بهزیستی ذهنی نوجوانان در مراقبت های مسکونی نسبت به جمعیت عمومی

Abstract The aim of this research is to study the subjective well-being of adolescents in residential care and compare it with that of the general population of the same age in Catalonia. Two samples were used: one from the general population in the first year of secondary education (n = 491; 50% boys; mean age = 12.1 years) and another from the residential care population aged 12–13 years (n = 226; 56% boys; mean age 12.5). The questionnaire of the International Survey of Children's Well-Being (ISCWeB) was used. It includes two psychometric scales: the Personal Well-Being Index—School Children (PWI-SC7) and the Overall Life Satisfaction (OLS); the former being adapted for the in-care population. To test the validity of the factorial structure of data for the two groups, a Confirmatory Factor Analysis (CFA) of the PWI-SC7 and different multi-group structural equation models (SEMs) were conducted. The CFA of the PWI-SC7 showed a good fit with the pooled sample and good comparability of correlations and regressions between the two groups. The SEM with constrained loadings allowed us to compare the contribution of the different items on the PWI-SC7 latent variable which was higher in all cases for adolescents in care. Likewise it showed a high correlation between OLS and PWI-SC7 in both populations, being it more intense among adolescents in care. Scores on the OLS and on the PWI-SC7 are significantly lower among adolescents in care. However, according SEM results mean scores of the PWI-SC7 are not strictly comparable between groups. Results challenge public policy concerning children by increasing efforts to promote equal opportunities for the in-care community and improve satisfaction with particular life domains, such as school and residential homes.

Introduction Children and adolescents who enter the protection system in Spain do so to be protected from a situation of risk of abuse or from actual abuse, as in most Western countries. Abuse is understood as a deliberate act, and includes sexual, physical or psychological abuse, as well as by default acts such as abandonment or neglect by parents or caregivers. In line with that proposed by Garbarino (1977) and Belsky (1993), we frame childhood abuse within developmental–ecological models based on systems theory, whereby cause and effect relationships are not identified but rather abuse is understood as the result of a multiple influence of diverse factors and the interaction among them (personal factors related to the children and parents, family relationships and parenting skills, and social context). The existence of these factors increases the likelihood of there being a situation which is harmful to the child, but with no certainty of this (Berger & Slack, 2014). In Catalonia, 70.4% of children in public care entered the child protection system due to neglect in 2009 (MTAS, 2011). As in other industrialized countries (Gilbert et al., 2009), the most common maltreatment among those in the Spanish child protection system is neglect, followed by psychological or emotional maltreatment. In third position we find physical abuse, and finally sexual abuse. Another major area of research is the study of how children are affected by the fact of suffering a situation of family abuse and the effects this can have on their future. Gilbert et al. (2009) used prospective studies to show a strong relationship between the abuse suffered in childhood and later behavioural problems and a moderate relationship with depression, educational achievements and having a job, among others. The results of a European study focusing on the educational pathways of young people who have been in the protection system (Montserrat, Casas, & Malo, 2013) show how there is evidence that children in both residential and family foster care seem to be at risk of exclusion due to unequal opportunities in compulsory and post-compulsory education, and that their pathways often display delays and dropout, even among those who demonstrate greater capabilities for study: the family context and lack of real support for and prioritizing of schooling often shown by the protection and educational systems have influenced the low educational achievements of this population. Currie and Widom (2010) studied the socioeconomic status of those who had stayed in the protection system into adulthood, finding lower levels of