Children and adolescents with behavioral problems show an increased risk for developing negative outcomes like school dropout, social exclusion, law breaking, psychopathology, and other health-related problems [1,2,3]. The costs are considerable, for each individual and for the society at large [4, 5]. Empirical findings show that child problem behaviors often operate in concert with poor social competence [6, 7]. This suggests that it may be advantageous to assess both concepts when children show high levels of resistant problem behaviors and are considered for intervention programs. However, in order to identify, monitor and treat behavioral problems and interpersonal challenges, reliable and validated screening instruments are required.
Problem behavior and social competence
A growing body of research has demonstrated an inverse association between child conduct problems and social competence [8,9,10,11]. That is, children who display high levels of conduct problems tend to show low levels of social competence, or the other way round. Persistent child problem behaviors is a well-known risk factor associated with a troublesome future pathway that may follow into adolescents and sometimes into adulthood [12]. On the other hand, social competence is considered a protective factor associated with positive development and healthy functioning [13, 14] that e.g., facilitate school success [15,16,17]. Whereas child problem behavior is apparent through aggressive, oppositional or inattentive behaviors that may cause considerable challenges for parents, social competence is characterized by e.g., awareness about others needs and situation- appropriate behaviors [18]. Children with poor social competence can show inadequate social information processing and poor adaption to situations, which further result in rejection and dislike from peers and exclusion from social groups. Nevertheless, both problem behavior and social competence should be considered in relation to a child’s age, as both concepts are characteristically dynamic and complex in nature. Some have advocated that the two constructs should be considered separate albeit related dimensions of social functioning [19].
A considerable number of studies show that there is a significant association between child problem behavior and social competence, however, not all children with problem behavior are socially unskilled, or vice versa. It may, in fact, be difficult to disentangle the temporal precedence of the two concepts. Some findings suggest that poor social competence may contribute to maintain or increase the intensity of child problem behaviors [20], as it fosters frustration and limits a child’s possibilities to develop positive interactions with others. In addition, it should be noted that the strength between problem behavior and social competence varies considerable across studies, and depend on whether the two concepts are assessed by the same respondents and the bandwidth of the constructs being measured. A recent meta-study based on 54 independent studies among children aged 3–13, showed an overall negative and significant correlation of medium effect size (r = − 0.42) between behavioral problem and social competence [21]. The magnitude was even higher (r = − 0.47, p < 0.001) when both constructs were assessed by parents, but considerable lower when one of the constructs were reported by a parent and the other by a teacher (r = − 0.17, p < 0.001). However, several of the studies relied on different measurement instruments to assess child problem behavior and social competence, and some date back to the 1980s and early 1990s, and have shown varying psychometric properties [22]. Overall, these findings suggest that it could be worthwhile to consider using instruments that are newer and more psychometrically sound, and that address both social competence and problem behavior. This would probably ease use and interpretation of both concepts.
The home and community social behavior scales
The Home and Community Social Behavior Scales (HCSBS [23]) is a parent-reported instrument that provides insight into children’s Social Competence (Scale A) and Antisocial Behavior (Scale B) using the same rating scale. The HCSBS scale is the home version of the School Social Behavior Scales (SSBS [24]). Whereas Scale A assesses the big picture of adaptive and positive social behaviors, Scale B measures the broad concept of socially related problem behaviors. Both scales comprise two subscales. The social competence scale measures Peer relations (e.g., assesses positive skills with peers) and Self-management/compliance (e.g., responses to adult expectations), whereas the Antisocial behavior scale includes the Antisocial/aggressive scale (e.g., overt violation, intimidation or harm to others) and the Disruptive/demanding scale (e.g., behaviors likely to disrupt ongoing activities and place inappropriate demands on others). Validation studies indicate that the HCSBS has good psychometric properties [25,26,27], in addition, two systematic reviews evaluating the Scale A have encouraged the use of the HCSBS to assess social competence [28, 29]. However, a drawback pertaining these studies is that they rely on classical test theory that e.g., treats the ordinal item scores as continuous, and puts emphasis on total scores where all items are equally weighted [30]. Although some studies have considered the social competence scale[31, 32], few studies have evaluated the dimensionality of the HCSBS or its parallel teacher form (SSBS). In the present study we aimed to extend previous knowledge about the HCSBS by applying item response theory (IRT), which is the preferred method for psychometric evaluations [33, 34].
Item response theory
Item response theory (IRT) comprises a family of flexible statistical models and techniques that can be used to evaluate the psychometric properties of a scale, improve scoring accuracy, increase measurement precision, and form the basis for evaluating item properties [35, 36]. In this study, we use polytomous IRT [37, 38] to model the response options for the ordinal observed variables of the HCSBS. Like classical test theory, IRT assumes that the latent construct cannot be measured directly, but instead is measured by a set of indicators. IRT, however, specifies a non-linear relationship between the continuous latent variable and the categorical observed variables. IRT models are typically characterized by the two parameters discrimination parameters and location parameters [39]. The discrimination parameters indicate how well the items can differentiate between participants with different values of the latent variable (similar to an index of item discrimination), while the location parameters inform about the average value of the latent variable associated with a particular response category (similar to item difficulty). With basis in the estimated item parameters, IRT allows for a detailed description of item and scale properties.
The present study
In this study, we utilized IRT to investigate three research questions. First, we investigate dimensionality of the Social competence (Scale A) and the Antisocial behavior (Scale B) scales as measured by HCSBS, by comparing one-dimensional models with two-dimensional models for each scale. For the two-dimensional models, the different subscales of Scale A and Scale B were considered as measuring separate but correlated dimensions of social competence and antisocial behavior, respectively. We evaluate the estimated models with respect to the model fit and item fit. Second, we infer the item and scale properties for the subscales Peer relations (Scale A1) and Self-management/compliance (Scale A2), and Antisocial/aggressive behavior (Scale B1) and Defiant/disruptive behavior (Scale B2), using item category probability curves, item information functions, and test information functions. Third, we estimate IRT model-based reliability indices for sum scores and IRT scores to evaluate the precision of measurement of the scale scores.
The benefits of IRT are a more realistic model for item responses (which accounts for the ordinal nature of the item scores), improved evaluation of item properties, increased measurement precision through pattern scoring, full-information methods for estimation, and access to detailed tools to evaluate scale properties such as scale difficulty via test characteristic curves and scale precision from test information functions. To our knowledge, this is the first study to evaluate the HCSBS using IRT and as such the study provides important additional information on the properties of the HCSBS, which is useful for practitioners and researchers alike.