Introduction
Attention deficit hyperactivity disorder (ADHD) is one of the most diagnosed mental health problems in children; meta-analyses estimate a world pooled prevalence of 5.3% in children and 2.5% in adults (Polanczyk et al. Reference Polanczyk, de Lima, Horta, Biederman and Rohde2007; Simon et al. Reference Simon, Czobor, Balint, Meszaros and Bitter2009). Characterized by difficulties of inattentiveness, hyperactivity and impulsivity, it is associated with significant problems in personal, social and occupational outcomes (Shaw et al. Reference Shaw, Hodgkins, Caci, Young, Kahle, Woods and Arnold2012). Co-morbid psychiatric problems are common presentations in both children and adults with ADHD (Pliszka, Reference Pliszka1998) and a US nationally representative household survey found that adults with ADHD were five times more likely to develop a mood disorder, four times more likely to develop an anxiety disorder, and three times more likely to develop substance use disorder (Kessler et al. Reference Kessler, Adler, Barkley, Biederman, Conners, Demler, Faraone, Greenhill, Howes, Secnik, Spencer, Ustun, Walters and Zaslavsky2006). Individuals with these co-morbid disorders are likely to experience greater occupational impairment, compared to people with ADHD alone.
Compared with population rates, there is robust evidence to support an over-representation of youths and adults with ADHD in the criminal justice system, most likely reflecting high rates of co-morbidity with conduct disorder (Langley et al. Reference Langley, Fowler, Ford, Thapar, van den Bree, Harold, Owen, O'Donovan and Thapar2010). A meta-analysis of 42 international studies reported that 30% and 26% of the youth and adult prison populations, respectively, had clinically diagnosed ADHD (Young et al. Reference Young, Moss, Sedgwick, Fridman and Hodgkins2014). There were no significant differences for gender and age, which does not parallel the corresponding differences observed in the general population. With respect to age, the implication is that young offenders with ADHD who come into contact with the criminal justice system remain within this system as repeat offenders.
More generally, rates of psychiatric disorders are over-represented among prisoners (Fazel & Seewald, Reference Fazel and Seewald2012) and it seems that the risk for developing co-morbid psychiatric disorders may be greater among offenders with ADHD (Abram et al. Reference Abram, Teplin, McClelland and Dulcan2003; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Young et al. Reference Young, Wells and Gudjonsson2011b ; Gudjonsson et al. Reference Gudjonsson, Wells and Young2012; Gonzalez et al. Reference Gonzalez, Gudjonsson, Wells and Young2013, Reference Gonzalez, Velez-Pastrana, Ruiz Varcarcel, Levin and Albizu-Garcia2015). The confounding effect of this high level of psychiatric co-morbidity is very likely to influence the behaviour, management and treatment of these individuals as, despite international guidelines (Seixas et al. Reference Seixas, Weiss and Muller2012) and treatment benefits on ADHD symptoms (Shaw et al. Reference Shaw, Hodgkins, Caci, Young, Kahle, Woods and Arnold2012), ADHD is rarely diagnosed and treated by offender mental health teams, as concluded by a multi-agency consensus on ADHD and offender management (Young et al. Reference Young, Adamou, Bolea, Gudjonsson, Muller, Pitts, Thome and Asherson2011c ). This problem needs to be addressed, especially since outcomes of treatment for ADHD may be highly effective, with both individual (Ginsberg et al. Reference Ginsberg, Hirvikoski, Grann and Lindefors2012) and societal gains (Lichtenstein et al. Reference Lichtenstein, Halldner, Zetterqvist, Sjolander, Serlachius, Fazel, Langstrom and Larsson2012).
It is therefore important to better estimate the risk of the development of co-morbid disorders (including type of disorder) among offenders with ADHD to inform healthcare practitioners, who in turn can deliver more appropriate treatment, management and care plans. Hence this study synthesized data obtained from a systematic literature search performed to ascertain the psychiatric co-morbid conditions associated with a clinical diagnosis of ADHD in both youth and adult prisoners using meta-analysis. Rates were compared with prisoners without ADHD.
