Introduction

Attention-deficit/hyperactivity disorder (ADHD) is among the most frequently diagnosed psychiatric disorders and is estimated to affect 5–10 % of school-age children (Scahill and Schwab-Stone 2000). Longitudinal studies have repeatedly shown that children with ADHD are at heightened risk of poor outcomes as they enter adolescence and early adulthood, and epidemiologic studies have identified elevated rates of substance use, personality disorders, and criminal behavior among young adults with symptoms of ADHD (Gudjonsson et al. 2012). Among poor outcomes, criminality (Barkley et al. 1990; Hechtman and Weiss 1986; Mannuzza et al. 1989, 1993) and substance use disorders (SUDs) (Mannuzza et al. 1998; Wilens et al. 1998; King et al. 2004) have the potential to cause significant hardship for the individual, their family, and society at large.

Several groups have shown elevated rates of SUDs and criminal behavior in ADHD youth when compared to non-ADHD peers. For example, Satterfield and colleagues (1997) showed significantly higher rates of arrests and incarcerations in adolescence and early adulthood for boys with ADHD when compared to controls (Satterfield and Schell 1997; Satterfield et al. 1982). Similarly, Hechtman et al. (1984) documented that young adults diagnosed as hyperactive in childhood reported greater police involvement when compared to matched controls (Hechtman et al. 1984). Biederman et al. (2006) reported elevated rates of antisocial disorders in adolescents/young adults diagnosed with ADHD in childhood when compared to non-ADHD controls (Biederman et al. 2006, 2008). Using official arrest records, Mannuzza et al. (2008) showed that adolescents and adults diagnosed with ADHD in childhood were arrested, convicted, and incarcerated significantly more often than controls (Mannuzza et al. 2008). Further, they reported evidence that criminal outcomes were at least in part mediated by the development of SUDs. Lastly, using official state arrest records, Barkley et al. (2004) found that clinic-referred children with ADHD had significantly more arrests and committed more felonies in early adulthood when compared to controls (Barkley et al. 2004). Interestingly, when separating criminal activities into predatory-overt and drug-related antisocial conduct, the ADHD group differed significantly from the control group only on drug-related activities, with the ADHD group having higher rates.

The aforementioned longitudinal studies have been extremely influential in shaping our current thinking with regard to ADHD and antisocial outcomes; however, several of these samples were initially recruited prior to 1980, before operationally defined criteria for the diagnosis of ADHD had been firmly established (Fischer et al. 1990; Hechtman et al. 1979; Mannuzza et al. 1998; Satterfield et al. 1982). Further, the majority of subjects in these studies were homogeneous in terms of ethnicity (mostly Caucasian) and socioeconomic profiles (mostly middle class) (Mannuzza et al. 1991; Satterfield et al. 1982; Barkley et al. 1990; Biederman et al. 2006), which limits generalizability. Considering the fact that poorer minorities represent the bulk of prison populations (Washburn et al. 2008), it follows that more diverse samples should be the target of research regarding antisocial behavior. Another key factor that has been overlooked in this line of research is the contribution of childhood maltreatment to poor outcome in ADHD youth, despite a wealth of literature outlining its risk in general population studies of antisocial outcomes. This is particularly striking given the facts that a) children with ADHD are at elevated risk of maltreatment (Briscoe-Smith and Hinshaw 2006; Cicchetti and Manly 2001; Ford et al. 2000) and b) population-based prospective studies carried out in the USA consistently report a robust link between a history of childhood maltreatment and later antisocial behavior (Widom 1989; Cicchetti and Manly 2001; Manly et al. 2001; Lansford et al. 2002; Smith and Thornberry 1995; Smith et al. 2005; Ireland et al. 2002).

The present study is unique in that it consists of a referred sample of children with ADHD who were recruited and assessed in childhood between the ages of 7 and 11 years. As the sample was recruited in the 1990s, standardized rating instruments, structured interviews, and DSM criteria were utilized. The children were re-assessed after approximately 10 years and compared with a well-matched, never-ADHD comparison group. Substance use and childhood maltreatment history were assessed for both probands and controls during the adolescent follow-up. Criminal data for the entire sample were obtained from the New York State Division of Criminal Justice Services approximately 3 years after the commencement of the adolescent follow-up. Finally, this sample, both probands and controls, is urban and ethnically and socioeconomically diverse, a group that has been underrepresented in the ADHD literature.