education and employment compared to the general population, with worse results for women from a care background. Hager and Runtz (2012) explored the relationship between childhood physical and psychological maltreatment and self-reported physical health in adult women, showing that physical and psychological child maltreatment were significantly associated with greater physical health concerns. But another question is yet to be addressed: how are they when they are in care? The statistical and social invisibility this phenomenon has often suffered from in many European countries (Casas & Montserrat, 2009) has contributed to prolonging problems associated with it. In Catalonia, a region located in north-eastern Spain and from which the results of the present study are drawn, the percentage of students in the school year corresponding to their age at 15 is 69.4% among adolescents in the general population, compared with 31.7% of the population in care, those in residential care faring even worse than children in non-kinship and kinship foster care (Montserrat, Casas, & Bertran, 2013). Difficulties at school, both social and academic, are also highlighted in other studies (Attar-Schwartz, 2009, Del Valle et al., 2009, Martín and Dávila, 2008 and Palacios and Jiménez-Morago, 2007). As for possible mental health problems suffered by those in residential care, in a study conducted in south-western Spain, Sainero, Bravo, and Del Valle (2013) found that 27% of these children aged between 6 and 18 were receiving psychological treatment. However, when they were administered the Child Behavior Checklist (CBCL) and the Youth Self Report (YSR) (Achenbach & Rescorla, 2001) within the context of the study, the percentage of children having a clinical score on one of the scales rose to 45%; according to the authors, this means that many of these children had not been diagnosed or received treatment. It also emerged from this study that 18% of the children in these centres had intellectual disabilities, a relatively unstudied phenomenon in Spain. The present study focuses on the population of children living in residential centres, which on average represents 50% of those in the protection system in Spain. The other half are in family foster care, mostly kinship care, and in general terms children in care represent 0.5% of the general population in Spain (Montserrat, Casas and Bertran, 2013 and Montserrat, Casas and Malo, 2013). However, the poor quality of official statistics on child protection at local and national level constitutes an important gap in Spain, with only a few snapshots available and some of the figures coming from research. Despite the efforts made by the regional autonomous governments to avoid placing children in residential care and to promote family foster care, the fact that half of the children in care in Spain are currently in residential homes reflects the lack of success of these policies. López et al. (2010) conducted a study to identify the factors determining such intensive use of residential care and the reasons why so many children stay for long periods of their lives in this type of placement, as well as obstacles to achieving either family reunification or foster care or adoption. Findings were related to (i) parents with a significant occurrence of alcohol problems and other addictions, with a poor likelihood of rehabilitation, which can explain the lack of family reunification. There were also situations of domestic violence (38%); (ii) children with some psychological problems and difficulties at school; (iii) half the children were placed with siblings, which is indeed a protective factor but a difficult situation for family integration (both for the original family and for foster carers); (iv) two thirds of children had previously been placed in another institution or family, so they were more reluctant to leave the resource; (v) difficulties in forecasting; and (vi) a lack of foster parents emerged as a factor contributing to long-term residential care. And finally, what do we know about the subjective well-being of children in care? There are very few studies on the subjective well-being of children in care. Generally speaking, it is only in recent years that studies have begun to appear which include the perspective of children, some of them focusing on their well-being using not only objective but also subjective data. Subjective well-being (SWB) refers to people's