the adolescents, compared to 6% found by Gevers et al. (2020). A study by Dirkse et al. (2018) revealed that a total of 32% of the adolescents deteriorated in their total problem behavior, as reported by their parents and sociotherapists. Overall, in relation to these previous studies, our findings on deterioration seem comparable and positive. In addition, a large part of the population shows no significant change in their problems (i.e., 25% of girls in regular care regarding externalizing problem behavior, up to 88% of girls in gender-specific care regarding maladaptive emotion regulation). Whereas these findings can partly be explained by the fact that some of these girls had no problems in this area at admission, it is again alarming that, despite the provision of highly specialized care, improvement is not achieved in a significant part of the population.

Limitations

Due to the relatively small sample size, some of our study’s analyses may have suffered from limited statistical power, in particular the MANCOVA. Accordingly, we used conservative statistical methods, which may have reduced the chances of finding statistically significant differences but increased the meaningfulness of the findings. Although the sample size at discharge was sufficient, the sample size for the follow-up measurements was relatively small. While this small sample size reduced the meaningfulness of the findings, these findings are valuable because adolescents from SRCs are difficult to reach for participation in longitudinal research, especially with regard to follow-up measurements. Another limitation resulting from this relatively small sample size is that it was impossible in the analyses to distinguish between adolescents with clinical pretest scores and those with nonclinical pretest scores. Furthermore, it was not possible to analyze the measurements at T1-T2-T3 together in one analysis, due to limited sample size. A second limitation is caused by using convenience sampling. Due to the lack of random assignment, initial differences between girls from “Hestia,” girls from “Midgaard” and/or boys from “Midgaard” can, as confounding variables, cause a threat to internal validity. For example, some girls of “Hestia” were victims of commercial sexual exploitation, whilst girls from “Midgaard” were not. The third limitation we would like to mention is that in the present study we tried to determine the potential added value of gender-specific care for girls, compared to regular SRC, by examining girls who stayed in a girls-only care facility, that uses a gender-specific care approach. Therefore, it is not possible to conclude whether this added value can be attributed to the fact that girls are treated without boys or to the substantially different, i.e., gender-specific, care approach. The final limitation of this study we would like to mention is that the level of treatment integrity and living group climate in the facilities were not part of this study. As we mention elsewhere in the discussion, both treatment integrity and living group climate can serve as a possible explanation for similarities and differences found in the effectiveness of regular and gender-specific treatment.

Implications for Clinical Practice and Future Research

Since previous research has shown that adolescents with externalizing rather than internalizing behavioral problems progress the most during their stay in SRC, and the findings in the present study emphasize that adolescents with the most severe behavioral problems improve during treatment, the accurate screening of problems prior to referring adolescents to SRC is essential. The information gathered should then be used to tailor treatment to adolescent’s risks and needs. Furthermore, the screening should be used to prevent adolescents who fail to improve during placement in SRC from entering this type of care. Alternative interventions have proven to be unsuitable for some adolescents, given the fact that most adolescents entering SRC have experienced failure during ambulatory care, foster care and residential care, and have lost their faith in and motivation for treatment. SRC is therefore mostly seen as a last resort. However, since these adolescents also fail to improve in SRC or even show deterioration, our findings underline the need for developing suitable alternative interventions that can guarantee the safety of the adolescents and prevent them from withdrawing themselves from the necessary care. In addition, less restrictive types of care, preferably family-style care, should be the first option of choice when children are not able to live at home. Especially, Treatment Foster Care Oregon for Adolescents (TFCO-A) shows promising results (Gutterswijk et al., 2020). Furthermore, a first exploration of small-scale residential care shows promising results as well. Professionals in these settings are able to establish a more positive working alliance with both adolescents and their parents and provided a more positive living group climate. Moreover, adolescents experience a more positive relationship with each other in small-scale residential care compared to regular care, feeling safer to express themselves (Nijhof et al., 2020). These types of care should be accompanied by effective family interventions, especially used to improve parenting skills of parents (Eltink et al., 2017; Martin et al., 2017).

Our findings seem to indicate that girls in gender-specific SRC develop more positively in terms of externalizing problem behavior, but not in other outcomes. Although diminishing externalizing problem behavior is the main goal of SRC (Ainsworth, 2017), it remains noteworthy that girls in gender-specific care develop similarly to girls in regular care in regard to PTSD symptoms, perceived competence, and family problems. This finding is remarkable, as the former type of care prioritizes the treatment of trauma and insufficient perceived competence. As previously mentioned, differences in, for example, working alliance (Ayotte et al., 2016), living group climate (Sonderman et al., 2015), or even lack of treatment integrity can serve as a possible explanation. Furthermore, it is conceivable that the girls in both settings differ from each other in terms of characteristics (e.g., treatment motivation, previous received care, adverse childhood experiences; Olver et al., 2011). However, previous research has shown that the girls from both facilities on average do not differ in the problems they experience at admission (i.e., internalizing and externalizing behavioral problems, PTSD symptoms, emotion regulation, perceived competence, parenting skills and the quality of parent–child relationships) (First author et al., under review). Another possible reason for the lack of differences is the treatment itself. Although gender-specific care places higher priority on trauma and family treatment than regular care does, facilities providing a regular approach may also offer trauma and family treatment when indicated. The professionals at Midgaard could be successful of tailoring treatment to meet the needs and risks of the adolescents they treat.

