Health Associations of Drug-Involved and Criminal-Justice-Involved Adults in the United States

Abstract

A burgeoning criminological literature has identified important intersections between public health, crime, and antisocial behavior. This study is based on public-use data collected between 2006 and 2010 as part of the National Survey on Drug Use and Health (NSDUH) and an analytical sample of men (N = 84,054) and women (N = 95,308) between the ages of 18 and 64. Latent class analysis (LCA) identified three classes: a large normative group, a small drug-involved group, and a criminal-justice-involved group. Chronic health conditions that are more closely associated with longer term medical problems and perhaps cumulative stress such as heart disease and diabetes are not linked to criminal-justice-system-involved or drug-involved offenders. Medical problems that are more closely related to an antisocial lifestyle such as sexually transmitted diseases, pancreatitis, and hepatitis were found to be more prevalent among antisocial subgroups in this sample.

Keywords

chronic health health mental health antisocial behavior crime medical conditions violence drug abuse

The idea that antisocial behavior is associated with increased odds of poor health, including mental health, has gained traction in recent years (Odgers et al., 2007; A. R. Piquero, Daigle, Gibson, Leeper Piquero, & Tibbetts, 2007; A. R. Piquero, Farrington, Nagin, & Moffitt, 2010; A. R. Piquero, Shepherd, Shepherd, & Farrington, 2011; Vaughn, Fu, et al., 2011). Although crime is commonly viewed as a matter to be handled by agents of the criminal justice system (police, courts, and corrections), it is also a public health issue. The intersection of criminal justice populations and public health systems has become increasingly intertwined (Bond & Gittell, 2010; Schroeder, Hill, Haynes, & Bradley, 2011; Stogner & Gibson, 2010; Vaughn & Howard, 2004). In fact, this overlap is becoming so marked that some scholars have called for the creation of a field known as epidemiological criminology or criminal justice epidemiology (Akers & Lanier, 2009; Vaughn, DeLisi, Perron, Beaver, & Abdon, 2012). Medical costs of antisocial behaviors are enormous stemming from emergency room visits, hospital stays, early mortality, and the spread of sexually transmitted disease. Indeed, health care costs associated with crime and drug abuse are in the hundreds of billions annually (Cohen, 1998; Cohen & Piquero, 2009; DeLisi & Gatling, 2003).

Prior investigators have explored the health consequences of an antisocial lifestyle. In a study using data from the Cambridge Study in Delinquent Development, Farrington (1995) revealed that male offenders were more likely to be treated in the hospital for serious illness and accidents. Also using the Cambridge data, Shepherd, Farrington, and Potts (2004) found that the predictors of delinquency also predicted injury and cardiovascular disease. Subsequent analyses of these data found that high-rate chronic offenders were also disproportionately more likely to die earlier than other types of offenders (Piquero, Farrington, Shepherd, & Auty, 2013). Using data from Baltimore, Piquero et al. (2007) compared life-course-persistent offenders to adolescent-limited offenders and reported that life-course-persistent offenders were at higher odds to experience adverse health outcomes (e.g., asthma, hypertension, kidney problems, diabetes, heart disease, and ulcers) and psychological distress. Most recently, Chassin, Piquero, Losoya, Mansion, and Schubert (2013) found that among youthful offenders several different distal and proximate variables predicted early death, but that the array of variables could not explain the heightened risk for African Americans. Despite the contributions of these studies, there remains a relative lack of large-scale studies and investigations that can pinpoint the causal mechanisms that link antisocial behavior to poor health (see Reingle, Jennings, Piquero, & Maldonado-Molina, 2013; Stogner, Gibson, & Miller, 2013).

Theoretical Context

The distal context that conditions the relationships between antisocial behavior and health is complex. The social determinants of health disparities, which are closely linked to social inequality, also give rise to crime and antisocial behavior. Thus, poverty, low levels of educational attainment, and social exclusion are connected to crime and health. The physiological demands placed on individuals residing in adverse or disadvantaged environments are a result of an increased number of stressors encountered (Mezuk et al., 2010). Indeed, stress serves as an important catalyst for the interplay between health problems and antisocial behavior (see Agnew, 1992; DeLongis, Folkman, & Lazarus, 1988).

Recent research that has articulated linkages between antisocial behavior, including violence, poor health, and stress, draws on multiple theoretical perspectives (see Reingle et al., 2013). Specifically, this research implicates two related theories, general strain theory (Agnew, 1992, 2006) and allostatic load theory (McEwen, 1998). From the point of view of general strain theory, individuals are pressured, due to life strains and resulting negative emotional states toward crime as an option (Stogner & Gibson, 2010). From the perspective of allostatic load theory, high-risk populations, such as offenders, often lead stressful lives that set into motion a hypervigilant state of arousal. The cumulative burden of a chronically engaged state of arousal results in high cortisol release that eventually wears down the body’s ability to mount successful adaptations to this stress load (McEwen, 1998). Allostatic load is known to contribute to chronic disease risk such as cardiovascular disease and is a primary linkage of antisocial behavior to health. Unlike allostatic load theory, which views stress as a result of antisocial behavior, researchers have used general strain theory as an explanation for crime rather than a consequence (e.g., Schroeder et al., 2011). Nevertheless, both theories link health outcomes to an antisocial lifestyle via stress mechanisms.

