Exploration and validation of clusters of physically abused children

Abstract

Objective

Cluster analysis was used to enhance understanding of heterogeneity in social adjustment of physically abused children.

Method

Ninety-eight physically abused children (ages 5–10) were clustered on the basis of social adjustment, as measured by observed behavior with peers on the school playground and by teacher reports of social behavior. Seventy-seven matched nonabused children served as a comparison sample. Clusters were validated on the basis of observed parental sensitivity, parents’ self-reported disciplinary tactics, and children's social information processing operations (i.e., generation of solutions to peer relationship problems and attributions of peer intentions in social situations).

Results

Three subgroups of physically abused children emerged from the cluster analysis; clusters were labeled Socially Well Adjusted, Hanging in There, and Social Difficulties. Examination of cluster differences on risk and protective factors provided substantial evidence in support of the external validity of the three-cluster solution. Specifically, clusters differed significantly in attributions of peer intent and in parenting (i.e., sensitivity and harshness of parenting). Clusters also differed in the ways in which they were similar to, or different from, the comparison group of nonabused children.

Conclusions

Results supported the contention that there were clinically relevant subgroups of physically abused children with potentially unique treatment needs. Findings also pointed to the relevance of social information processing operations and parenting context in understanding diversity among physically abused children.

Practice implications

Pending replication, findings provide support for the importance of considering unique treatment of needs among physically abused children. A singular approach to intervention is unlikely to be effective for these children. For example, some physically abused children might need a more intensive focus on development of prosocial skills in relationships with peers while the prosocial skills of other abused children will be developmentally appropriate. In contrast, most physically abused children might benefit from training in social problem-solving skills. Findings also point to the importance of promoting positive parenting practices in addition to reducing harsh discipline of physically abusive parents.

Introduction

There is compelling evidence that physically abused children, as a group, differ markedly from nonabused children on many indicators of social adjustment. They tend to exhibit aggression (Bolger & Patterson, 2001), are often rated by their peers and teachers as socially withdrawn (Salzinger, Feldman, Ng-Mak, Mojica, & Stockhammer, 2001), and they tend to have low peer status and to be socially rejected at high rates (Bolger & Patterson, 2001; Rogosch & Cicchetti, 1994). There is a long history of research designed to examine such differences between physically abused and nonabused children. With the exception of recent literature on resilience among abused children, however, limited attention has been focused on diversity in outcomes for these children.

A developmental psychopathology framework (Cicchetti & Cohen, 1995) suggests that children who experience physical abuse can differ markedly in developmental outcomes (i.e., multifinality). Indeed, some abused children show evidence of resilient functioning (Haskett, Nears, Sabourin Ward, & McPherson, 2006). Although there are evidence-based interventions that show promise for maltreated children, effectiveness of most treatments has been rather disappointing (National Call for Action, 2004). Traditionally, interventions for physically abused children have been designed without considering diversity in adjustment. Understanding heterogeneity could assist in development of interventions tailored for subgroups of physically abused children who share common treatment needs. The purpose of this study, therefore, was to contribute to the understanding of diverse outcomes for physically abused children. Cluster analysis, an approach to detecting homogeneous groups within a sample of diverse individuals (Magnusson, 1998), was employed. This approach has been used to identify subtypes within samples of abusive parents (Oldershaw, Walters, & Hall, 1989), sexually abused children (Trickett, Noll, Reiffman, & Putnam, 2001), and child witnesses of domestic violence (Grych, Jouriles, Swank, McDonald, & Norwood, 2000), but there are no published accounts of cluster analysis used to derive subtypes of physically abused children.

Although there are no published reports of efforts to subtype physically abused children, some investigators have applied cluster analysis to samples of child witnesses of domestic violence. Those studies illustrate the utility of using cluster analysis to examine heterogeneity in adjustment of children exposed to family violence because characteristics of physically abused children and child witnesses of domestic violence show remarkably similar patterns of social cognitive and behavioral functioning, including risk for externalizing problems (Hughes, Parkinson, & Vargo, 1989; Marcus, Lindahl, & Malik, 2001; Wolfe, Jaffe, Wilson, & Zak, 1985). Hughes and Luke (1998) clustered a sample of 58 children exposed to domestic violence. Support was obtained for five clusters based on parent- and self-report of child behavior problems, internalizing distress, and self-esteem. One cluster was comprised of children who had few behavior problems, no anxiety, and high self-esteem. Another was comprised of children who were considered to be “hanging in there,” with low levels of problematic behavior and average self-esteem. The final three clusters were comprised of children who experienced high levels of behavior problems and/or high distress. Individual differences in internalizing and externalizing problems of witnesses of domestic violence also were examined by Grych et al. (2000). One cluster was comprised of children who experienced no behavioral or emotional problems. A second cluster consisted of children with low levels of behavior problems and only minor internalizing problems. The other three clusters included children with high internalizing and/or externalizing problems. In summary, both studies indicated there were five clusters, with one characterized by positive functioning, one with low levels of maladjustment, and other clusters demonstrating various patterns of adjustment problems.

