Assessing change in the behavior of children and adolescents in youth welfare institutions using goal attainment scaling

An Internet-based instrument developed in cooperation with youth welfare professionals (CJD; Christian Association of Youth Villages[45]) and a software company (arielgrafik[46]) was introduced as a standard evaluation tool in 25 day care and residential care institutions of a large youth welfare organization in Germany. Professionals (social workers, caregivers, psychologists, educators) were trained in using the computer-based tool and asked to complete the questionnaires at the beginning of youth welfare services. Children and adolescents were shown by their guardians how to complete the self-report versions. Professionals and young people worked together to set goals at the beginning and rated change in goal behavior about every six months, depending on the procedures of the respective youth welfare service. Follow-up measures with questionnaires (by caregiver and self-report) were performed after the same time interval as that of goal attainment scaling.

Since this was a naturalistic study, time intervals between the beginning of services, initial measurement, and follow-ups, as well as the duration of youth welfare services varied from client to client. Moreover, not every instrument was used with every client. The following analyses show the results for all clients with individual goal attainment scores and, where available, the outcomes concerning social goals as well as mental health or behavior problems. Goal attainment and changes in mental health and behavior problems were calculated by the differences between the initial measurement (t1) and last available follow-up (tn).

The instrument contains two goal attainment scales developed in earlier studies[42, 47]. One scale measures the attainment of individual goals, while the other measures changes in areas of social behavior that are important to most children and adolescents (see Table 1). These generally applicable goals were derived from a Delphi method[48] performed with professionals (social workers, psychologists, teachers, pedagogical practitioners, nurses) and adolescents in participating day care and residential care institutions. The chosen topics were expressed using eight social goals defined by the worst possible behavior (1) and the best possible behavior (7).

Goal attainment on both scales was recorded on a seven-point scale: Goal behavior is exhibited almost never (1), rarely (2), sometimes (3), occasionally (4), frequently (5), usually (6), always (7). In addition, the motivation of the client to change the targeted behavior is documented on a five-point scale ranging from ‘not motivated’ (1) to ‘very motivated’ (5).

In a pilot study, both goal attainment scales were found to be practical and methodically adequate[47]. Inter-rater reliability was good (ICC-coefficient: .68 - .88) with regard to social goals and even very good (ICC-coefficient: .90 - .96) for individual goals. Both scales were sensitive to changes with statistically significant t-values from 4.13 to 7.41 (p < .001)[42]. Construct validity was tested by means of correlations between goal attainment and the decrease of emotional and behavioral problems measured using the Child Behavior Checklist/4-18 (CBCL)[49].

The tool is supplemented by questions about the socioeconomic background, family history, school-related and health problems (basic documentary sheet based on the official German youth welfare statistics). Quality of life was determined using the Inventory for Assessing Quality of Life in Children and Adolescents[50]. There are two versions of this inventory, one for caregivers and one for children and adolescents. Seven items covering different aspects of quality of life are aggregated into a single problem score (0 = no problems, 7 = problems in all areas).

Moreover, mental health and behavior problems were assessed using standardized rating scales as well. The Child Behavior Checklist/4-18 (CBCL)[49] was completed by caregivers and the Youth Self-Report (YSR)[51] by the clients themselves. Both questionnaires comprise eight scales with 120 items: withdrawal, anxiety/depression, somatic complaints, social problems, thought problems, attention problems, delinquent behavior, and aggressive behavior. These scales can be combined into three broadband scores: internalizing behavior, externalizing behavior, and total problems.

Changes in social and individual goals as well as in mental health and behavior problems were tested with t-tests for dependent variables. To describe the extent of change, Cohen’s d was used as a measure of effect sizes[52]. Correlations in between the social goals, between social goals and CBCL/YSR broadband scales as well as between the extent of change and the length of the time interval between measurements were tested using Pearson’s correlation coefficients. Changes in social and individual goals in relation to changes in mental health and behavior problems were tested my means of one-way analysis of variance (ANOVA), where changes in the CBCL total problem score was classified into four groups: (1) no (borderline) clinical behavior at the beginning as well as at the last measurement (T < 60 → T < 60), (2) problematic behavior at the beginning but not at the last measurement (T ≥ 60 → T < 60; decrease of problems), (3) no (borderline) clinical behavior at the beginning but at the last measurement (T < 60 → T ≥ 60; increase of problems) and (4) problematic behavior at the beginning as well as at the last measurement (T ≥ 60 → T ≥ 60). This categorization was chosen to illustrate how changes from clinically relevant CBCL scores to normal scores relate to changes in social goals. Cohen’s f was used as effect size[52]. Since converting CBCL scores into a categorical variable reduces the information contained, correlations between changes in goals and changes in CBCL/YSR were additionally tested using Pearson’s correlation coefficients. The level of significance was set at p < .01. To account for the large number of analyses, we adjusted p-levels using the Bonferroni correction. Effect sizes were calculated using MS Excel, while all other analyses were conducted with SAS 9.3.