Although these viewpoints are plausible, theory pertaining to self-control provides another lens from which to view the associations between antisocial behavior and health outcomes. From this vantage point, poor health outcomes are the result of living for immediate rewards at the expense of longer-term care of self (de Ridder, Lensvelt-Mulders, Finkenauer, Stok, & Baumeister, 2012). Gottfredson and Hirschi (1990) advanced that self-control was the quintessential individual-level predictor of crime and problem behaviors. In their view, low self-control is characterized by impulsiveness, reactive aggression, action (as opposed to verbal reasoning), and preference for quick and easy payoffs as opposed to planning for the long-term. In the Dunedin birth cohort study, Moffitt and colleagues (2011) found that childhood self-control predicted a health outcome composite measure at age 32. These results held even after controlling for the confounding effects of socioeconomic status and IQ. Childhood assessments of self-control in this same study also predicted substance use and dependence and crime. At least in the near term, low-self-control is plausibly linked to health outcomes reflecting substance misuse (e.g., hepatitis) and a more here-and-now orientation (e.g., sexually transmitted disease) rather than health conditions more closely tied to stress pathways such as cardiovascular illness (see, for example, A. R. Piquero, Gibson, & Tibbetts, 2002).

Sex Differences

Distinct differences between males and females in the prevalence of and general liability to engage in antisocial behaviors are also seen in various behavioral disorders and other problem behavior conditions reflective of externalizing (Eme, 2010). Sex ratios for externalizing disorders such as Conduct Disorder, Antisocial Personality Disorder, psychopathy, life-course-persistent offending, career criminality, and other typologies of serious crime indicate that males are significantly more likely to display impairments compared with females (DeLisi, 2013; DeLisi & Piquero, 2011; Eme, 2007). Studies of severe antisocial behavior show that males are more likely to possess greater impairments in self-control including control over emotions and greater resistance to aversive conditioning and relative fearlessness (Chapple, Vaske, & Hope, 2010; C. L. Gibson, Ward, Wright, Beaver, & DeLisi, 2010; Lykken, 1995; Vaughn, Beaver, DeLisi, & Wright, 2009). In a recent meta-analysis of over 250 studies, Cross, Copping, and Campbell (2011) found that females had better self-control, were more sensitive to punishment, and were less likely to take risks. Thus, there is reason to believe that the relationships between antisocial behavior and health are contingent on gender. Furthermore, examining the extent to which females who have contact with the criminal justice system experience similar or different rates of chronic health problems, including mental health disorders, is important for building a knowledge base on crime and public health. Unfortunately, the knowledge base regarding adolescent females who have had contact with the criminal justice system, as well as the nature and extent of their physical and mental health, has been sorely lacking (Odgers, Robins, & Russell, 2010). Existing studies suggest that females in the juvenile justice system in particular are more likely to report mental health and substance use disorders and experience a disproportionate amount of physical and sexual health problems (Robins, Odgers, & Russell, 2010). Clarifying the theoretical underpinnings of this overall discrepancy across sex, as well as the extent to which differential involvement in particular forms of antisocial behavior help to generate these sex differences has also been ill-researched (though see Repetti, Taylor, & Seeman, 2002). However, that which does exist has focused on the early life stressors (maltreatment, victimization, exposure to violence, risky families) that females are differentially exposed to compared with males and/or the extent to which females respond to these stressors in different ways compared with males (internalization vs. externalization; see, for example, Broidy & Agnew, 1997; N. L. Piquero & Sealock, 2004). In short, the little evidence that does exist exploring the interrelationships between sex, antisocial behavior, and negative health outcomes tends to suggest that there may be unique mechanisms and/or a female-specific pathway linking antisocial behavior to poor health (Odgers et al., 2010, p. 431).

The Current Study

Although there is a growing recognition in the literature that there are important intersections between public health and crime and antisocial behavior (cf., Akers & Lanier, 2009; Jennings & Reingle, 2012; Reingle et al., 2013; Vaughn et al., 2012), relatively few investigations have accrued that have explicitly examined associations between health and patterns of antisocial involvement derived from large nationally representative epidemiologic databases. The present study uses 5 years of data derived from the National Survey of Drug Use and Health to assess the associations between health status and criminal and antisocial behaviors among latent subgroups of males and females in the United States. We take advantage of 5 years of independent cross-sectional surveys with identical variables, which enables us to explore heterogeneity in the form of latent subgroups and examine the associations between subgroup membership and health among over a quarter of a million persons in the United States. Although not longitudinal and predictive, we are nevertheless able to paint a strong empirical and generalizable portrait of health and antisociality. The investigation has three objectives: (a) Provide a comprehensive epidemiologic statement about the patterned nature of health and crime and antisocial behaviors. (b) Determine and specify the heterogeneity of criminal behavior and drug use/abuse among latent subgroups. (c) Examine sex differences in these patterns.