Method
Eligibility criteria
The systematic review was performed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA guidelines) (Liberati et al. Reference Liberati, Altman, Tetzlaff, Mulrow, Gotzsche, Ioannidis, Clarke, Devereaux, Kleijnen and Moher2009). Initial searches were carried out in a variety of databases and websites to gain an understanding of the amount of information available. Reports published since 1980 and in English were included.
Data sources
Searches were conducted in OvidSP Medline (1948 to present) and EMBASE (1988 to present segment), Datastar PsycINFO (unrestricted) and Social SciSearch (1972 to date; limited to English and added since 1 January 1980), including the literature published between 1 January 1980 and 3 May 2011; the search was further extended to 31 May 2014 on 25 June 2014. One additional study (Konstenius et al. Reference Konstenius, Larsson, Lundholm, Philips, van de Glind, Jayaram-Lindstrom and Franck2015) was added following reviewer feedback.
Search
Search terms were developed, refined and tested for relevance by cross-checking results against a list of known relevant articles provided by the lead author. The following descriptors were utilized in OvidSP EMBASE (1988 to present segment): ADHD; attention deficit disorder; [EMTREE] crime; criminals; criminology; criminal behaviour; criminal justice; criminal law; court; criminal psychology; delinquency; juvenile delinquency; gang; legal evidence; legal procedure; police; legal liability; mandatory programs; violence; prisons; prisoner; probation; law enforcement; recidivism; jurisprudence; punishment; offender; drug abuse; drug misuse. Studies were examined to identify those containing information about co-morbid psychiatric disorders in an incarcerated ADHD population.
Study selection
Articles obtained from the final searches were first de-duplicated, then an inclusion/exclusion screening process was undertaken based on the following exclusion criteria: non-English-language articles, articles published before 1980, animal studies, articles that were not peer reviewed (e.g. dissertations), and articles that obviously did not hold relevance (e.g. they did not focus on ADHD or crime and/or contained information about co-morbid psychiatric disorders).
Review articles pre-2006 were excluded, and post-2006 reviews were kept with the sole purpose to examine bibliographies to check for any other articles not identified in the search; these review articles were not included in the final prevalence calculation; only primary research articles were included. Articles with no abstract (including initial PsycINFO and Social SciSearch search outputs) were also excluded unless the title or other information (e.g. key terms) suggested they may hold relevance. The inclusion/exclusion review was first completed based on title/abstract/key words by two researchers, and if the relevance of an article was unclear, the full text was retrieved before a final decision was made. Once the initial inclusion/exclusion process was completed, the full texts of included articles were retrieved for detailed evaluation against eligibility criteria:
-
• Studies must use samples composed entirely of incarcerated individuals.
-
• Studies must define a clear ADHD group, i.e. give the prevalence of their sample who met diagnostic/screening criteria for ADHD using a validated measure. Studies in adults which reported only retrospective self-report of childhood symptoms were not included due to the likelihood that the rates would be overinflated (Young et al. Reference Young, Moss, Sedgwick, Fridman and Hodgkins2014).
-
• Studies must give additional prevalence on at least one of the following: conduct disorder, substance use disorder, mood disorder, anxiety disorder, or personality disorder. The studies must report these variables by ADHD group status (or provide information so that this can be deduced) so the characteristics can be compared between an ADHD group and a non-ADHD group from within the same sample.
