Systematic review and meta-analysis of interventions relevant for young offenders with mood disorders, anxiety disorders, or self-harm

,


Introduction
Numerous studies have demonstrated that young offenders experience high levels of mental health problems (e.g.Vermeiren, 2003;Chitsabesan et al, 2006).Depression, anxiety disorders, and self-harm occur particularly frequently in this population (Vermeiren 2003;Callaghan et al, 2003;Stallard et al, 2003;Ahrens and Rexford 2002;Dixon et al, 2004;Sukhodolsky and Ruchkin 2006;Carswell et al, 2004) and with high levels of comorbidity (Axelson and Birmaher 2001).Moreover, there is considerable evidence that young offenders, both incarcerated and in the community, do not receive the referrals and interventions for mood and anxiety disorders, and self-harm that they need (Gunn et al, 1991;Stallard et al, 2003;Callaghan et al, 2003;Carswell et al, 2004;Chitsabesan et al, 2006), and that such problems predict offending status -the more psychiatric diagnoses a youngster has, the more likely they are to be an offender (Dixon et al, 2004).
In order to identify which interventions should be recommended for young offenders experiencing problems with mood or anxiety disorders, or self harm, we have conducted a systematic review of trials of interventions using Cochrane Collaboration methods (Higgins and Green, 2008).A systematic review is a scientific and robust methodology which aims to provide a synthesis of data which is as free from bias as possible (Khan et al., 2003).In particular, it is important that clinical and policy related decisions are made using the highest quality evidence and the systematic review is seen as the gold standard in evidence-based research (Egger et al., 2001).Moreover, evidence-based research and practice helps to ensure that public money is not wasted on interventions which have exaggerated effect sizes due to biases (Schulz et al., 1995), or on interventions which may do more harm than good.This is particularly crucial in the domain of mental health, especially when working with a vulnerable group like young offenders.It is crucial that decision-making in relation to health services and treatments for this group are based on current best evidence.
The main objective of the review was to determine what interventions are relevant to, and effective in, alleviating the symptoms and behaviours associated with mood and anxiety disorders and self-harm.Where possible we wished to determine whether specific interventions are effective for different types of disorder.

METHOD Search strategy
We searched the following electronic databases (until August 2007) using Ovid Online:

Inclusion criteria
Studies were eligible for inclusion in the study if they (i) examined interventions relevant to the treatment of mood or anxiety disorders, or self-harm, in young offenders; (ii) included participants with a mean age of 19 years or under (which is in line with the current structure of the UK Youth Justice system (Youth Justice Board, 2006); (iii) included a specific mental health assessment which measured suicidality, anxiety symptoms or depressive symptoms; (iv) were randomised controlled trials or systematic reviews of randomised controlled trials.Where studies also included non-offenders we included those trials where 75% of the sample was young offenders.

Selection of studies and data extraction
Two authors independently screened potential studies for inclusion in the review.Data extraction for key outcome variables was completed separately by two authors and any disagreements were resolved by consensus and, where needed, a third reviewer's decision was sought.Data extraction forms were standardised and included the information on baseline, post-treatment and follow-up outcome measures (means, standard deviations, and proportions), participant information and information about interventions and the setting of the study (see Tables 1 and 2 for full details of information extracted).

Assessment of the quality of studies
This was carried out independently by two reviewers (blind to authorship and publication information) using a standardised form.The form used for randomised controlled trials was based on the criteria described in the Cochrane Collaboration Handbook (Higgins and Green, 2008), which focus on the relationship between concealment of allocation to treatment, and the strength and direction of the treatment effect (Shulz et al, 1995).We also examined blinding of assessors and withdrawals after randomisation.

RESULTS
We identified 708 papers for potential inclusion in this review.Of these, 10 studies reported in 12 papers were evaluated as suitable for inclusion in this review.The flow of trials in this review is shown in the Quorum Diagram in Figure 2.This diagram describes how trials identified in the searches are handled through the review process with the aim of using review methods that are transparent and reproducible (Moher et al., 1999).
Figure 2 about here.

