Practice Profile: Psychological Treatments for Adults With Histories of Violent Offending

Papalia and colleagues (2020) conducted a meta-analysis to examine whether psychological treatments delivered in correctional or inpatient mental health settings to adults with histories of violent offending were related to changes measured at intermediate posttreatment of outcomes, including trait anger, impulsivity, social problem solving, antisocial cognitions, and general social skills.

The literature review process followed the PRISMA (Preferred Reporting Items for Systemic Reviews and Meta-Analysis) guidelines. A variety of resources were searched from January 1990 through November 2017. Electronic databases that were searched included Cochrane Central Register of Controlled Trials, PsycINFO, ISI Web of Science, MEDLINE, Criminal Justice Abstracts, ProQuest, EBSCOhost, and CINCH Australian Criminology Database. Google, Google Scholar, and relevant government websites (e.g., Australian Institute of Criminology, Forum for Corrections Research) were searched to identify relevant gray literature and unpublished research. To gather additional records, hand searches were also conducted of existing literature review and key journals.

All studies were required to be available in English and have been published between 1990 and 2017. The population of interest was adults with a history of violent offending, defined by law enforcement records or self- or informant-reports. All types of structured psychological treatments were eligible for inclusion. Psychological treatments were defined as “talking therapies” derived from psychological principles that intervened in the domains of thoughts, feelings, and/or behaviors. The aims of the treatments must have been broadly stated to reduce violent, aggressive, or otherwise antisocial behavior. Studies conducted in correctional settings and forensic mental health settings were eligible for inclusion. Samples were eligible for inclusion if either 1) at least 80 percent of the sample comprised people with a history of violent offending or 2) fewer than 80 percent of the sample comprised individuals with past violence but study outcomes were reported separately for the violent subsample.

All definitions of violence were accepted. However, studies with a sole focus on sexual or domestic violence were excluded. Studies evaluating violence treatments for adults with intellectual and/or learning disabilities were excluded. Studies were eligible for inclusion if the psychological treatment was contrasted with at least one control condition. Studies were excluded if they used only a treatment attrition/dropout group as the comparison group. Both randomized controlled trials and quasi-experimental designed studies were included in the meta-analysis. Nonrandomized designed studies were included if they matched the treatment and comparison groups on a pretreatment risk variable (e.g., prior offending history, risk of reconviction score; showed no significant differences between the treatment and comparison groups on a pretreatment measure of at least one offending behavior outcome variable [e.g., mean number of aggressive incidents in the previous 6 months]) or demonstrated another type of pretreatment comparability of the treatment and control groups (e.g., comparability on index sentence length, scores on treatment target measures).

Twenty-two independent studies were included in the review, with 18 reporting sufficient data for meta-analyses. Studies were conducted in the United Kingdom (59.1 percent), the United States (18.2 percent), Australia (13.6 percent), Canada (4.5 percent), and the Netherlands (4.5 percent). Treatments were performed in prison settings (50.0 percent), in an inpatient forensic mental health environment (45.5 percent), and in a community setting (4.5 percent). Treatment types included several versions of the reasoning and rehabilitation programs (31.8 percent), anger management programs (27.3 percent), other cognitive–behavioral–based treatments (18.2 percent), violence reduction programs (13.6 percent), and third-wave cognitive–behavioral therapies (i.e., dialectical behavior therapy, schema-focused therapy; 9.1 percent). The average duration of treatment was 54.3 hours, with a range between 16 to 300 hours. The combined sample size of the included studies was 1,969 individuals. Most studies examined male-only samples (86.4 percent). The mean age of the included samples was 34.2 years, with a range between 24 and 42 years. Studies varied in when they measured outcomes posttreatment. It was unspecified in 12 studies, reported as commencing directly following treatment in 5 studies, within 2 weeks of the end of treatment in 3 studies, during and posttreatment measurements combined in 1 study, and combined measurements taken during treatment, posttreatment, and during follow-up in 1 study. Half of the studies (n = 9) reported follow-up measurements taken after posttreatment testing, with follow-up times ranging from 1 to 12 months.

Studies were required to measure and report at least one quantitative outcome variable of interest for the treatment and control groups. Primary outcomes were behavioral acts (e.g., criminal recidivism), whereas secondary outcomes were psychological constructs (e.g., anger, impulsivity). For this meta-analysis, outcomes were aggregated into categories with similar measures. The trait anger measure is defined as a long-term disposition to experiencing anger as a general tendency or in response to provocation. Impulsivity was defined as a tendency to engage in behavior with little or no planning or forethought. Social problem solving was defined as the extent to which an individual attempts to identify or engage in adaptive coping response for everyday social problems. Antisocial cognitions were defined as cognitions that encourage criminal or harmful behavior (i.e., criminal attitude, violent attitude, and hostility). General social skills were defined as the absence of problems in social functioning.

Effect size was measured using the standardized mean change for pretest–posttest control designs and was calculated by obtaining mean scores, standard deviations, and the number of participants for treatment and comparison groups at both pretreatment and posttreatment stages for each study. Individual effect sizes were pooled across studies using the inverse variance random-effects model.