life satisfaction, both overall and for different domains. Although overall satisfaction comprises two components, one more cognitive (how people evaluate their life) and one more affective (emotions associated with life experiences) (Casas, 2011), in this study we refer only to the former. Dinisman, Montserrat, and Casas (2012) studied SWB among adolescents, taking into account recent changes they had experienced in their lives. They found that those who had undergone few changes in terms of parents or caregivers, school, home or area where they lived reported significantly higher well-being than those who had experienced more changes (who were mainly living in single parent families or in care). Therefore, in this study stability appears as a key factor in the SWB of the adolescents surveyed. Tomyn (2013) found that adolescents with unstable living arrangements and who have experienced situations of domestic violence score very low on SWB using the Personal Well-Being Index-School Children (PWI-SC) and have a higher risk of depression. In his Australian study on adolescents at risk, especially with problems at school, and including absenteeism, the author shows that the SWB of those at risk is significantly lower than that of the general population, although the two samples are both within the normal 70 + range on 100. The author attributes this to the resilience shown by many young people at risk despite their difficult situation (Tomyn, 2013). This may be related to the homeostasis theory (Cummins, 2003), whereby SWB is normally quite positive and stable (within a range of values between 70 and 90 points) for most people with an evolved mechanism to maintain their personal well-being. However, this homeostatic system can be challenged when life events exceed people's capacity to cope, and hence the importance of providing resources to compensate for this situation. In the aforementioned study (Tomyn, 2013), adolescents at risk are almost two times more likely than the general sample to be at a high risk of depression. They scored very low, especially for the domains ‘Standard of living’, ‘Future Security’ and “Health”, compared to the general population sample. In addition, women at risk scored lower in all seven PWI-SC domains, being more prone to depression. Satisfaction with school was also much lower among adolescents at risk (also found by Dinisman et al., 2012) compared with the general population. Moreover, like other authors Casas (2011) and Tomyn (2013) observe that personal well-being decreases with age from 12 to 19 for both genders in general population samples. Regarding children in care, different authors have already highlighted the negative effects of instability suffered by children in the child protection system (Sinclair et al., 2007 and Wade et al., 2011). Research focusing on young people who had been in residential care shows that one of the factors that seems to have most influence is the number of centres they have been in rather than the number of years they have spent in the protection system (Del Valle, Bravo, Álvarez, & Fernanz, 2008; Sala et al., 2009 and Silva and Montserrat, 2014). In this field, we find authors such as Montserrat and Casas (2007) and Palacios and Jiménez-Morago (2007), who explore the satisfaction of children and adolescents in kinship care with regard to life and the care they receive. Rees et al. (2012) analysed the SWB of a sample of English children aged 8 to 16 – using both satisfaction with overall life and satisfaction with several life domains as indicators – and observed that children who were not living with their family (children in foster care, in residential homes or in other non-family arrangements) scored significantly lower than the general population. 1.1. The research question The aim of this study is to analyse the subjective well-being (SWB) of adolescents in residential care and compare this to the general population of the same age. We explore their overall life satisfaction and life domains regarding satisfaction with (i) health, (ii) how secure they feel, (iii) opportunities in life, (iv) things they have, (v) their relationships in general, (vi) the school they attend, and (vii) their use of time.