Data collection process
A data extraction sheet was developed in Microsoft Excel. The full text publications were divided randomly among two researchers who performed the data extraction independently. Data were reviewed for consistency and any queries were resolved by discussion among the researchers and the lead author. The lead author also made the final decision whether to include/exclude data by reviewing the identified publications. Some cohorts of prison populations were published more than once (Retz et al. Reference Retz, Retz-Junginger, Hengesch, Schneider, Thome, Pajonk, Salahi-Disfan, Rees, Wender and Rosler2004; Rosler et al. Reference Rosler, Retz, Retz-Junginger, Hengesch, Schneider, Supprian, Schwitzgebel, Pinhard, Dovi-Akue, Wender and Thome2004; Young et al. Reference Young, Gudjonsson, Wells, Asherson, Theobald, Oliver, Scott and Mooney2009, Reference Young, Wells and Gudjonsson2011b; Gudjonsson et al. Reference Gudjonsson, Wells and Young2011, Reference Gudjonsson, Wells and Young2012). To avoid double counting data, multiple reports of the same cohort were pieced together by juxtaposing author names, sample sizes and rates of co-morbidity.
Data items
For each included study, the following variables were extracted:
-
(1) Number of people with ADHD and the selected co-morbidities, (2) number of people without ADHD but with the selected co-morbidities, (3) number of people with ADHD but without the selected co-morbidities, (4) number of people with neither ADHD nor the co-morbidities, (5) youth/underage and adult sample, (6) sex of sample. In addition to the co-morbidities that are reported on, we attempted to obtain rates of autism spectrum disorder and learning disability, but too few papers reported on these diagnoses.
-
(2) It was noted that there was no clear definition across publications on the age of an adult v. a youth. For the purposes of this meta-analysis, we designated 18 years to be the cut-off point for a youth becoming an adult (i.e. youths were ⩽18 years and adults were >18 years); papers reporting youth data often cited 18 years as the upper limit of the age range of participants. In those papers where an age range was given that spanned this cut-off point (e.g. 15–28 years), the mean (or median if mean was not provided) was used to define whether the study population should be listed as ‘youth’ or ‘adult’. For studies reporting on both genders separately, information was recorded as two separate observations (‘study strata’) linked by the study number.
Quality control
To ascertain the validity of eligible publications, one of the researchers checked and independently reviewed all papers selected for data extraction and interpretation consistency. Disagreements were resolved by reviewing the data source and by discussion between two reviewers.
Statistical analysis
First, pooled random effect meta-analysis estimates of the prevalence of each psychiatric disorder within participants with ADHD were computed. Second, meta-analyses of risk for co-existing disorders among prisoners with ADHD were performed calculating pooled odds ratios (OR) with 95% confidence intervals (CIs).
Considering heterogeneity of ADHD assessment in research, differences in the criminal justice population, and sample characteristics across countries, a random-effects model approach was established a priori. A random-effects approach presupposes that studies are too dissimilar to assume they theoretically come from the same sample. Studies in each domain were pooled using the DerSimonian & Laird (Reference DerSimonian and Laird1986) method for random effects meta-analysis. Heterogeneity is reported by the I 2 index, which describes the percentage of variation across studies that is from actual heterogeneity rather than chance. The I 2 statistic does not entirely depend on the number of studies considered (Higgins & Thompson, Reference Higgins and Thompson2002). Values of 25, 50 and 75 respectively denote low, moderate and high heterogeneity.
We explored the risk for co-morbidity overall and by age group separately for each of the following psychiatric disorders: conduct disorder (CD), substance use disorders (SUD), mood disorders (MD), anxiety disorders (AD) and personality disorders (PD). PD was only reported for adult studies. Subgroup analysis based on gender was not possible due to the low number of studies that resulted by stratifying results for male, female and mixed samples for each co-morbid category. Abram et al. (Reference Abram, Teplin, McClelland and Dulcan2003) and Plattner et al. (Reference Plattner, The, Kraemer, Williams, Bauer, Kindler, Feucht, Friedrich and Steiner2007) reported their data separately by gender. We combined these sources of data by performing a ‘within study’ meta-analysis, therefore obtaining a common standard error for all participants of those studies, which was then used in the overall pooled meta-analyses and meta-regressions.