We hypothesize that highly deviant subgroups will exhibit a higher prevalence of impulse-control-related health conditions such as sexually transmitted diseases and hepatitis rather than those reflective of cumulative physiological stress and inflammation such as high blood pressure and cardiovascular disease. We also hypothesize that the associations between health and antisocial behavior in the form of our latent subgroups will be contingent on gender such that relationships will be magnified among males compared with females.

Method

Sample and Procedures

This study is based on public-use data collected between 2006 and 2010 as part of the National Survey on Drug Use and Health (NSDUH; SAMHSA, 2011). Consistent with previous studies, NSDUH data were pooled to increase the analytic sample size and improve population estimates (Hedden & Gfroerer, 2011; Nakawaki & Crano, 2012). The NSDUH provides population estimates of drug use and health-related behaviors in the U.S. general population. It utilizes multistage area probability sampling methods to select a representative sample of the U.S. civilian, noninstitutionalized population aged 12 years or older for participation in the study. Study participants include household residents, civilians residing on military bases, and residents of shelters, rooming houses, and group homes.

A total of 280,098 respondents aged 12 years or older completed the survey between years 2006 and 2010. Weighted response rates for these years were approximately 90% for household screening and 75% for interviewing (SAMHSA, 2011). Each independent, cross-sectional NSDUH sample is considered representative of the U.S. general population aged 12 years or older. No individuals were interviewed more than once. To improve the precision of behavioral and drug use estimates for subgroups, adolescents aged 12 to 17 years and young adults aged 18 to 25 years were oversampled. A more detailed description of the NSDUH sampling and data collection procedures are documented in greater detail elsewhere (SAMHSA, 2011).

The current study restricted analyses to men (N = 84,054) and women (N = 95,308) between the ages of 18 and 64. This was due to the lack of health and mental health variables in the adolescent subsample interview. The median age of respondents was between 26 and 29 years of age. In terms of race/ethnicity, approximately two thirds of the respondents were White (66.5%), 12.1% were African American, and 14.8% were Hispanic. The annual family income of the majority of the sample was greater than $50,000 per year (51.6%) with roughly one third (33.4%) of the sample residing in households earning more than $75,000 per year; however, more than one in six respondents (17.1%) reported residing in households with a total income of less than $20,000 per year, and nearly one third (31.3%) resided in households earning between $20,000 and $49,000 per year.

Measures

Indicator Variables

This study identified latent classes among males and females using five independent, cross-sectional NSDUH samples in the general population. The variables used to identify latent subgroups were based on the prevalence of 16 behavioral items in the domains of illegal behavior, criminal justice system involvement, and illicit substance use as described below.¹

Illegal behaviors

Four indicator variables measured respondents’ participation in illegal behaviors over the previous 12-month period, including theft, driving while intoxicated, drug selling, and violent attacks. Respondent participation in the aforementioned illegal behaviors was determined on the basis of questions regarding the frequency of engagement in such behaviors. For instance, respondents who took part in theft were identified by responding affirmatively to the question, “During the past 12 months, how many times have you stolen or tried to steal anything worth more than $50?” Respondents in each survey who reported any participation in each of these illegal behaviors were coded as 1 while those who reported no participation were coded as 0.

Criminal justice system involvement

Six indicator variables measured respondents’ involvement in the criminal justice system over the last year for a variety of offenses, including public drunkenness, drug offenses, driving under the influence, theft-related offenses such as larceny, burglary, breaking and entering, and motor vehicle theft, and violent crimes such as robbery or assault and battery. Respondent involvement in the criminal justice system was determined on the basis of questions relating to being arrested and booked for the aforementioned crimes. For example, respondents who were arrested for driving under the influence responded affirmatively to the question, “In the past 12 months, were you arrested and booked for driving under the influence of alcohol or drugs?” Respondents were also queried regarding having been arrested and booked for any offenses (with the exception of minor traffic violations) during the previous 12 months. Respondents who reported involvement in the criminal justice system for any of the aforementioned crimes were coded as 1 while those who were not arrested and booked were coded as 0.

Illicit substance use

Six indicator variables measured respondent use of illicit substances over the previous 12 months, including marijuana, cocaine/crack, hallucinogens (including ecstasy), stimulants (including methamphetamine), tranquilizers, and heroin or Oxycontin. Respondent use of the aforementioned illicit substances was determined on the basis of questions regarding the frequency of drug use. For example, respondents who were identified as having used marijuana responded affirmatively (i.e., nonzero) to the question, “On how many days in the past 12 months did you use marijuana or hashish?” Respondents who reported using the aforementioned illicit substances one or more times during the previous 12 months were coded as 1, while those who reported no use of the substances were coded as 0.

Indicator covariates

Several demographic variables were included as covariates to provide additional data to refine behavioral subgroups and assess how the latent subgroups vary across important characteristics such as age, race, and income. The following variables were used: age, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and other [American Indian or Alaska Native, Asian, other Pacific Islander or Native Hawaiian, and persons reporting more than one race]), and total annual family income (less than $20,000, $20,000 to $49,999, $50,000 to $74,999, and $75,000 or more). Sex was not used as an indicator covariate given that the identification of latent subgroups was conducted separately for men and women.