Interventions
A range of interventions were examined in the studies included in this review.Some studies focussed on young offenders with a diagnosis of a particular mental health problem (e.g.PTSD), whilst others did not impose diagnostic inclusion criteria.Brief details about the participants, settings and interventions of each of the studies included in this review are shown in Tables 1 and 2. In this review we have included outcomes relating to each of the mental health problems under investigation.It should be noted that a range of other outcomes (e.g.recidivism) were also reported in trials which are not included here as our focus in this review is on mental health outcomes.
Tables 1 and 2

Quality assessment
A key aspect of trial quality is the adequacy of the allocation of participants to treatment groups.Allocation concealment ensures that a researcher cannot influence which arm of a trial a given participant is allocated to.An example of adequate concealment would be the use of an off-site trial unit which generates the allocation sequence independent of the research team.Where a trial is found to have used inadequate allocation concealment (e.g.alternation) effect sizes are found to be considerably inflated (Schulz et al., 1995).We used Cochrane Collaboration criteria to evaluate the allocation concealment methods used in the studies included in this review (Higgins and Green, 2008).In two trials additional pre-existing control groups were reported as a second control group (Taylor et al, 1967;Rohde et al, 2004a), but because participants were not randomised to these groups they were excluded from the review.We contacted the authors of trials where allocation concealment was unclear in the report of the trial.In one case this changed the allocation score from 'unclear' to 'adequate' (Rohde et al, 2004b) (randomisation had been carried out by a data analyst who had no contact with any of the participants in the trials).However, in one case (Biggam and Power 2002) the further information provided did not clarify the adequacy of allocation concealment.
The assessment of quality of concealment of allocation resulted in four trials (40%) being rated as 'A' quality (the highest rating -for adequate concealment), five trials (50%) being rated as B (for unclear concealment), and one trial (10%) being rated as C (inadequate concealment).
The quality scores in terms of allocation concealment, blinding of assessors, withdrawals after randomization and length of follow-up period may be viewed in Table 3 (summarized as per Khan et al, 2003).

1) Effectiveness in individual studies.
Here we report the results of the effectiveness of each trial in terms of the descriptive statistics and significance levels as reported in the original trials (see Table 4).As is standard practice in systematic reviews we selected the longest follow-up point from each trial (Higgins and Green, 2008).In the eight trials in which depression was measured as an outcome, significant improvements in depressive symptoms were seen in those receiving experimental interventions as compared to control groups in five trials (see Table 4).In the six trials that measured anxiety outcomes, significant improvements in anxiety symptoms were seen in those randomised to experimental interventions as compared to those in control groups in three trials (see Table 4).Only one trial measured self harm as an outcome and reported no significant reduction in the experimental group as compared to a control group (Rohde et al., 2004a).
It is interesting to note that in the Rohde et al., (2004b) study some significant differences were found between treatment groups immediately post treatment, leading the authors to conclude that their group-based CBT intervention is a useful acute treatment for the youth offending population.
Table 4 about here 2) Data aggregation and meta-analysis.We aggregated data from three trials included here with respect to the outcome of depression as they had each examined a groupbased CBT intervention.Ahrens and Rexford (2002) (Beck et al, 1996) and one used the Hospital Anxiety and Depression Scale (HADS) (Zigmond and Snaith, 1983).Previous meta-analyses have demonstrated that it is feasible to combine these measures using Standardised Mean Difference (SMD) calculations (Townsend et al, 2001).The SMD is a measure of effect size for continuous data and is used when trials have used different scales to measure the same outcome (Khan et al., 2003).Figure 3 shows the outcome of this analysis, which demonstrates that outcomes for depression were improved in participants who received group-based CBT as compared to those in the control groups [SMD = 0.38 (-0.69, -0.07)].
Figure 3 about here Insufficient data were available to permit full intention-to-treat analyses of outcome data.We were, however, able to obtain sufficient information to permit an available case analysis (Higgins and Green, (2008).Available case analyses: "Include data on only those whose results are known, using as a denominator the total number of people who had data recorded for the particular outcome in question" (Higgins and Green, 2008, Chapter 16, section 2.2).We have included proportion of participants in each study who did not provide outcome data in Table 2 where this information was provided.