Here, we first characterize the antisocial outcomes among probands and controls. Regression analyses were then used to investigate the relative effect of childhood maltreatment on antisocial outcomes among four groups based on the presence or absence of ADHD and maltreatment. We hypothesized that (1) individuals diagnosed with ADHD in childhood would have elevated rates of criminality and substance use when compared to typically developing controls and (2) maltreatment history would differentially influence criminal and substance use outcome across the four groups.

Method

Participants

Ninety-eight adolescents/young adults who were evaluated in a research protocol during childhood (Halperin et al. 1997) participated in a follow-up evaluation approximately 10 years later. They were drawn from a group of 169 youth who were recruited between 1990 and 1997. Of these, 18 refused participation, one was deceased, five were incarcerated, and 47 were lost to follow-up. Those who were and were not followed did not differ significantly in age at initial evaluation, race/ethnicity, sex, childhood comorbidity, socioeconomic status (SES), or ADHD behavior ratings at baseline. All participants were 7–11 years old; mean (SD) age at initial evaluation was 8.9 (1.3) years. The original sample was comprised of 21.2 % Caucasian, 26.5 % African-American, 40.4 % Hispanic, and 11.9 % mixed ancestry. The group was primarily of lower to lower-middle SES [mean (SD) = 31.6 (14.51), range = 11–66] (Nakao and Treas 1994), with a large portion at the poverty level. All participants were English-speaking. Individuals were referred for behavioral difficulties by schools and mental health providers. The childhood sample was rated as having significant behavior problems by both parents and teachers, and all participants were diagnosed with ADHD according to Diagnostic and Statistical Manual for Mental Disorders 3rd edition, revised (i.e., DSM-III-R) or 4th edition (i.e., DSM-IV) criteria (depending on the date recruited, as the diagnostic system changed in the middle of the study). Parent and teacher reports using the Child Behavior Checklist (Achenbach 1991) and IOWA Conners Rating Scale (Loney and Milich 1982), respectively, were also obtained. Childhood assessments determined using DSM-III-R were re-coded to be consistent with DSM-IV; it is likely that most, if not all, participants met criteria for ADHD combined type (ADHD-C). Children were assessed cognitively using the WISC-III (Wechsler 1974, 1991). Table 1 shows childhood characteristics of the ADHD sample. Diagnoses of schizophrenia, pervasive developmental disorder, or Tourette’s disorder, or a full scale IQ below 70 were excluded. Eighty-two (48.5 %) subjects met criteria for ODD, 53 (31.4 %) met criteria for CD, 55 (32.5 %) met criteria for at least one anxiety disorder, and 22 (13.0 %) had a mood disorder.

Table 1 Baseline Characteristics of the Childhood ADHD Sample
Full size table

In addition, 85 never-ADHD controls were recruited during the adolescence/young adulthood follow-up via advertisements in neighborhoods that matched the ADHD sample by zip code. Control participants were screened for lifetime presence of ADHD symptoms, treatment for ADHD symptoms, and previous difficulties related to ADHD symptoms. Those with evidence of prior or current ADHD symptoms were excluded from the study. Like the original ADHD sample, prospective controls were excluded if they had any chronic medical/neurological condition or psychosis or were non-English-speaking. Controls were not excluded for the presence of psychiatric disorders other than those that were also excluded for the ADHD children and therefore did not represent a group of ‘super-normals.’

Among the 183 probands and controls who participated in the follow-up study, 161 (88.0 %) were male and all but three were between the ages of 16 and 21 years of ageFootnote 1 [mean (SD): 18.4(1.7)]. The adolescent sample, like the childhood sample, was ethnically diverse—with 26.8 % African-American, 35.5 % Latino, 25.1 % Caucasian, and 12.6 % mixed ethnicity. Participants were generally of lower to lower-middle SES (mean = 36.5, SD = 17.8, range 11–85) (Nakao and Treas 1994) and almost exclusively urban. There were no significant differences between the ADHD and control groups in age (18.4 vs. 18.5 years), sex (88.8 vs. 87.1 % male), SES (43.6 vs. 40.9), or FSIQ (93.0 vs. 96.8) (all p values >.05).