To determine the degree to which clusters are clinically meaningful, investigators typically compare clusters on factors on which the clusters should differ if the distinction between them is relevant to clinical practice. In the current study, each cluster of physically abused children was compared to a sample of nonabused children closely matched to the abuse sample. In addition, clusters were compared to each other on several protective and vulnerability factors known to predict individual differences or resilient functioning among physically abused children. Using a multilevel approach, we considered characteristics of the individual child and features of the caregiving environment associated with positive adjustment. At the individual child level we measured social information processing operations, and characteristics of parenting included level of sensitivity and harshness of discipline.

According to the highly influential social information processing (SIP) framework (Crick & Dodge, 1994), the manner in which children receive and process social cues available during interactions with peers impacts their responses in peer interactions. A large body of literature documents a strong link between SIP and children's social behavior (Crick & Dodge, 1994). Of interest in this study were attributions for peer intentions and problem-solving skills in problematic peer situations. Those SIP operations were included due to their strong association with social adjustment. That is, children who are aggressive and who are rejected by their peers are less accurate in interpreting the intentions of peers and are more likely to attribute hostile intent to neutral peer behavior (de Castro, Veerman, Koops, Bosch, & Monshouwer, 2002). Further, socially competent children generate more effective, socially appropriate solutions to peer-related problems than do aggressive children (Lochman, Meyer, Rabiner, & White, 1991). In terms of the social information processing abilities of maltreated children (i.e., physically abused and/or neglected children), research shows that they tend to have a hostile attributional bias for peers’ intentions (Price & Glad, 2003) and such attributions predict externalizing behavior and conduct problems of maltreated children (Price & Landsverk, 1998). Deficits in problem-solving skills also have been found among maltreated children (Smith & Walden, 1999; Trickett, 1993), and generation of solutions to peer conflicts predicts maltreated children's social competence (Price & Landsverk, 1998).

Extant research shows parents who are warm and responsive tend to have children who engage in more prosocial behavior (Whiteside-Mansell, Bradley, Tresch Owen, Randolph, & Cauce, 2003) and less externalizing behavior compared to children of parents lacking in warmth (Booth, Rose-Krasnor, McKinnon, & Rubin, 1994). Further, parental sensitivity can serve a protective function for children exposed to a variety of developmental risks (Pianta, Egeland, & Sroufe, 1990). In contrast, harsh discipline is associated with negative outcomes, including aggression (Strassberg, Dodge, Pettit, & Bates, 1994) and low rates of child prosocial behaviors such as helping and comforting others (Hastings, Rubin, & DeRose, 2005). Many physically abusive parents are characterized by harsh discipline and low parental sensitivity so one might assume that those parenting dimensions would be stable across abused children. A few studies, however, indicate parental warmth and discipline predict individual differences within samples of maltreated children (Farber & Egeland, 1987; Lansford et al., 2006). Thus, even for children who experience abusive parenting, there appears to be sufficient variability in sensitivity and harshness of discipline to predict differential outcomes.

The abused child participants in this study were divided into groups, using cluster analysis, on the basis of similarities and differences in their social behavior (as observed on the playground and reported by teachers). Cluster analysis is an exploratory procedure and there are no extant studies on classification of physically abused children by which to predict the number of clusters that would emerge. Based on prior efforts to subtype children exposed to domestic violence, however, we expected that there would be a cluster of children characterized by relatively positive social adjustment and at least one that would show poor social adjustment on most indices. To evaluate the validity of the clustering process, clusters of abused children were compared to nonabused children and to one another on social information processing operations and the parenting context. We predicted that clusters would vary in the degree to which they differed from nonabused children, such that clusters comprised of relatively well adjusted children would be more similar in social adjustment to nonabused children. It was expected that children in the cluster characterized by positive social adjustment would have greater breadth and higher quality of solutions to peer problems and less hostile attributional bias. Further, they were expected to experience greater parental sensitivity and less harsh discipline.