. Results 3.1. Exploratory analysis Table 1 shows the mean response for each of the scales studied (PWI-SC7 and OLS) and for each of the items comprising the PWI-SC7 scale among both the in-care and general populations. Table 1. Descriptive results for the different items and the PWI-SC7 and OLS scales. Satisfaction with: General population Residential care Effect size Cohen's d Your health Mean 9.48 ⁎8.61 .51 Std. dev. 1.13 2.15 How secure you feel Mean 8.88 ⁎7.61 .59 Std. dev. 1.66 2.57 Opportunities in life Mean 8.84 ⁎7.31 .64 Std. dev. 1.60 2.98 Things you have Mean 9.27 ⁎7.27 .98 Std. dev. 1.23 2.61 Your relations in general Mean 9.05 ⁎8.18 .51 Std. dev. 1.29 2.05 The school you attend Mean 8.87 ⁎7.60 .55 Std. dev. 1.63 2.86 Your use of time Mean 8.47 ⁎7.37 .50 Std. dev. 1.85 2.52 PWISC7 Mean 89.78 ⁎77.07 .91 Std. dev. 8.94 17.52 OLS Mean 9.08 ⁎7.10 .82 Std. dev. 1.39 3.14 ⁎ Statistically significant differences: p < .001. Table options Both the overall scores for both scales and those of each of the items studied show statistically significantly lower (p < .001) mean satisfaction scores among in-care adolescents than those of the general population. All effect sizes calculated support that there are differences between the two groups studied — the effect size is medium (d > = 5) or large (d > = 0.8) in all items and scale indexes. Both scales show a relatively high correlation, although this is markedly higher among the in-care population (0.546 in the general population and 0.665 in the in-care population). Table 2 shows the mean response for each of the scales studied (PWI-SC7 and OLS) and for each of the items comprising the PWI-SC7 scale according to the gender of adolescents in care. Boys averaged higher on both scales and almost all items comprising the PWI-SC7 scale (with the exception of Things you have and Your relations in general, where girls scored slightly higher). In addition, we find that with How secure you feel and the OLS these gender differences are statistically significant. The effect sizes calculated show that the differences by gender are meaningless (d < 20) in some items or small (between d ≥ 20 & d < 50) in the others. In fact, the greater effect sizes are found in the two variables mentioned above: How secure you feel (d = 0.33) and OLS (d = 0.36). Table 2. Descriptive results for the different items and the PWI-SC7 and OLS scales by gender among the in-care population. Satisfaction with: Residential care Effect size Boys Girls Cohen's d Your health Mean 8.84 8.31 .24 Std. dev. 1.85 2.45 How secure you feel Mean 7.99 ⁎7.13 .33 Std. dev. 2.27 2.85 Opportunities in life Mean 7.57 6.97 .20 Std. dev. 2.91 3.04 Things you have Mean 7.23 7.32 − .03 Std. dev. 2.61 2.62 Your relations in general Mean 8.08 8.31 − .11 Std. dev. 2.09 2.01 The school you attend Mean 7.85 7.28 .20 Std. dev. 2.52 3.24 Your use of time Mean 7.50 7.21 .11 Std. dev. 2.54 2.50 PWISC7 Mean 78.68 75.04 .21 Std. dev. 16.40 18.72 OLS Mean 7.60 ⁎6.47 .36 Std. dev. 2.99 3.23 ⁎ Statistically significant differences: p < .05. Table options 3.2. Confirmatory factor analysis of the PWI-SC7 An initial model with the aggregated samples relating the items on the PWI-SC7 scale to a latent variable, without constraints and without allowing covariance of errors, showed only an acceptable fit, with a RMSEA of 0.058 (see Table 3, Model 1). We then tested the same model including an error covariance between satisfaction with the things you have and satisfaction with your opportunities in life. This modified model shows a better fit ( Table 3, Model 2; Fig. 1 shows the standardized factor loads with the aggregated samples), which led to us to test that it as an unconstrained multigroup model (Model 3), and then the same model with constrained loadings (Model 4) and then with constrained loadings and intercepts (Model 5). Table 3. Statistical fit of the different structural equation models analysed using the PWI-SC7. Model χ2 df p-value CFI RMSEA (confidence interval) SRMR 1 Initial PWI-SC7 Aggregated samples 47.19 14 .000 .974 .058 (.040–.076) .030 2 PWI-SC7 + 1 with error cov Aggregated samples 39.31 13 .000 .979 .053 (.035–.073) .028 3 PWI-SC7 Model 2 w/o constraints Multigroup 61.12 26 .000 .959 .043 (.029–.058) .024 4 PWI-SC7 Model 2 with constrained loadings Multigroup 75.16 32 .000 .950 .043 (.031–.056) .033 5 PWI-SC7 1 + constrained loadings + intercepts Multigroup 127.22 38 .000 .896 .057 (.046–.068) .048 6 PWI-SC7 + OLS + gender + 1 error cov + constrained loadings Multigroup 105.88 56 .000 .957 .035 (.025–.045) .036 Table options Subjective well-being. Model 2. Aggregated data. Fig. 1. Subjective well-being. Model 