Assessment of heterogeneity
Examination of the forest plot (which shows measure of effect sizes and confidence intervals for individual studies and data aggregated via meta-analysis (Lewis and Clarke, 2001)) in Figure 3 reveals a degree of heterogeneity with respect to the effectiveness for group CBT interventions for young offenders with any type of mental health problem.There are a number of potential sources of heterogeneity.First, the studies focussed on young offenders with differing mental health problems (MDD, PTSD and DSH).Secondly, there were differences in quality of the studies in terms of the adequacy of allocation concealment.Only one of the studies was found to have adequately concealed the allocation to groups (Rohde et al, 2004b).The heterogeneity (I 2 ) for this meta-analysis (which is a measure that quantifies inconsistency) was substantial and significant (see Figure 3).Hence, we conducted a random effects analysis (which assumes that included studies estimate intervention effects which are different, but related in some important way) (Higgins and Green, 2008).This type of analysis is commonly used to explore any observed heterogeneity (Higgins and Green, (2008), and revealed a non-significant overall effect for group-based CBT [SMD= -0.69 (-1.18, 0.37)].

Types of studies and interventions
A thorough search of the worldwide literature revealed that very few high quality studies (RCTs) have been conducted to evaluate the effectiveness of interventions relevant to young offenders experiencing problems with mood and anxiety disorders, or self-harm.The lack of evidence for interventions found here for young offenders has been noted by other researchers (e.g.Sukhodolsky and Ruchkin 2006) and reflects an overall paucity of well-designed trials for children and young people with mental health problems (Shapiro et al, 2006).The lack of randomised studies to evaluate the effectiveness of interventions in the criminal justice system in general has been highlighted by Farrington (2003), who argued that individual differences in policy makers is, rather worryingly, the key factor in determining whether RCT-type studies are promoted and funded.
A variety of interventions have been tried which are relevant to young offenders with mood or anxiety disorders, or problems with self harm.Half of the trials identified in this review used a cognitive behavioural approach to intervene with these problems.
Three of the 10 studies we identified focussed on young offenders who met criteria for one or two specific disorders (PTSD, co-morbid MDD and CD, and self-harm).Seventy percent of the studies recruited participants from the general juvenile offending population (regardless of their mental health status), but measured mental health outcomes that the intervention had targeted.Eighty percent of the trials had been carried out with incarcerated male young offenders only.This is significant because the vast majority of young offenders are supported in the community, including a large proportion of those who were previously incarcerated.Moreover, it appears that, although incarcerated young offenders and those in the community have similar mental health needs, those of community-based offenders are not being met to the same degree as those in custodial settings (Chitsabesan et al, 2006).Just one study included community-based male and female young offenders (Rohde et al, 2004b) and two further studies had examined incarcerated female offenders (Taylor, 1967;Reardon, 1977).