Health and Mental Health Variables

Health characteristics

Thirteen items (0 = no, 1 = yes) were used to assess the health status of respondents in terms of current diagnoses of serious diseases, including asthma, cirrhosis, diabetes, heart disease, hepatitis, lung cancer, pancreatitis, pneumonia, sexually transmitted diseases, stroke, tuberculosis, and stomach ulcers. Determination of these illnesses was based on whether respondents indicated they were informed by a physician or medical professional that they met diagnostic criteria for these disorders.

Mental health

Four items (0 = no, 1 = yes) were used to assess various manifestations of mental health disorders and related outcomes including depression, anxiety, dual diagnosis, and suicidal ideation. Specifically, determination of lifetime depression and anxiety (0 = no, 1 = yes) was based on whether respondents had ever been informed by a doctor or medical professional that they met criteria for either of these disorders. Dual diagnosis (0 = no, 1 = yes) was determined if respondents were found to (a) meet criteria for mental illness according to the World Health Organization Disability Assessment Schedule (WHODAS; World Health Organization, 2012) and (b) meet criteria for a substance use disorder according to the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association [APA], 1994). Suicidal ideation was determined on the basis of the following question: “At any time in the past 12 months, did you seriously think about trying to kill yourself?” Respondents who reported having recently thought about committing suicide were coded as 1 and all other respondents coded as 0.

Chemical dependence

Two items (0 = no, 1 = yes) were used to assess various forms of chemical dependence: nicotine dependence and alcohol dependence. Alcohol dependence was determined based on the criteria of the DSM-IV (APA, 1994). Nicotine dependence was determined on the basis of the current score on the Nicotine Dependence Syndrome Scale (Shiffman, Waters, & Hickcox, 2004) and the Fagerstrom Test of Nicotine Dependence (Fagerstrom, 1978), as administered by the NSDUH.

Statistical Analyses

Latent class analysis (LCA) and multinomial regression analyses were carried out in successive steps to identify and distinguish latent subgroups of adult men and women in the study sample. LCA is a statistical procedure designed to assign individual cases to their most likely latent subgroups on the basis of observed data (McLachlan & Peel, 2000). Multinomial regression is a statistical procedure designed for nominal outcomes that contain categories that can be assumed to be unordered (Long & Freese, 2006). In this two-part analysis, LCA was carried out to identify latent behavioral subgroups and multinomial regression was utilized to examine associations between identified subgroups and health variables. Thus, the odds of health and mental health being associated with a particular latent class are assessed.

Beginning with the LCA, a sequence of latent class models were identified between one and five classes using Latent GOLD® 4.5 (Vermunt & Magidson, 2008). Three statistical criteria were used to identify the best fitting model: the Bayesian Information Criterion (BIC), Log Likelihood, and Entropy. While multiple information criterion statistics are available, simulation studies tend to support the BIC as a superior indicator compared with other statistics such as Akaike’s Information Criterion (AIC) and Consistent Akaike’s Information Criterion (CAIC; Yang, 2006). In interpreting these criteria, lower BIC values, higher log likelihood values, and higher entropy values reflect better model fit and, in the case of entropy, greater accuracy of classification. Importantly, in addition to these quantitative criteria, the parsimony and substantive interpretability of the latent class solutions also function as key criteria for the selection of the final model.

After identifying latent behavioral subgroups using LCA and assigning subjects to classes on the basis of the probability of membership (i.e., persons are assigned to the class to which they have the highest probability of belonging to), multinomial regression is used to examine associations between class membership and health variables. As such, we are not predicting health outcomes as a function of group membership but are instead parsimoniously examining associations between subgroup membership and health conditions. It is parsimonious because if we used each health condition as a dependent variable this would entail 18 regression models (i.e., one for each health condition including mental health). As is typically practiced, the class containing the greatest number of respondents was identified as the reference category. Using multinomial regression, relative risk ratios and confidence intervals were estimated. In the case of multinomial regression with latent classes, relative risk ratios refer to the likelihood of membership in one particular class versus a specified reference class and are interpretably akin to odds ratios (Zhang & Yu, 1998). Statistical procedures involving multinomial regression models were conducted using Stata 12.1SE survey data functions (StataCorp, 2011). This system implements a Taylor series linearization to adjust standard errors of estimates for complex survey sampling design effects including clustered multistage data.

Results

Identifying Latent Subgroups Using Lca

An examination of the statistical and substantive criteria indicated that a three-class solution was the optimal modeling of the data for male and female adults ages 18 to 64. As seen in Table 1, while the BIC and log likelihood values for the four class models are slightly better to those of the three class models, the small differences between the three and four class models suggest that the addition of a fourth class may not be parsimonious. In addition, for male and female adults, the highest entropy values are observed for the three-class solutions, indicating that the three-class models also are the most accurate in terms of classification. In addition to these statistical criteria, the clear interpretability of the three-class solution suggests that this solution offers a statistically sound and a conceptually clear modeling of the behavioral heterogeneity within the data.