2. Aggregated data. Figure options According to Chen (2007) and Cheung and Rensvold (2001) a rule to accept a model with additional constraints is that fit statistics (particularly CFI) do not change more than 0.01. We find that Model 5 shows a bigger decrease in its fit statistics and therefore strong factorial invariance cannot be accepted, suggesting that response styles are different in both groups, with the result that means cannot be compared. We therefore use Model 4 (Table 3) to carry out a comparison of correlations and regressions between the groups, as it shows a good fit and allows us to compare the standardized factorial loadings of the two samples. The data in Table 4 show that all of the standardized factorial loadings on the latent variable PWI-SC7 are higher among the in-care population than among the general population of the same age, and we can therefore state that they contribute more to the subjective well-being of the former than of the latter. Table 4. Confirmatory Factor Analysis of the PWI-SC7 using the multigroup model, with constrained loads. Standardized factor loads (Model 4). Bootstrap ML. 95% confidence intervals. Resamples = 500 General population Residential care Estim Lower Upper Estim Lower Upper Your health ← PWI-SC7 .510 .403 .621 .553 .461 .649 How secure you feel ← PWI-SC7 .570 .485 .664 .740 .656 .813 Opportunities in life ← PWI-SC7 .535 .425 .637 .622 .472 .737 Things you have ← PWI-SC7 .420 .336 .509 .432 .318 .542 Your relations in general ← PWI-SC7 .477 .378 .575 .587 .481 .682 The school you attend ← PWI-SC7 .447 .358 .533 .527 .412 .647 Your use of time ← PWI-SC7 .534 .440 .638 .775 .670 .864 Table options Further analysis shows that among the in-care population the item with the greatest loadings on the latent variable is satisfaction with your use of time, followed by satisfaction with security and, further behind, satisfaction with opportunities in life. Although these items contribute less to the latent PWI-SC7 variable among the general population, they display higher factorial loadings, but in a different order: satisfaction with security, satisfaction with opportunities in life and satisfaction with use of time. It is also interesting to note that the item on the scale with the lowest contribution in both groups is satisfaction with things you have, while this item's contribution on the overall index is very similar for the two groups (0.420 among the general population and 0.432 among the in-care population). Finally, we incorporated the OLS and gender into Model 4 and found that it fits well (Model 6 in Table 3). Table 5 displays standardized estimates with confidence intervals calculated using the bootstrap method (see also Fig. 2 for the in-care adolescent population sample). Table 5. Multigroup structural equation model relating OLS and gender to the PWI-SC7, with constrained loads. Standardized estimates (Model 6). Bootstrap ML. 95% confidence intervals. Resamples = 500 General population Residential care Estimate Lower Upper Estimate Lower Upper PWISC7 ← OLS .654⁎ .541 .750 .743⁎ .643 .825 PWISC7 ← Gender .054 − .033 .137 .008 − .107 .116 OLS ↔ Gender − .012 − .091 .075 − .18⁎ − .309 − .039 Your health ← PWISC7 .494⁎ .372 .616 .547⁎ .452 .65 How secure you feel ← PWISC7 .559⁎ .472 .643 .747⁎ .666 .815 Opportunities in life ← PWISC7 .566⁎ .472 .644 .651⁎ .527 .749 Things you have ← PWISC7 .439⁎ .356 .523 .452⁎ .357 .585 Your relations, in general ← PWISC7 .454⁎ .366 .542 .573⁎ .473 .677 The school you attend ← PWISC7 .459⁎ .374 .536 .547⁎ .446 .659 Your use of time ← PWISC7 .511⁎ .411 .598 .758⁎ .671 .835 ⁎ Statistically significant (p ≤ .05). Table options Multigroup SEM relating the PWI-SC7 to the OLS and gender. Standardized weights ... Fig. 2. Multigroup SEM relating the PWI-SC7 to the OLS and gender. Standardized weights for the in-care population. Constrained loads (Model 6 in Table 3). Figure options We find that the correlation between gender and OLS reaches statistical significance only for the in-care population, correlating lower with in-care girls than boys. However, gender does not display a statistically significant relationship with the latent variable PWI-SC7 in either of these groups at this age. We also find a strong relationship between the OLS and the PWI-SC7 in both populations, although it is stronger among the in-care population (0.65 among the general population and 0.74 among the in-care population).