This study was approved by the Institutional Review Boards of the Mount Sinai School of Medicine and Queens College, City University of New York. Participants over the age of 18 signed their own statement of informed consent for participation in the study. When participants were under the age of 18, parents signed written statements of informed consent for their own and their child’s participation, and assent was obtained from all participants younger than 18. Participants were compensated for their time and travel expenses.

Measures

Assessment of childhood maltreatment

Childhood maltreatment was assessed during the follow-up evaluation using the short form of the Childhood Trauma Questionnaire (CTQ-SF (Bernstein et al. 1994, 2003), a 28-item self-report measure that screens adults and adolescents for histories of childhood abuse and neglect. The CTQ-SF is a brief, reliable, and valid means of retrospectively assessing childhood maltreatment, with test–retest reliability coefficients ranging from .79 to .86 and internal consistency coefficients ranging from .66 to .92 across samples (Bernstein et al. 1997, 2003). Subjects rated statements about childhood trauma on a 5-point Likert scale as ‘never true,’ ‘rarely true,’ ‘sometimes true,’ ‘often true,’ and ‘very often true.’ Minimization and denial of abuse and neglect are rated on a three-item scale that is incorporated into the questionnaire to detect false-negative trauma reports. The CTQ-SF assesses five types of maltreatment: emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. The CTQ-SF provides cutoff scores from none to low, low to moderate, moderate to severe, and severe to extreme exposure for each of the five types of maltreatment. For the purposes of this study, a dichotomous variable of maltreated/not-maltreated was created where individuals were categorized as maltreated if they met criteria for one or more subtypes of maltreatment using the cutoff scores provided in the CTQ manual.

Assessment of socioeconomic status

Socioeconomic status was assessed using a measure of socioeconomic prestige developed at the National Opinion Research Center (Nakao and Treas 1994). This measure approaches the issue of measuring SES by ranking the relative prestige of an individual’s occupation on a scale from 1 to 100. Information used to determine the socioeconomic prestige score was obtained from the parents during both the baseline and follow-up assessments as most participants were still living with their parents.

Assessment of adolescent substance use

Adolescent substance use was assessed using several measures. The Rutgers Alcohol and Drug Use Questionnaire (Labouvie et al. 1997) was used to systematically evaluate the subject’s overall drug and alcohol use. The RADQ assesses current and past use of cigarettes, alcohol, marijuana, cocaine, and other prescription and non-prescription drugs. Respondents report the frequency and amount of drug and alcohol use in the past 3 years. Secondly, the substance abuse supplemental module of the Kiddie-SADS-Present Lifetime version (K-SADS-PL; (Kaufman et al. 1996) was used to ascertain current and lifetime psychiatric disorders based on DSM-IV criteria, including SUDs. Interviews were conducted by trained graduate students and postdoctoral fellows, separately with adolescents and their parents as informants. Responses were combined across informants; if either informant indicated that the item caused significant distress or impairment, the symptom was judged to be present. Finally, a urine toxicology screen was collected from each subject on the day of evaluation, analyzed for the presence of marijuana, cocaine, amphetamines, and opiates, and used to corroborate subject report, although this could not be used to determine the proband’s diagnostic status. In addition, to facilitate honest responding and to maintain strict measures of confidentiality, we obtained a certificate of confidentiality from the National Institute of Health. Information gathered from all sources was used to inform the clinician determination of the individual subject’s drug and alcohol habits and guided the clinician in probing for the specific criteria needed to make a diagnosis of drug or alcohol abuse and/or dependence in accordance with the criteria set forth in the DSM-IV. Measures of abuse and dependence were collapsed to create a dichotomous variable of substance abuse/dependence versus no substance abuse/dependence.