Meta-regression was used to examine the impact of age (continuous) and gender (male, female, mixed) as covariates on the pooled OR calculated for co-existing disorders. All analyses were performed using Stata v. 13 (StataCorp, USA) metan and metareg commands.
Results
In total, 9410 publications were identified following the OvidSP Medline, EMBASE, Datastar, PsycINFO and Social SciSearch database search. One additional publication was added through personal communication. Publications not specifically relating to ADHD and criminality and duplicates between databases were excluded electronically, leaving 339. Once further duplicates had been identified manually, 325 publications remained for which the full text was retrieved. 281 publications were further excluded as they did not report on incarcerated ADHD samples and/or did not provide associated co-morbid disorders data.
A further 26 studies were then removed for the following reasons: (1) used inadequate methods to assess ADHD status (e.g. studies that only used retrospective symptom assessment, self-report of ‘behavioural problems’ as a child, no consideration of childhood symptoms before adulthood diagnosis given, ADHD and specific developmental disorder considered synonymous; nine studies), (2) provided co-morbid information but only for the whole sample, not specified by ADHD status (six studies), and (3) did not have an adequate control group (i.e. all included participants had ADHD, or the control comparison group was not incarcerated; five studies). Although data on ADHD and CD were provided, we excluded Forehand et al. (Reference Forehand, Wierson, Frame, Kempton and Armistead1991) from this analysis because all their cases had CD (i.e. two 0 count cells). Data from Chang et al. (Reference Chang, Chen and Huang2007) specific to SUD were also removed because all their cases were co-morbid with ADHD.
For those papers that presented the co-morbidity data continuously (e.g. as T scores, mean scores on diagnostic screens, etc.; six authors relating to eight studies), the author was contacted to attempt to gain rates for the co-morbid diagnosis (i.e. number of participants who fulfilled clinical criteria). One author responded with additional data relating to a study conducted in Scotland (Gudjonsson et al. Reference Gudjonsson, Wells and Young2012), which was subsequently included in the meta-analysis. The remaining five studies were excluded as their data could not be meaningfully utilized.
Thus, a total of 18 studies of prisoners with/without ADHD and co-morbid disorders were selected to provide data for the meta-analysis (see Fig. 1 for a flow chart of the manual screening process). These 18 studies included nine adult and nine youth studies, for a combined sample of 1615 participants with ADHD and 3128 without the disorder (See Table 1). For each co-morbid category, pooled effects are presented by the overall sample of studies, followed by subgroup analysis for youth and adult samples.
Flow diagram of the manual screening process for eligible literature inclusion.
Summary of characteristics of studies included in the meta-analyses, by psychiatric disorders and pooled co-morbidity rates, by psychiatric disorders
ADHD, Attention deficit hyperactivity disorder
a Not included in the meta-analytic prevalence rate because of one 0 count cell.
b There were two missing cases on these co-morbidity rates.
c 17 cases excluded as questionable depression.
d Only adult studies included.
Pooled random-effects meta-analysis estimates
Table 2 shows all pooled prevalence rates and 95% CIs of co-morbid disorders. Among youths with ADHD, pooled prevalence rates of CD, SUD, MD, depression and AD were 61%, 70%, 25%, 13% and 21%, respectively. In adults with ADHD, the respective prevalence rates were 29%, 74%, 66%, 55%, 68% and 60% for personality disorders.
Pooled random effects meta-analysis estimates of the prevalence of psychiatric disorders co-existing with ADHD
ADHD, Attention deficit hyperactivity disorder; CI, confidence interval.
a Only adult studies included.