Furthermore, our findings support the presence of iatrogenic effects. Most of the adolescents in our sample were referred to SRC because of behavioral problems. However, based on parent reports, 7.0% of the adolescents had neither externalizing nor internalizing problems in the clinical range at admission. It is unlikely that SRC is the appropriate type of care for these adolescents because deterioration of problem behavior was found in these adolescents in particular. To minimize iatrogenic effects, it is important to match the intervention to the risks and needs of adolescents to improve the effectiveness of the intervention (cf. Andrews & Bonta, 2010) and work on a therapeutic residential group climate (Van der Helm et al., 2018).

Our dataset suffered from many missing values and our sample was relatively small, limiting statistical power. This reduced the chances of finding statistically significant differences. To prevent respondents dropping out during the study, we recommend using a more ‘wraparound’ approach, where the measurements are an integral part of the care. Furthermore, to increase the sample size, in future research data should be collected at more SRC institutions. The length of stay of the adolescents in our sample varied greatly (range = 43–636 days). Since the policy in The Netherlands is to ensure that the duration of the placement is as brief as possible, and scientific evidence shows conflicting results for the ideal duration of treatment (Strijbosch et al., 2015; Van Ijzendoorn et al., 2020), it is important to investigate in future research whether placements in (secure) residential youth care with a duration of over 6 months are justified. Another recommendation is to use a repeated measures design to study the development of adolescents during treatment. Using repeated measurements makes it possible to study the link between behavioral change and the length of stay.

Conclusion

In an attempt to better match the content of SRC to the risks and needs of adolescent girls, gender-specific care has been developed in recent years. The findings of this study only partially support the importance of gender-specific care. Adolescent girls in gender-specific SRC do develop more positively during their stay, since their externalizing behavioral problems show a significantly stronger decrease than the externalizing behavioral problems of girls in regular care. However, internalizing behavioral problems, PTSD-symptoms and perceived competence emphasized by the gender-specific approach, show a comparable progress in both gender-specific and regular care. Furthermore, our findings seem to not confirm the criticism that regular care is more suitable for boys. Boys do not develop significantly better in regular care than girls do. Nevertheless, our study confirms an important part of the criticism, that a significant proportion of adolescents fail to show improvement in their problems, and some even deterioration. Despite its intensity and costs, SRC fails to achieve convincing results. Most adolescents fail to show a positive development in their behavioral problems. Organizations should do everything within their power to prevent adolescents, for whom SRC is not appropriate, to be referred to SRC. In addition, the development of alternative interventions is highly necessary. Regarding this development, small-scale residential care and TFCO-A show promising results.^{1, 2, 3}

Disclosure Statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/0886571X.2022.2100561

Additional information

Funding

This research was funded by Reformed Civil Orphanage Rotterdam.

Notes

1. Univeriate tests using the identical subset of subjects included in the initial MANCOVA also indicated only a significant effect for parent-reported externalizing behavioral problems, F(2,32) = 7.24, p < .01, partial η2 = .31, and bootstrapped simple contrasts revealed that the T2-T1 differences were significantly larger for girls in gender-specific care vs regular care (b = – 4.18, SE_b = 1.24, BCa 95% CI [−6.42, −1.61]), but not for girls vs boys in regular care (b = – 2.22, SE_b = 1.33, BCa 95% CI [−4.76, 0.35]).

2. Bootstrapped simple contrasts, using the identical subset of subjects included in the initial MANCOVA, indicated that the T3-T1 decreases were relatively larger for girls in gender-specific care vs regular care, b = – 4.75, SE_b = 1.85, BCa 95% CI [−8.06, −1.35]). No significant difference was observed for girls vs boys in regular care (b = – 1.62, SE_b = 1.76, BCa 95% CI [−4.77, 1.135]).

3. For example, participants may conceivably have dropped out early due to a lack of improvement in externalizing behavior (or larger improvement) and participants may conceivably been started later due to higher levels of externalizing behavior (or lower levels). If so, then T1-T2 slopes would differ for participants with both T1 and T2 data vs. those with only T1 or T2.