Criminality

Detailed juvenile and adult criminal records for the entire sample were obtained from the New York State Division of Criminal Justice Services, Albany, which houses the official data for all offenses committed in New York State. A detailed description of our prospective follow-up study (including methods, sample characteristics, goals, and significance) was submitted to the Division of Criminal Justice Services for review. On approval, a non-disclosure agreement was signed by the principal investigator (J.M.H), stipulating that arrest history data would not be disclosed in a manner that could identify any individual, would be used only for research purposes, and would be treated as strictly confidential. This large criminal records data set includes numerous variables such as age of first arrest, number of arrests, type of offense (drug related or not), conviction records, as well as details of recidivism.

Statistical procedures

Chi-square analyses were used to determine group (ADHD/control) differences in rates of childhood maltreatment, SUDs, arrest, conviction, incarceration, and drug-related crime (dichotomous measures). Student’s t tests were used to determine group differences in age at criminal database (CDB) collection, childhood maltreatment severity, age of first arrest, and rates of recidivism (continuous variables). To further characterize the sample, subjects were assigned to four groups based on the presence or absence of ADHD and maltreatment: (1) without ADHD, without maltreatment (n = 31); (2) without ADHD, with maltreatment (n = 51); (3) ADHD without maltreatment (n = 23); and (4) ADHD with maltreatment (n = 61). Arrest and SUD outcomes were compared between these four groups using logistic regression analyses with the no-ADHD/no-maltreatment group serving as a reference group. Age and SES were entered into the first step of the analysis as control variables.

Additionally, we used logistic regression analyses to explore whether SUD status was associated with arrest or, more specifically, drug-related arrest. Age and SES were entered as control variables on the first step, SUD diagnosis and group status were entered as predictor variables on the second step, and an SUD-by-group interaction term was entered on the third step.

Results

At the time of the criminal database collection, 41 probands (41.8 %) and 30 controls (35.3 %) had been arrested. Rates of arrest, conviction, incarceration, and recidivism did not differ significantly between probands and controls (all p values >.05), although probands were significantly younger than controls at the time of their first arrest [16.9 (1.7) vs. 18.2 (2.3); t (63) = − 2.6, p = .011]. Group differences for select variables are listed in Table 2. When divided into four groups based on group status and maltreatment history, probands with maltreatment were significantly more likely to have been arrested when compared to controls without maltreatment (OR, 3.12; 95 % CI 1.02–9.57; p = .05). Controls with a history of maltreatment did not have significantly higher rates of arrest when compared to the reference group.

Table 2 Group differences on select variables
Full size table

In terms of SUD outcome, probands had significantly higher rates of SUD diagnoses at the adolescent follow-up (47.4 vs. 29.4 %; χ 2 = 6.2, p = .01). Probands and controls with maltreatment were more likely to be diagnosed with an SUD when compared to controls without maltreatment (OR 10.27; 95 % CI 3.05–34.61; p < .001, OR 4.07; 95 % CI 1.18–14.07; p = .03). Individuals with an ADHD diagnosis and a history of maltreatment were two and a half times more likely to have an SUD diagnosis when compared to controls with a history of maltreatment (see Fig. 1). Surprisingly, although the two groups differed significantly in rates of SUDs, with probands almost twice as likely to obtain an SUD diagnosis when compared to controls, rates of drug-related arrests did not differ between the two groups. Lastly, SUD diagnosis, group status, and group-by-SUD-diagnosis did not significantly predict arrests or drug-related arrests.

Fig. 1
figure 1

Risk of arrest and substance use disorder diagnosis by maltreatment history and ADHD status. *p < .05; ***p < .001