Effectiveness of interventions
Individual trials reported varying degrees of success in terms of the effectiveness of experimental interventions decreasing depression, anxiety or self harm.We aggregated data from three of the trials indentified here with respect to the outcome depression, as each trial had examined a group-based CBT intervention with young offenders.The results of this analysis revealed that depressive symptoms were significantly reduced in young offenders receiving a cognitive-behavioural intervention, compared to those receiving 'usual care' or a 'no treatment control'.This is a promising finding and is especially important when viewed in the context of the findings of Barrett et al, (2006), who demonstrated that depressed mood in young offenders predicts higher costs in terms of health and educational services used, and Dixon et al, (2004), who showed that having multiple psychiatric diagnoses increases the risk of offending behaviour.
However, it should be noted that a random effects analysis of the depression data did not yield a significant pooled effect.Hence, the results of this meta-analysis should be interpreted with caution given the significant heterogeneity observed.If heterogeneity was not a problem in the review the results of the fixed and random effects analyses would be identical.Nonetheless, the heterogeneity arising in the metaanalysis was to be expected given that the three trials combined here included participants who presented with a range of problems (PTSD, depression and self harm), were dealt with in various settings, had different follow-up periods and differed in the adequacy of allocation concealment.Despite these problems we feel that the data aggregation presented is useful clinically, given the paucity of data in the field, and the problems with co-morbidity of mental health problems in the youth offending population.
More research is needed on treatment effectiveness and efficacy to bolster the evidence base, but these preliminary data suggest that group-based CBT interventions may be helpful for young people in the criminal justice system with mental health problems like mood and anxiety disorders, and self harm.As Desai et al, (2006) noted, CBT-based interventions are extremely well suited to application in the criminal justice system -especially in secure settings.CBT is short-term, time-limited and focused on current problems, which is ideal for settings where length of stay is short and outcomes need to be as effective as possible.CBT provides clients with a new set of skills and is both collaborative and empowering.Many CBT interventions are manualised, which make them relatively easy to teach to clinicians from a variety of backgrounds.Indeed, CBT interventions which focus on problem behaviours and reducing recidivism have been extensively researched within criminal justice settings (Lipsey et al, 2000).Sukhodolsky and Ruchkin (2006) note that CBT for 'internalising disorders' such as mood and anxiety disorders is likely to share techniques with CBT used to tackle aggression and delinquency.This, they suggest, could be a benefit and lead to the development and refinement of interventions for young people with multiple problems.
On the other hand, they point out that CBT for internalising disorders may suffer from the limitations of treatments for delinquent behaviour, such as possible negative effects of group-based treatments (Dishion et al, 1999).It is noteworthy that most of the CBTbased studies discussed here used a group-based format to deliver treatment to young offenders.More research is required relating to the effectiveness of group-based CBT treatments as compared to individual CBT treatments in this population.Generally it is acknowledged (in the non-offending literature) that CBT is effective for depressed adolescents (Reinecke et al, 1998), and for adolescents with anxiety disorders (Soler and Weatherall 2005).
In non-offending adolescent populations it has been shown that there is no difference in effectiveness of CBT across different delivery formats (e.g.individual, family or group) (James et al., 2005).Moreover, the studies included in the James et al., (2005) systematic review reported significant positive outcomes when measured at longer follow-up periods than in the present review.
It is possible that CBT type interventions for young offenders may not yield the longer term effects seen in other populations and this is likely to be due to the intense emotional difficulties, and co-morbidity of mental disorders evident in this population.
However, one study (Rohde et al 2004b) highlights the possibility that CBT-based interventions are a useful 'acute' treatment in this group given that outcome measures differed significantly between the experimental and control groups immediately posttreatment.

Methodological issues
Unfortunately, most of the trials included in this review suffer from methodological weaknesses, such as a lack of information about randomisation and blinding procedures, incomplete information about drop-outs, and short follow-up periods.This is unfortunate given that these factors are known to bias the outcomes of trials (Schulz et al, 1995) and that there have been repeated calls in the literature for researchers to report studies in a transparent manner (Moher et al, 2001).
There were also problems with follow-up periods, with most of the trials reporting outcomes for very short follow-up periods (most trials reported post-treatment data only).This could mean that the long-term effectiveness of these interventions is not clear.Given the small numbers of participants in all of the studies reviewed here it is highly likely that they were underpowered (ie.they did not recruit significant numbers of participants to yield adequately powered statistical tests).In future researchers should conduct power analyses to determine how many participants should be recruited to a trial in order to ensure adequate power for statistical analyses.
Finally, the RCTs included in this review used a wide variety of comparison interventions including no treatment controls, usual care, or another type of intervention.Ideally, experimental interventions in this population should be tested against a control group involving 'usual care' which is fully described.

