|
Abstract
|
A naturalistic study of violent behavior in a sample of 100 boys residing within an inpatient psychiatric facility is reported. Over 1000 recorded incidents of interpersonal aggression were used to explore patterns of male violence in a contextually defined setting. Findings include a proportionately greater incidence of violence in younger boys. Staff members, particularly males, were more likely to be the victims of interpersonal violence, but peers and staff were equally likely to be the target of aggressive behavior in general. Observable provocation in the social context, the age of the child, and choice of victim (staff vs. peer) were found to be important predictors within a logistic regression model of violence. An estimate of the predictive utility of a logistic model for the occurrence of violence was found to be too low for clinical purposes. An application of a descriptive statistical method to collections of incidents for five highly aggressive boys yielded idiopathic models of violent behavior. Predictive accuracy estimates from these models suggest that multivariate description of behavioral data can be a clinically useful tool in understanding and possibly preventing violence within defined contexts such as residential psychiatric facilities.
VioLit summary:
OBJECTIVE:
The aim of this study by Garrison was to examine the levels of predictive power in the prediction of violent behavior of boys within an inpatient psychiatric treatment facility.
METHODOLOGY:
The author employed a quasi-experimental multiple cross-sectional design, collecting data upon 1,038 aggressive incidents that occurred over a two year period between 100 male children and adolescents, between the ages of 7 and 15, residing within a psychiatric institution in the northeastern United States. Participant-observer reports upon aggressive behavior were collected during five 5-hour sessions a week in a psychoeducational unit, and they included information concerning time, location, antecedents and form of aggression, as well as data about victim, objects used, consequences of aggression and staff observations about external provocation. Observations were conducted at random times over a twelve-month period, with moderate inter-rater reliability. Violence was defined by the author as intense physical attacks, whilst threats, brief physical contact, play and aggressive verbal behaviors were not considered to be violence. Data analysis included examination of frequencies, Chi-Square analysis, functional category analysis and logistic regression analysis. Five highly aggressive children were chosen as subjects for logistic analysis in order to determine the usefulness of these statistical methods for individual cases.
FINDINGS/DISCUSSION:
The author initially examined frequency distributions, and found that violent behaviors were recorded in nearly half of all the incidents, with older boys accounting for 14% more of the aggression. Staff members were more often targets of aggression than were peers, with males also more likely to be victims. External provocation was found to be present in 64% of the incidents, and involved demands related to social circumstances, attacks from peers or desires of the subject. If no external provocation was found, the child was often anxious or irritated, or often withdrawn. Most of the incidents occurred in the classroom, during the morning hours. In an examination of bivariate relationships, the author found that whilst more aggression was found in the older group, younger boys accounted for a proportionately greater number of violent incidents. Violence was not more likely to occur at any particular time or location. Only one higher-order interaction was found, with male staff members being the most likely victim of violence. Logistic regression with violence as the possible outcome found three significant variables: antecedent (if provocation was present), age group (11 and under or 12 and above) and status of victim (staff or peer). Whilst the overall model with these three variables was highly significant, the goodness-of-fit measure was low, and overall predictive power of the model was only slightly better than chance. The model also exhibited moderate levels of false positives and false negatives. Predictive utility was found to be much better for the five individual cases, with lower false classifications than for the whole sample. The author suggested that unmeasured variables might have led to the low predictive ability for the entire sample, and he concluded that this methodology might be more useful for predicting violence in individual cases than for aggregated data.
AUTHOR'S RECOMMENDATIONS:
The author suggested that attention must be paid to specific environmental factors, as well as to internal sources of aggression. Developmental changes in, and functional properties of, aggressive behavior in various samples of children should be considered, in order to determine the role of normal development and psychopathological conditions. The author concluded that prediction could be maximized only with future research that is aimed at identifying a homogeneous group of violent individuals, and only via use of multiple theories of aggression.
EVALUATION:
The author presents an informative examination of the predictive ability of a number of models. However, the measures seem to be less than optimal. The definition of a violent act seems to be somewhat vague, and other environmental and personal variables could have been included in the analysis. The use of participant observation has its own problems, as subjective ratings of the subjects' behaviors might lead to results that are not easily generalized and are not seen as thoroughly solid. A more thorough discussion of the rating of violence would have been appreciated, as would have a discussion of the implications of the findings for future treatment, intervention and prevention planning. Despite these limitations, the author presents a study that provides an excellent foundation upon which to base further research into the complicated and confusing area of prediction of violence. (CSPV Abstract - Copyright © 1992-2007 by the Center for the Study and Prevention of Violence, Institute of Behavioral Science, Regents of the University of Colorado)
KW - Juvenile Violence
KW - Juvenile Male
KW - Juvenile Patient
KW - Juvenile Offender
KW - Child Violence
KW - Child Patient
KW - Child Male
KW - Child Offender
KW - Male Offender
KW - Male Violence
KW - Patient Violence
KW - Middle Childhood
KW - Late Childhood
KW - Early Adolescence
KW - Violence Prediction
KW - Mental Health Institution
KW - Mental Health Patient |