Full size image

Discussion

This prospective study reported on criminal and substance use outcomes in an ethnically diverse, lower-SES sample of urban youth with and without ADHD and investigated the influence of maltreatment on outcome. Consistent with the extant child abuse literature, our results show that children with a history of maltreatment have elevated rates of later criminality (Widom 1989; Ireland et al. 2002; Smith et al. 2005) and SUDs (Liebschutz et al. 2002; Moran et al. 2004; Ondersma 2007; Wall and Kohl 2007). In contrast to most of the major prospective studies of children with ADHD (Mannuzza et al. 1989; Barkley et al. 1990; Fischer et al. 2002; Weiss et al. 1985), we did not identify significant discrepancies between probands and controls in rates of arrest, drug-related arrest, conviction, incarceration, or recidivism. Arrest rates were quite high for controls in this sample (35.3 %) when compared to controls in other studies, most probably a reflection of the sociodemographic composition of the sample. This suggests that a diagnosis of ADHD may have less of an impact when studying criminal outcomes among lower-SES, minority urban youth. The same cannot be said for substance use outcomes. Despite possible ‘mitigating’ sociodemographic variables, we identified significant discrepancies between ADHD and control groups in rates of SUDs, in accordance with extant studies of ADHD youth.

Similar to Mannuzza et al. (1989), we found that subjects with an arrest record did have significantly higher rates of SUDs compared with subjects without an arrest record. (66.2 vs. 21.4 %; χ 2 = 36.2, p = < .001). This suggests that criminality may be at least in part mediated by substance use, although our data do not provide sufficient information to discern the temporal sequence of events (e.g., substance use preceded or following arrest). In contrast to the findings of Barkley et al. (2004), SUD diagnosis was not a significant predictor of drug-related crime in our sample.

When compared to a no-ADHD/no-maltreatment reference group, both ADHD youth and controls with maltreatment had significantly higher rates of SUD. When compared to the control group with a history of maltreatment, ADHD youth with maltreatment were two and a half times more likely to carry an SUD diagnosis. This result highlights the need to consider childhood maltreatment when assessing and treating ADHD youth, particularly with regard to substance use risk.

There are several limitations to be considered when interpreting these findings. First, there is a relatively high attrition rate from the original sample. The difficulty to find many individuals from this highly mobile, largely lower-SES group is not surprising—especially given the passage of nearly 10 years. While available data suggest that the follow-up subsample is largely representative of the original group, it is possible that those lost to follow-up had different and perhaps worse outcomes. Secondly, identification of childhood maltreatment was based solely on retrospective reports from each participant. Although considerable data support the reliability and validity of the CTQ (Bernstein et al. 1997, 2003; Fink et al. 1995; Scher et al. 2001), self-report measures are susceptible to a variety of biases including social desirability, mood at the time of report, and memory limitations. Nevertheless, had maltreatment been assessed during childhood, there would be increased likelihood for false negatives due to parental underreporting as well as the possibility of the childhood maltreatment occurring subsequent to our childhood evaluation. Furthermore, as is true of other longitudinal research in the area of ADHD, this sample is largely male, and it is therefore difficult to generalize the findings to females. While the use of official state criminal records captures more data on criminal activity, this method overlooks those individual cases that were redirected outside the state and courts or when charges were dropped due to errors. Finally, the present study did not control for possible contributory effects of comorbid childhood psychopathologies (e.g., CD) or familial factors (e.g., parental psychopathology). The contributions of childhood psychiatric comorbidities, mainly CD, have been established as a potent risk factor for both criminal and SUD outcomes in several studies of ADHD youth (Armstrong and Costello 2002; Brook et al. 1995; Disney et al. 1999; Barkley et al. 2004) and should be included in future investigations into the salience of maltreatment as an independent risk factor for poor outcome in ADHD youth.

In conclusion, we found that a well-replicated finding—namely an increase in criminal outcome for ADHD youth—is not necessarily applicable to samples of lower-SES, ethnically diverse urban youth. Results of existing longitudinal studies in Caucasian children with ADHD may have limited the ability to fully understand the key issues regarding antisocial outcomes in non-Caucasian, urban children of lower SES. Considering the fact that minorities are much more likely to be arrested and convicted and subsequently represent the bulk of prison populations (Washburn et al. 2008), more diverse samples should be the target of future ADHD research in order to develop a clearer understanding of the mechanisms that underlie antisocial behavior in this group. Additionally, maltreatment history emerged as a potent predictor of SUD and criminal activity among both ADHD youth and non-ADHD peers, highlighting the need for maltreatment screening to identify at-risk youth. These results suggest that ADHD, in the absence of maltreatment, may not confer added risk, at least in an ethnically diverse, lower-SES sample.