Implications for future research
The trials included in this review demonstrate is that it is possible to carry out RCTs for mental health problems with young offenders in both incarcerated and non-incarcerated settings.Large (adequately powered), long-term multi-centre RCTs are now urgently required in both incarcerated and community-based settings.There is a particular need to evaluate the impact of interventions in non-incarcerated young offenders.In the UK the vast majority of young offenders are dealt with on community-based orders and incarceration is very much viewed as a last resort.Moreover, the evidence concerning the effectiveness of interventions specifically for female young offenders is especially sparse.This is of concern because prevalence studies estimate female young offenders to experience more mental health problems than males (Jasper et al, 1998).
There are a number of pragmatic issues that researchers undertaking RCTs with young offenders should be aware of.Firstly, the engagement of these young people can be a considerable challenge (which may explain why 80% of studies were solely based in secure units).Secondly, this is a population with considerable heterogeneity in terms of diagnostic comorbidity, offending profiles, and related social and developmental characteristics.Thirdly, it may be that the 'dose' of treatment required for this population is greater than in other groups of individuals.Hence, trialists should consider implementing longer treatment durations for this group as effect sizes may diminish significantly after six months in the case of group-based CBT (Rohde et al 2004b).
Finally, it can be difficult to control for the environmental and judicial interventions that are imposed on young offenders.Hence, there are a variety of confounding factors facing researchers in this area.However, such problems can be overcome and where possible these factors should be controlled for in analyses within trials.
To summarize, the results presented here suggests group-based cognitivebehavioural interventions appear to be useful for young offenders experiencing depression, anxiety disorders or problems with self harm.Future studies should be careful to adopt adequate randomisation procedures, compare experimental treatments to 'usual care' (which should be fully described) and evaluate outcomes (with blind assessment) over a 12-month follow-up period with dropouts fully described.This evidence is crucial given the high unmet needs in relation to these problems in young offenders (Chitsabesan et al, 2006), and that such problems are associated with increased rates of recidivism and suicidality (Vermeiren 2003).
AMED (from 1985), CINAHL (from 1982), EMBASE (from 1980), Ovid Medline ® In-Process and other non-indexed citations and Ovid Medline (from 1950), and PsycINFO (from 1967).Our search strategy included keywords to encompass: (i) specific characteristics of young offenders using person and population-based terms like 'young offender' or institution-based terms like 'prison', (ii) specific types of literature such as Randomized Controlled Trials (RCTs) and systematic reviews and (iii) specific mood and anxiety disorders (eg.depression) and self harm.To implement the search across all disorders and literature types eight separate strategies were executed based on the information in Figure 1.These were (i) Anxiety and RCT, (ii) Anxiety and Systematic review, (iii) PTSD and RCT, (iv) PTSD and Systematic review, (v) Depression and RCT, (vi) Depression and Systematic review, (vii) Self harm and RCT, (viii) Self Harm and Systematic review.The details of the separate strategies are available from the corresponding author.

Figure 1
Figure 1 about here had examined Cognitive Processing Therapy (CPT) with young offenders with PTSD.Biggam and Power (2002) examined group problem solving with vulnerable young offenders, and Rohde et al, (2004b) investigated group Cognitive Behavioural Therapy (CBT) with young offenders with Major Depressive Disorder (MDD) and Conduct Disorder (CD).Two of the trials examined depression with the Beck Depression Inventory (BDI)

Figure
Figure 1.Complete electronic search strategy

Table 3 Summary of results of quality assessments performed in this review Figure 3 Effectiveness of CBT group-based interventions in young offenders (with any disorder) with respect to depression
Test for heterogeneity: Chi² = 20.55,df = 2 (P < 0.0001), I² = 90.3%Test for overall effect: Z = 2.38 (P = 0.02)