Caregivers (in our study they are invariably staff in the participating day care or residential care institutions) used the goal attainment scales with 433 children and adolescents from 2006 to 2010. Ages ranged from 6 to 18 years (M = 14.7, SD = 2.6). Girls (M = 15.5, SD = 1.9) were older than boys (M = 14.3, SD = 2.8; t = 5.52, df = 411.8, p < .001). The young people had been in their current institution for about 8 months before starting goal attainment within our project. Many of the children/adolescents experienced problems in their families, at school, or with regard to health issues (see Table 2). 86.8% indicated at least one critical life event in their past, with on average 3.3 (SD = 1.9) events being reported. Quality of life was rated as impaired by 41.5% of the young people, while caregivers considered it to be even up to 58.3% of the children and adolescents.

Caregivers did not necessarily use all the instruments of the computer-based tool with every child or adolescent: Social goals were repeatedly assessed with 415 young people. Quality of life was evaluated with 429 children and adolescents. Mental health and behavior problems were rated for 406 young people in CBCL and by 398 adolescents in YSR. On average, individual and social goals were assessed 2.72 and 2.93 times respectively. The time lag between the first two measurements was about eight months (individual goals: M = 7.76, SD = 5.83; social goals: M = 8.08, SD = 5.80; CBCL: M = 7.75, SD = 5.16; YSR: M = 8.00, SD = 5.21).

At the initial measurement, caregivers as well as clients rated mental health and behavior problems of children and adolescents as borderline clinical on average (see Table 3). 56% showed less emotional and behavioral problems in caregiver reports after an average of 14 months, in youth self-reports, as many as 64% reported fewer problems. About 20% of the clients were rated as showing borderline clinical or clinical behavior at the beginning and subsequently improved to normal behavior (range: 16.1-21.5%). Internalizing behavior problems and overall problem behaviors in CBCL decreased over the course of time. Adolescents displayed significantly less problem behaviors on all broadband scales of YSR. Effect sizes, however, were small (d = 0.10 – d = 0.31). Correlations between the extent of changes in mental health and behavior problems and the time lapse between the initial and last measurements in both the caregiver and youth reports were negative, suggesting that the decrease of problem behavior was greater after a longer time interval. However, these correlations were not significant.

At the initial measurement, all social goals were rated between “goal behavior is shown occasionally (4)” and “frequently (5)” on average. There were medium to high correlations in between the eight social goals at the initial as well as at the last measurement (see Table 4). Moreover, there were small to medium correlations between social goals and mental health / behavior problems (broadband scales of CBCL and YSR) at the initial measurement (see Table 5). Children’s and adolescents’ motivation to increase competencies ranged from 3.72 for “reliability / compliant to rules” to 4.11 for “independence”. Between 40% and 54% of the clients showed some improvement in social goals, and 65% showed overall improvement (see Table 3). On average, clients were rated significantly more competent in six domains over the course of time. There was no significant change in the goals “reliability / compliant to rules” and “behavior in school / vocational training” (both on the “adaptation” behavior axis). Effect sizes, however, were small (d = 0.14 - d = 0.44). The last column of Table 3 shows the correlations between the extent of changes in goal behavior and the duration between the initial and last measurements. There were significant correlations with regard to two of the goals, namely “independence” and “integration into (peer) groups / friendship”, with greater improvement after a longer time interval.

The magnitude of changes in goal behavior differed in relation to the extent of competencies already exhibited at the beginning of child welfare. Children and adolescents whose social goal behaviors were rated low (total score < 4) at the beginning showed improvement more often (55% - 71%) than young people who were quite competent already (33% - 47%). Young people with deficits became more competent in all eight domains with medium to large effect sizes, whereas goal behaviors of already competent children and adolescents (total score ≥ 4) changed to a lesser degree with only small effect sizes (see Table 6).

An average of 3.50 (SD = 1.81) goals were defined for each child/adolescent (range = 1–15). All in all, 1494 goals were rated. Clients exhibited improvement in 62% of the goals. However, only in 37% was the targeted characteristic met. On average, they displayed a significant goal attainment over the course of time (d = 0.75; see Table 7). Effect size was medium for goals that were classified as “developing a resource” (d = 0.66) and large for goals that were classified as “reducing a problem” (d = 0.84). There was a significant but small correlation between the extent of changes in individual goal behavior and the duration between the initial and last measurements, with more improvement after a longer time interval (r = 0.15, p < .0001).

Furthermore, individual goals were classified by their titles into 20 categories to allow more detailed analyses. Table 7 shows the ten most frequently used goal categories. The goals that were set most often involved behavior and progress in school and vocational training (35.5% of all goals). Depending on the goal category, children and adolescents exhibited improvement in 52% – 75% of the individual goals. The biggest changes occurred for goals having to do with “social competencies” (d = 0.98) and “relationship to family members” (d = 0.94).