Meta-analyses of risk for co-existing disorders
Conduct disorder
Seven studies (Kaplan & Cornell, Reference Kaplan and Cornell2004; Gordon & Moore, Reference Gordon and Moore2005; Chang et al. Reference Chang, Chen and Huang2007; Stahlberg et al. Reference Stahlberg, Anckarsater and Nilsson2010; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Misch, Collins and Gudjonsson2011a ; Grieger & Hosser, Reference Grieger and Hosser2012) provided data on the association between ADHD and CD (Fig. 2a ). The overall association was not significant (OR 1.16, 95% CI 0.61–2.20). Subgroup analysis by youth and adult studies revealed no significant pooled effects for youth samples, but a significant effect based on two adult studies (OR 2.10, 95% CI 1.19–3.70; Fig. 3a ]. Heterogeneity was high in youth studies and low in adults (I 2 = 77.1 v. 33.9%).
Forest plots with overall odds ratios (OR), effect size (ES), and homogeneity statistics for meta-analyses of six domains of co-morbid disorders.
Forest plots with odds ratios (OR), effect size (ES), and homogeneity statistics for meta-analyses of five domains of co-morbid disorders stratified by age ⩽18 years and adult studies. Only adult studies for personality disorders, included in Fig. 2.
Substance use disorders
Eleven studies (Milin et al. Reference Milin, Halikas, Meller and Morse1991; Abram et al. Reference Abram, Teplin, McClelland and Dulcan2003; Retz et al. Reference Retz, Retz-Junginger, Hengesch, Schneider, Thome, Pajonk, Salahi-Disfan, Rees, Wender and Rosler2004; Gordon & Moore, Reference Gordon and Moore2005; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Rosler et al. Reference Rosler, Retz, Yaqoobi, Burg and Retz-Junginger2009; Stahlberg et al. Reference Stahlberg, Anckarsater and Nilsson2010; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Wells and Gudjonsson2011b ; Grieger & Hosser, Reference Grieger and Hosser2012; Konstenius et al. Reference Konstenius, Larsson, Lundholm, Philips, van de Glind, Jayaram-Lindstrom and Franck2015) provided data on the association between ADHD and SUD (Fig. 2b ). There was a significant overall effect size (OR 2.48, 95% CI 1.30–4.72), with substantial heterogeneity (I 2 = 84.3%). Subgroup analysis indicated a significant pooled effect for adults (OR 2.41, 95% CI 1.22–4.79) but not for youths (OR 2.28, 95% CI 0.73–7.12). Heterogeneity was high (I 2 = 91.9%) in youth studies and moderate (I 2 = 54.8%) in adults (Fig. 3b ).
Mood/affective disorders and depression
Twelve studies (Eyestone & Howell, Reference Eyestone and Howell1994; Abram et al. Reference Abram, Teplin, McClelland and Dulcan2003; Gordon & Moore, Reference Gordon and Moore2005; Anckarsater et al. Reference Anckarsater, Nilsson, Stahlberg, Gustafson, Saury, Rastam and Gillberg2007; Chang et al. Reference Chang, Chen and Huang2007; Plattner et al. Reference Plattner, The, Kraemer, Williams, Bauer, Kindler, Feucht, Friedrich and Steiner2007; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Rosler et al. Reference Rosler, Retz, Yaqoobi, Burg and Retz-Junginger2009; Stahlberg et al. Reference Stahlberg, Anckarsater and Nilsson2010; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Wells and Gudjonsson2011b ; Konstenius et al. Reference Konstenius, Larsson, Lundholm, Philips, van de Glind, Jayaram-Lindstrom and Franck2015) provided data on mood/affective disorders, which included depression, mania and adjustment disorder with associated mood symptoms (Fig. 2c ). The overall effect size was significant (OR 2.96, 95% CI 1.86–4.71), with moderate heterogeneity (I 2 = 64.2%, p = 0.001). The pooled effect was significant for studies in youth (OR 1.89, 95% CI 1.09–3.28) and adults (OR 4.50, 95% CI 2.69–7.51), with low heterogeneity in both (youth I 2 = 48.3%; adults I 2 = 39.0%) (Fig. 3c ).