Table 1:

Fit Indices for Latent Classes Among Men and Women

Class solution	BIC	Log likelihood	Entropy
Men
1 Class	511,193.96	−255,506.26	n/a
2 Classes	433,758.01	−216,652.22	0.80
3 Classes	417,951.38	−208,612.84	0.84
4 Classes	412,700.75	−205,851.45	0.77
Women
1 Class	385,459.69	−192,638.12	n/a
2 Classes	328,391.56	−163,966.48	0.75
3 Classes	318,729.11	−158,997.68	0.78
4 Classes	314,339.02	−156,665.06	0.70

Note. BIC = Bayesian Information Criterion.

For adult males and females, the substantive interpretability of the three class models were examined by plotting the prevalence estimates of the four illegal behavioral variables, the six criminal justice system variables, and the six illicit substance use variables across each of the latent classes. As seen in Figure 1, very similar class solutions were identified for male and female adults. Among men and women, the three-class solution comprised a normative class (men: 81.83%; women: 86.80%), a drug-involved class (men 11.29%; women: 10.69%), and a criminal-justice-system-involved class (men: 6.88%; women: 2.51%). Notably, the normative class is slightly larger and the criminal-justice-system-involved class slightly smaller among female adults than among male adults. For men and women, the normative class comprises more than four fifths of the sample and, with the exception of driving while intoxicated (men: 16.2%; women: 8.8%) and marijuana use (men: 11.7%; women: 6.1%), is characterized by extremely low levels of participation in illegal behaviors, criminal justice system involvement, and illicit substance use. In contrast, the male and female drug-involved class is characterized by elevated levels of illegal behavior, extremely low criminal justice system involvement, and elevated levels of illicit drug use. The criminal justice-system-involved class, which as noted above is larger in the sample of adult men than in the sample of adult women, is characterized by very high levels of recent arrests and a variety of arrests for particular criminal behaviors as well as levels of illegal behavior and illicit substance use that are elevated but slightly lower than those of the members of the drug-involved class. In sum, the three-class solution represents an interpretable and substantively meaningful modeling of the heterogeneity of illegal behavior, criminal justice system involvement, and illicit drug use behavior among adult males and females in this large nationally representative data source.

Figure 1:

Past Year Prevalence of Antisocial Behavior, Criminal Justice System Involvement, and Substance Use for Men and Women Ages 18-64 Across 3 Latent Classes

Bivariate Sociodemographic Characteristics of Latent Classes

Table 2 presents the percentages and confidence intervals for age, race/ethnicity, and family income among adult men and women across the three latent behavioral classes. For men and women, the normative class has the largest proportion of respondents between the ages of 35 and 64 (men: 66.6%; women: 66.7%). Across sex, this proportion of older respondents among the normative class is approximately twice that of the drug-involved (men: 24.9%; women: 24.8%) and criminally engaged (men: 32.9%; women: 34.3%) classes. In terms of race/ethnicity, among men and women, the class with the highest proportion of White respondents is the drug-involved class (men: 74.0%; women: 77.6%) and the class with the highest proportion of African American respondents is the criminal-justice-system-involved class (men: 20.8%; women: 20.3%). Among Hispanics, a slightly different pattern is observed among male and female adults; the class with the largest proportion of Hispanic men is the criminal-justice-system-involved class (16.9%) whereas the class with the highest proportion of Hispanic women is the normative class (14.3%). As for family income, for adult men and women, the normative class has the smallest proportion of respondents with family income levels lower than $20,000 per year (men: 13.9%; women: 17.5%) and the highest proportion from families earning more than $75,000 per year (men: 37.0%; women 32.4%). Among men and women, the criminal-justice-system-involved class has the largest proportion of respondents living in families with income levels below $20,000 per year (men 31.9%; women 43.4%) and in households earning between $20,000 and $49,000 per year (men: 37.7%; women 35.9%). Across all three latent classes, the chi-square tests were significant (p < .001), suggesting that sociodemographic differences are observed across the three latent classes among men and women.

Table 2:

Sociodemographic Characteristics of Latent Classes of Men and Women Ages 18-64

	Class 1 (normative)		Class 2 (drug involved)		Class 3 (CJ system involved)
	%	95% CI	%	95% CI	%	95% CI	χ² Significance
Men
Age
18-25	14.68	[14.4, 14.9]	51.22	[49.5, 52.9]	43.16	[41.0, 45.4]	<.001
26-34	18.74	[18.3, 19.2]	23.83	[22.3, 25.5]	23.96	[21.9, 26.1]
35-64	66.57	[66.1, 67.1]	24.94	[23.1, 26.9]	32.88	[30.2, 35.7]
Race/ethnicity
White	66.70	[66.1, 67.3]	74.04	[72.5, 75.6]	57.70	[55.2, 60.1]	<.001
African American	10.84	[10.5, 11.2]	10.27	[9.2, 11.4]	20.78	[18.9, 22.8]
Hispanic	15.90	[15.4, 16.4]	11.53	[10.4, 12.7]	16.90	[14.9, 19.1]
Other	6.56	[6.2, 6.9]	4.16	[3.6, 4.8]	4.63	[3.8, 5.6]
Family income
< $20,000	13.88	[13.5, 14.3]	27.67	[26.3, 29.1]	31.93	[29.7, 34.2]	<.001
$20,000-$49,000	30.42	[29.9, 31.0]	33.78	[32.2, 35.4]	37.69	[35.3, 40.1]
$50,000-$74,000	18.71	[18.2, 19.2]	14.69	[13.5, 15.9]	13.80	[12.3, 15.5]
> $75,000	37.00	[36.4, 37.6]	23.86	[22.3, 25.4]	16.58	[14.8, 18.5]
Women
Age
18-25	14.72	[14.5, 15.0]	54.34	[52.6, 56.1]	40.60	[37.5, 43.7]	<.001
26-34	18.62	[18.2, 19.0]	20.82	[19.4, 22.4]	25.07	[22.0, 28.5]
35-64	66.66	[56.2, 67.1]	24.84	[23.0, 26.7]	34.32	[30.5, 38.3]
Race/ethnicity
White	65.59	[65.0, 66.1]	77.58	[76.2, 78.9]	62.94	[59.5, 66.2]	<.001
African American	13.10	[12.7, 13.5]	9.14	[8.2, 10.2]	20.35	[17.7, 23.2]
Hispanic	14.31	[13.9, 14.7]	8.91	[8.0, 9.9]	12.50	[10.4, 15.0]
Other	7.00	[.7, 7.3]	4.36	[3.8, 5.0]	4.20	[3.2, 5.4]
Family income
< $20,000	17.53	[17.1, 17.9]	31.32	[29.9, 32.8]	43.41	[39.9, 47.0]	<.001
$20,000-$49,000	31.48	[31.0, 32.0]	34.25	[32.7, 35.8]	35.94	[32.6, 39.4]
$50,000-$74,000	18.61	[18.2, 19.0]	13.35	[12.1, 14.6]	10.70	[8.6, 13.3]
> $75,000	32.39	[31.8, 32.9]	21.08	[19.7, 22.5]	9.95	[7.9, 12.4]

Health Characteristics of Latent Classes

Table 3 displays the associations between adult male and female sociodemographic and health characteristics and membership in the three latent behavioral classes. Male members of the drug-involved class were significantly less likely to be older (RR = 0.73, CI = 0.72-0.74) and significantly less likely to be African American (RR = 0.61, CI = 0.53-0.70), Hispanic (RR = 0.46, CI = 0.40-0.52), or “other” race (RR = 0.47, CI = 0.40-0.56) compared with the normative class. Male members of the criminal-justice-system-involved class were also significantly less likely to be older (RR = 0.78, CI = 0.77-0.80) and significantly less likely to be Hispanic (RR = 0.84, CI = 0.71-0.99) or “other” race (RR = 0.64, CI = 0.52-0.77) compared with the normative class; however, in contrast to the drug-involved class, members of the criminal-justice-system-involved class were significantly more likely to be African American (RR = 1.59, CI = 1.38-1.82). In terms of health characteristics, compared with the normative class, male members of the drug-involved class were significantly more likely to have received a diagnosis of asthma (RR = 1.18, CI = 1.05-1.32), hepatitis (RR = 2.04, CI = 1.42-2.95), pancreatitis (RR = 2.53, CI = 1.48-4.32), and a sexually transmitted disease (RR = 2.53, CI = 2.10-3.06). Notably, however, male members of the drug-involved class were also significantly less likely to have received diagnoses of diabetes (RR = 0.48, CI = 0.35-0.68) and heart disease (RR = 0.51, CI = 0.35-0.76), which may manifest in later ages. Male members of the criminal-justice-system-involved class were significantly more likely to have received a diagnosis of hepatitis (RR = 1.74, CI = 1.02-2.98), pancreatitis (RR = 2.32, CI = 1.05-5.15), and a sexually transmitted disease (RR = 1.50, CI = 1.15-1.95).

Table 3:

Health Characteristics of Men Ages 18-64 Across 3 Latent Classes^a

	Adult males (age 18-64 years)				Adult females (age 18-64 years)
	Class 2 (drug involved)		Class 3 (CJ system involved)		Class 2 (drug involved)			Class 3 (CJ system involved)
Variables	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Sociodemographic variables
Age	0.73	[0.72, 0.74]	0.78	[0.77, 0.80]	0.71	[0.70, 0.72]	0.79	[0.77, 0.81]
Race/ethnicity
African American	0.61	[0.53, 0.70]	1.59	[1.38, 1.82]	0.38	[0.33, 0.44]	0.96	[0.80, 1.16]
Hispanic	0.46	[0.40, 0.52]	0.84	[0.71, 0.99]	0.37	[0.32, 0.42]	0.60	[0.48, 0.75]
Other	0.47	[0.40, 0.56]	0.64	[0.52, 0.77]	0.45	[0.38, 0.52]	0.53	[0.39, 0.71]
Family income
< $20,000	2.32	[2.07, 2.60]	3.66	[3.09, 4.33]	2.11	[1.90, 2.35]	5.96	[4.49, 7.90]
$20,000-$49,000	1.65	[1.48, 1.84]	2.41	[2.06, 2.81]	1.60	[1.44, 1.78]	3.34	[2.54, 4.40]
$50,000-$74,000	1.16	[1.02, 1.32]	1.56	[1.30, 1.88]	1.06	[0.92, 1.22]	1.75	[1.25, 2.46]
Health characteristics
Asthma	1.18	[1.05, 1.32]	1.17	[0.99, 1.38]	1.11	[1.01, 1.23]	1.38	[1.12, 1.69]
Cirrhosis	1.64	[0.46, 5.77]	1.05	[0.41, 2.67]	4.07	[1.98, 8.36]	2.15	[0.77, 6.02]
Diabetes	0.48	[0.35, 0.68]	0.71	[0.49, 1.01]	0.55	[0.40, 0.76]	0.70	[0.45, 1.11]
Heart disease	0.51	[0.35, 0.76]	0.73	[0.48, 1.10]	0.61	[0.44, 0.86]	1.60	[0.82, 3.11]
Hepatitis	2.04	[1.42, 2.95]	1.74	[1.02, 2.98]	2.03	[1.41, 2.91]	2.91	[1.72, 4.91]
Lung cancer	1.21	[0.44, 3.39]	2.40	[0.62, 9.27]	0.48	[0.14, 1.63]	5.56	[1.07, 28.95]
Pancreatitis	2.53	[1.48, 4.32]	2.32	[1.05, 5.15]	1.53	[0.90, 2.62]	1.34	[0.49, 3.67]
Pneumonia	0.96	[0.79, 1.17]	0.87	[0.64, 1.18]	0.95	[0.79, 1.13]	0.96	[0.67, 1.38]
STD	2.53	[2.10, 3.06]	1.50	[1.15, 1.95]	3.03	[2.68, 3.44]	1.98	[1.56, 2.52]
Stroke	1.31	[0.54, 3.21]	0.65	[0.31, 1.36]	1.00	[0.52, 1.92]	0.76	[0.31, 1.85]
Tuberculosis	1.27	[0.66, 2.46]	0.75	[0.35, 1.59]	1.07	[0.52, 2.20]	0.77	[0.21, 2.83]
Stomach ulcers	1.18	[0.93, 1.50]	1.08	[0.71, 1.16]	1.14	[0.92, 1.41]	1.22	[0.83, 1.79]

Note. Risk ratios in bold are significant at p < .05 or lower. Reference = Class 1 (normative). STD = sexually transmitted disease.

Given the very large sample sizes, interpretation of results should be guided by effect sizes (RR) and not merely statistical significance.

As seen in Table 3, consistent with males, female members of the drug-involved class were significantly less likely to be older (RR = 0.71, CI = 0.70-0.72) and to be an ethnic or racial minority compared with members of the normative class. With one notable exception, a similar pattern could be observed among female members of the criminal-justice-system-involved class who were significantly likely to be younger (RR = 0.79, CI = 0.77-0.81) and to be Hispanic (RR = 0.60, CI = 0.48-0.75) or “other” race (RR = 0.53, CI = 0.39-0.71) but not African American. In terms of health characteristics and like their male counterparts, female members of the drug-involved class were significantly more likely to have received a diagnosis of asthma (RR = 1.11, CI = 1.01-1.23), cirrhosis (RR = 4.07, CI = 1.98-8.36), hepatitis (RR = 2.03, CI = 1.41-2.91), and a sexually transmitted disease (RR = 3.03, CI = 2.68-3.44). Similar to male members of the drug-involved class, female members were also significantly less likely to have received a diagnosis of diabetes (RR = 0.55, CI = 0.40-0.76) or heart disease (RR = 0.61, 0.44-0.86) than members of the normative class. Compared with the normative class, female members of the criminal-justice-system-involved class were significantly more likely to have received a diagnosis of asthma (RR = 1.38, CI = 1.21-1.69), hepatitis (RR = 2.91, CI = 1.72-4.91), lung cancer (RR = 5.56, CI = 1.07-28.95), and a sexually transmitted disease (RR = 1.98, CI = 1.56-2.52). In sum, empirical findings indicate a relatively similar pattern of health characteristics for male and female antisocial subgroups.

Mental Health and Chemical Dependence Characteristics of Latent Classes

Table 4 reveals the associations between membership in the three latent classes and mental health and chemical dependence outcomes. Male members of the drug-involved class were significantly less likely to be older (RR = 0.72, CI = 0.70-0.74) and significantly less likely to be an ethnic or racial minority compared with members of the normative class. Members of the criminal-justice-system-involved class were also significantly less likely to be older (RR = 0.78, CI = 0.74-0.81), but significantly more likely to be African American (RR = 1.83, CI = 1.39-2.42). In terms of family income, male members of the drug-involved and criminal-justice-system-involved engaged classes were more likely to report total family incomes of less than $50,000 per year. In terms of mental health, compared with the normative class, members of the drug-involved class were significantly more likely to have been diagnosed with depression (RR = 1.47, CI = 1.03-2.10) and a dual diagnosis (RR = 4.92, CI = 3.60-6.72), while members of the criminal-justice-system-involved class were significantly more likely to have received a dual diagnosis (RR = 3.67, CI = 2.49-5.41) and to report suicidal ideation (RR = 1.82, CI = 1.16-2.86). In terms of chemical dependence, members of the drug-involved and criminal-justice-system-involved classes were significantly more likely to meet criteria for nicotine or alcohol dependence compared with members of the normative class.