The magnitude of changes in goal behavior differed in relation to changes in mental health and behavior problems in the following four domains: future perspective, conflict management, ability to criticize / take criticism, and reliability / rule compliance. Effect sizes ranged from small to medium (see Table 8). The greatest increase in goal behavior occurred in all eight domains if mental health and behavior problems decreased over time (T ≥ 60 → T < 60). In cases where behavior problems became (borderline) clinical while in child welfare services (T < 60 → T ≥ 60), mean goal behavior decreased in six of the domains, but the changes were not significant. In addition, correlations between the extent of changes in social goals and in emotional and behavioral problems (broadband scales of CBCL & YSR) were analyzed. The magnitude of changes in seven goal behaviors correlated with changes in the CBCL total score, while none correlated with the CBCL internalizing problem scale. Changes in the YSR total score correlated with changes in four goal behaviors (see Table 9).

The magnitude of changes in individual goals did not differ significantly between young people for whom the extent of emotional and behavioral problems changed over time and for those with a constant high or low level of problems (ANOVA: F = 1.53, df = 3/1081, p = .206, Cohen’s f = 0.07). However, there were negative correlations between changes in those instruments, meaning that improvement in individual goals was to some extent linked to a decrease of externalizing behavior (see Table 9).

In the youth welfare studies mentioned above, about two-third of the children and adolescents showed overall improvement while in care[53, 54]. In our study, mental health and behavior problems decreased significantly for young people receiving youth welfare services from the adolescents’ as well as the caregivers’ point of view. However, we do not know if this decrease is a result of the environment inside the day care and residential care institutions or if this decrease occurs in relation to other changes in the children’s and adolescents’ lives, e.g. changes having to do with their parents and friends or with usual child development.

Moreover, young people exhibited significantly more socially competent behavior after receiving youth welfare services. The greatest progress was in the areas of “independence” and “management of conflicts”. In four of the eight social goals, young people showed more improvement after a prolonged stay in youth welfare, implying that a longer and thus more expensive care may be worthwhile. Even so, social competencies usually increase with age. Therefore, a comparison with young people who do not receive youth welfare services is needed in order to estimate how much of the improvement may be due to general child development.

In addition, young people with a lack of competent behavior were found to improve more than children and adolescents who were fairly competent to begin with. This could be due to a ceiling effect of the goal scale or a regression to the mean over time. However, even smaller changes of this kind still represented improvement, which goes against the apprehension that young people with competent behavior will imitate the poorer behavior of their fellow youth welfare recipients. Our results so far are similar to those found in other youth welfare studies[53], both using internationally established screening instruments (CBCL, YSR) as well as using a newly developed tool for youth welfare contexts (social goals).

It was particularly in goals involving individual social competencies that young people showed significant improvement in individual goal behavior. Most often, the goals that were set had to do with progress and behavior in school or vocational training. Effect sizes were larger than for social goals and problem behavior. This confirms earlier findings: that individual goal attainment scaling is more sensitive to individual change than are standardized questionnaires and global measurements[34, 36].

While behavior improved in most of the individual goals, the targeted characteristic was only met in about one third of the cases. It seems like the goals were not realistic, suggesting that caregivers need to be trained how to set achievable goals.

In the JES study, more than half of the goals were reached[53]. Comparing this to our findings, however, is difficult, since goal attainment was rated by the professionals alone, whereas in our study, caregivers and young people rated the goals cooperatively, thereby meeting a requirement of the “German law to improve protection of children and adolescents” with regard to enabling young people to participate[55].

Children and adolescents whose mental health and behavior problems decreased over time exhibited the greatest improvement in social goal behavior. Whenever behavior problems increased, there were no significant changes in social goals. There was no significant difference in individual goal attainment regardless of whether mental health and behavior problems of children and adolescents decreased, remained constant, or increased. Individual goal behavior improved under all conditions. Correlations between changes in those instruments, however, suggested that improvement was greater if the problems - especially those involving externalizing behavior - decreased. Considering that individual goals may refer to all kinds of youth behaviors and not only to mental health and behavior problems recorded by CBCL, these findings are in accord with assumptions about goal attainment scaling, namely that it is a flexible instrument that is able to measure even small, but relevant, changes[34].

There are some limitations that need to be mentioned. In this study, a variety of services in the heterogeneous German youth welfare system were analyzed together. There may well be differences between institutions as well as types of treatments that were not examined. Furthermore, there was no control group to compare our findings with, and thus we cannot know how much of the behavioral change might be due to the typical maturation of young people. Another aspect that should be pointed out is that the evaluations were performed by caregivers and young people, who may have overestimated the progress of the children and adolescents, instead of by a non-participating third-party rater. However, it was not feasible in this naturalistic study to also use uninvolved evaluators. A further consequence of the naturalistic study design is the wide range of time intervals between measurements as well as missing data. While a more restrictive study design could eliminate these problems, it would not portray youth welfare routines as accurately.