To partial-out how much of the effect of mood disorders was likely due to depression, we performed a meta-analysis on nine studies (Eyestone & Howell, Reference Eyestone and Howell1994; Gordon & Moore, Reference Gordon and Moore2005; Anckarsater et al. Reference Anckarsater, Nilsson, Stahlberg, Gustafson, Saury, Rastam and Gillberg2007; Chang et al. Reference Chang, Chen and Huang2007; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Stahlberg et al. Reference Stahlberg, Anckarsater and Nilsson2010; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Wells and Gudjonsson2011b ; Konstenius et al. Reference Konstenius, Larsson, Lundholm, Philips, van de Glind, Jayaram-Lindstrom and Franck2015) that provided data specific for depressive disorder (Fig. 2d ). The overall effect size was significant (OR 3.00, 95% CI 1.70–5.29), with moderate heterogeneity (I 2 = 62.9%). The effect of ADHD on depression was not significant in youth participants (OR 1.59, 95% CI 0.61–4.14) but was significant in adults (OR 4.66, 95% CI 2.92–7.45), with moderate and no heterogeneity, respectively (youth I 2 = 57.3%; adults I 2 = 24.5%).
Anxiety disorders
On the seven studies(Abram et al. Reference Abram, Teplin, McClelland and Dulcan2003; Chang et al. Reference Chang, Chen and Huang2007; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Rosler et al. Reference Rosler, Retz, Yaqoobi, Burg and Retz-Junginger2009; Stahlberg et al. Reference Stahlberg, Anckarsater and Nilsson2010; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Wells and Gudjonsson2011b ) with data on associations between ADHD and AD (Fig. 2e ), there was no significant overall effect (OR 1.82, 95% CI 0.78–4.28; I 2 = 79.3%). Stratification revealed no effect in youth studies (OR 0.72, 95% CI 0.13–3.95; I 2 = 84.9%), but was significant among adults (OR 3.58, 95% CI 2.32–5.53), with no heterogeneity (Fig. 3e ).
Personality disorders
Only studies in adults (Black et al. Reference Black, Arndt, Hale and Rogerson2004; Retz et al. Reference Retz, Retz-Junginger, Hengesch, Schneider, Thome, Pajonk, Salahi-Disfan, Rees, Wender and Rosler2004; Einarsson et al. Reference Einarsson, Sigurdsson, Gudjonsson, Newton and Bragason2009; Rosler et al. Reference Rosler, Retz, Yaqoobi, Burg and Retz-Junginger2009; Westmoreland et al. Reference Westmoreland, Gunter, Loveless, Allen, Sieleni and Black2010; Young et al. Reference Young, Wells and Gudjonsson2011b ; Konstenius et al. Reference Konstenius, Larsson, Lundholm, Philips, van de Glind, Jayaram-Lindstrom and Franck2015) were considered for the effect size of the association with PD (Fig. 2f ). Seven studies showed a significant overall effect of ADHD in PD (OR 3.22, 95% CI 2.07–5.01) with low heterogeneity (I 2 = 37.6%).
Meta-regression
To further assess potential sources of heterogeneity, the effect size (OR) of each co-morbid category was regressed on age and gender. Age was significantly (p < 0.05) associated with an increase in the OR between ADHD and mood/affective disorders including depression. Age was not a significant factor on the remaining co-morbid categories. There were no effects of gender on any of the effect sizes. The latter result might be a reflection of low power to detect differences in addition to a potential true finding.
Discussion
This systematic review of the prevalence of psychiatric morbidity associated with ADHD in prison included 18 studies comprised of 1615 individuals with ADHD and 3128 without the disorder. Subgroup analysis allowed to partial out heterogeneity due to age of participants, and its impact on the effect estimates. Several keys findings were produced from this review. First, the risk of (non-ADHD) psychopathology is significantly increased among adult prisoners with ADHD compared with prisoners without ADHD. Second, significant associations with all psychiatric morbidity categories were present in adult offenders with ADHD, but in youth offenders these were limited to mood/affective disorders, highlighting important differences by age. Third, there was greater heterogeneity in effect sizes for youth samples.
Adult prevalence estimates of psychiatric co-morbidity associated with ADHD were up to 74% for SUD and lowest (29%) for CD. However, the largest effect sizes measured by OR were found for mood/affective disorders with an almost 5-fold increase, and anxiety disorders with more than a 3-fold increase in the adult co-existence odds relative to non-ADHD adult prisoners. Adult ADHD prisoners were twice more likely than adult non-ADHD prisoners to have a co-morbid history of CD, SUD and PD. For youth offenders, risk was increased 3-fold for mood/affective disorders. Contrary to expectations, CD was not associated with ADHD in youth studies. This most likely reflects that CD is a common co-morbidity in offender samples in general and this combined problem therefore does not distinguish between prisoners with and without ADHD, e.g. the prevalence estimates by Chang et al. (Reference Chang, Chen and Huang2007), Stahlberg et al. (Reference Stahlberg, Anckarsater and Nilsson2010) and Young et al. (Reference Young, Wells and Gudjonsson2011b ) reported rates of CD at 75%, 77% and 61%, respectively.
Surprisingly, there was a lack of association between ADHD and SUD in youth studies. Nevertheless, considering the exceedingly high prevalence rate of SUD in youths with ADHD (70%), it is likely that this negative finding was due to low power to detect potential associations. The overall substantial prevalence of co-morbidity between youth and adult prisoners with ADHD, and the robust associations among adults, suggest that SUD may be an important mediator in the association between ADHD, delinquency and incarceration. Population studies indicate that 12.5% of adults meet full DSM criteria for a SUD (Kessler et al. Reference Kessler, Adler, Barkley, Biederman, Conners, Demler, Faraone, Greenhill, Howes, Secnik, Spencer, Ustun, Walters and Zaslavsky2006). There is evidence that untreated ADHD is a predictor for the development of SUD (Wilens et al. Reference Wilens, Biederman, Mick, Faraone and Spencer1997), and pooled prevalence rates of ADHD at 23.1% have been reported among treatment seeking substance dependent patients (van Emmerik-van Oortmerssen et al. Reference van Emmerik-van Oortmerssen, van de Glind, van den Brink, Smit, Crunelle, Swets and Schoevers2012). Indeed it is hypothesized that people with undiagnosed or untreated ADHD engage in substance misuse as a means of self-medication (Khantzian, Reference Khantzian1985). The significant effect for ADHD and co-morbid mood/affective disorders (which included depression, mania and suicidal behaviour) suggests that, compared with non-ADHD peers, these youth offenders with ADHD were three times more likely to develop an affective disorder. Emotional dysregulation is increasingly recognized as a presenting problem for adults with ADHD (Gudjonsson et al. Reference Gudjonsson, Sigurdsson, Adalsteinsson and Young2013) and the pooled effects reported here suggest that emotional instability presents from childhood in young offender populations. This has implications for treatment as emotional instability, low mood and substance misuse are problems that are likely to hamper the rehabilitation process. When examining depression separately however, we found an important distinction: adult, but not youth offenders, had the greater risk (x5), which has implications for the management of risk of suicide within correctional services.
The strengths of this meta-analysis are the comprehensive review of studies and the range of psychiatric disorders studied among youth and adult offenders with and without ADHD. However, these findings should be interpreted in the context of several limitations. We found considerable variation from studies conducted in youths, and heterogeneity that was unaccounted for in some adult estimates. This may reflect variation from differences in the methodology employed in these studies, cultural differences and in instruments used for the measurement of ADHD. The latter has been found to produce important differences in meta-analytical rates (Young et al. Reference Young, Moss, Sedgwick, Fridman and Hodgkins2014). Increased heterogeneity in minors may also reflect a process of stabilization in the characteristics of adult inmate populations over time, may be due to more similar individual ‘types’ being repeat offenders or some convergence of disorders with maturity. Additionally, extracted data was not sufficient to perform sensitivity analysis on potential moderators, such as gender, or to model meta-analytical regressions with sufficient power on all outcomes. For instance, limited amount of studies performed in female inmates hinders our ability to conclude whether there is an absence of gender effects on the meta-analytical regression estimates, or if we failed to attain sufficient power to detect such effects (i.e. Type II error). We established rigorous criteria for the ADHD assessment and excluded studies that only used retrospective symptom assessment, but it was not feasible to apply these same principles to the measurement of the other psychiatric categories, therefore some variation in the estimates may derive from their assessment.
This study adds to the evidence regarding the co-morbid presentation of offenders with ADHD and the findings have implications for clinical intervention and for criminal justice policy. Clinical symptoms of ADHD in youth and adult offenders are often missed or misdiagnosed (Baillargeon et al. Reference Baillargeon, Penn, Knight, Harzke, Baillargeon and Becker2010; Young et al. Reference Young, Adamou, Bolea, Gudjonsson, Muller, Pitts, Thome and Asherson2011c ) and it seems that for youth offenders, ADHD is most likely to be misdiagnosed as mood/affective disorders. On the basis of these results, children with ADHD in prison are likely to present additional mood/affective disorders, which may manifest more recognizable symptoms, thus making it likely that an ADHD diagnosis goes unnoticed. Similar factors may be at play for adults who are likely to present with additional symptoms of mood, anxiety or substance use disorders. Research suggests that the pathway from ADHD to depression involves both anxiety and disruptive behaviour, and increases with age due to the continued negative reactions of others to their ADHD and disruptive symptoms (Roy et al. Reference Roy, Oldehinkel, Verhulst, Ormel and Hartman2014). The differences between the effect estimates for youths and adults found in the present study suggest an incremental effect (in both frequency and severity) for the development of co-morbidities. It is likely that a cumulative effect of psychiatric morbidity takes its toll on the young person's social and personal development, leading them to become emotionally less resilient and increasing the risk for the development of clinical and personality co-morbid pathology as they mature.
With re-offending rates related to psychiatric co-morbidity (Baillargeon et al. Reference Baillargeon, Penn, Knight, Harzke, Baillargeon and Becker2010), the identification of effective interventions for this patient group should be a primary concern. In developmental terms, onset of ADHD symptoms most likely will onset before these co-morbid conditions, therefore primary prevention strategies targeting children at risk may help improve long-term outcomes. Further research on this area should aim to better understand the contribution of ADHD and co-existing conditions in the pathways to delinquency and crime.
Acknowledgements
We thank E. Watts (Fishawack Communications Ltd, UK, funded by Shire AG, Switzerland) and G. Stoudt (Shire, USA) for performing the initial searches; J. Wright and K. Lay (Caudex Medical, UK, funded by Shire AG, Switzerland) and K. Brown (Institute of Psychiatry, Psychology and Neuroscience, King's College London) for support with extracting relevant data from the studies. O. Sedgwick receives funding support from the National Institute for Health Research (NIHR) Mental Health Biomedical Research Centre at South London and Maudsley National Health Service (NHS) Foundation Trust and King's College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. R. González was partially funded by NIDA grant 5R24DA024868-05.
Declaration of interest
S.Y. and G.G. have received honoraria for consultancy, travel, educational talks and /or research from Janssen, Eli Lilly, Shire, Novartis, HB Pharma, Flynn Pharma and/or Shire. P.H. was an employee of Shire. He is now Vice President, Global HEOR, Vertex Pharmaceuticals, USA. Shire develops and markets drugs to treat psychiatric disorders, including ADHD. M.F. is a paid statistical consultant to Shire. The other authors have no conflicts of interest.
Open access