Table 4:

Mental Health Characteristics of Men Ages 18-64 Across 3 Latent Classes^a

	Adult males (age 18-64 years)				Adult females (age 18-64 years)
	Class 2 (drug involved)		Class 3 (CJ system involved)		Class 2 (drug involved)		Class 3 (CJ system involved)
Variables	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Sociodemographic variables
Age	0.72	[0.70, 0.74]	0.78	[0.74, 0.81]	0.71	[0.69, 0.73]	0.80	[0.76, 0.85]
Race/ethnicity
African American	0.63	[0.46, 0.86]	1.83	[1.39, 2.42]	0.37	[0.27, 0.51]	1.45	[0.94, 2.24]
Hispanic	0.58	[0.43, 0.80]	1.02	[0.65, 1.60]	0.46	[0.33, 0.64]	0.97	[0.60, 1.57]
Other	0.50	[0.33, 0.77]	0.74	[0.49, 1.14]	0.51	[0.35, 0.74]	0.54	[0.30, 0.98]
Family income
< $20,000	1.80	[1.40, 1.33]	1.80	[1.40, 2.33]	1.70	[1.32, 2.19]	3.64	[2.14, 6.17]
$20,000-$49,000	1.38	[1.08, 1.75]	1.38	[1.08, 1.75]	1.48	[1.41, 1.93]	2.36	[1.35, 4.15]
$50,000-$74,000	1.16	[0.87, 1.54]	1.16	[0.87, 1.54]	1.12	[0.80, 1.58]	1.32	[0.69, 2.53]
Mental health and chemical dependence
Depression	1.47	[1.03, 2.10]	1.06	[0.69, 1.63]	1.51	[1.15, 1.98]	2.16	[1.35, 2.44]
Anxiety	1.21	[0.85, 1.72]	1.37	[0.86, 2.16]	1.08	[0.81, 1.44]	1.11	[0.66, 1.88]
Dual diagnosis	4.92	[3.60, 6.72]	3.67	[2.49, 5.41]	8.97	[6.54, 12.32]	7.88	[4.29, 14.49]
Suicidal ideation	1.09	[0.78, 1.50]	1.82	[1.16, 2.86]	1.37	[0.99, 1.89]	1.30	[0.77, 2.19]
Nicotine dependence	2.76	[2.20, 3.47]	3.52	[2.69, 4.61]	1.77	[1.37, 2.29]	3.32	[2.26, 4.87]
Alcohol dependence	1.89	[1.33, 2.69]	2.87	[1.90, 4.31]	1.95	[1.27, 3.00]	1.81	[0.84, 3.88]

Note. Risk ratios in bold are significant at p < .05 or lower. Reference = Class 1 (normative).

Given the very large sample sizes, interpretation of results should be guided by effect sizes (RR) and not merely statistical significance.

A very similar pattern was observed among women in terms of the associations between membership in the latent behavioral classes and mental health and chemical dependence outcomes. Female members of the drug-involved class were significantly less likely to be older (RR = 0.71, CI = 0.69-0.73) and significantly less likely to be a member of an ethnic or racial minority group compared with the normative class. Female members of the criminal-justice-system-involved class were also significantly less likely to be older (RR = 0.80, CI = 0.76-0.85), but the only significant association observed in terms of race and ethnicity was that members of this class were significantly less likely to be of “other” race (RR = 0.54, CI = 0.30-0.98). Consistent with males, female members of the drug-involved and criminal-justice-system-involved classes were significantly more likely to report total family incomes of less than $50,000 per year compared with the normative class. In terms of mental health and chemical dependence, female members of the drug-involved class were significantly more likely to have received a diagnosis of depression (RR = 1.51, CI = 1.15-1.98), have a dual diagnosis (RR = 8.97, CI = 6.54-12.32), and to meet criteria for nicotine (RR = 1.77, CI = 1.37-2.29) and alcohol dependence (RR = 1.95, CI = 1.27-3.00). With the exception of alcohol dependence, a similar pattern was observed among female members of the criminal-justice-system-involved class who were significantly more likely to have received a diagnosis of depression (RR = 2.16, CI = 1.35-2.44), have a dual diagnosis (RR = 7.88, CI = 4.29-14.49), and meet criteria for nicotine dependence (RR = 3.32, CI = 2.26-4.87).

Discussion

Much of the literature investigating antisocial behavior has tended to focus on its antecedents and while this emphasis is undoubtedly important for theoretical and policy matters, recent scholarship has investigated the linkages and associations that offending has with health-related behaviors and adverse health outcomes (Akers & Lanier, 2009; A. R. Piquero et al., 2011; Reingle et al., 2013; Vaughn, 2011). Although some recent studies have used generalizable samples such as the Add Health (Reingle et al., 2013; Stogner et al., 2013), one of the limitations of prior investigations is their relegation to small, specific (mainly male, offender based) samples, thereby precluding more expansive analyses among large, nationally representative data sources with attention to gender differences. The current study performed such an investigation.

Overall, results show that drug-involved and criminal-justice-system-involved males and females are significantly more likely to have health and mental health problems that are associated with shorter term behavioral risk rather than chronic diseases that are more likely to occur over time and in older populations such as cardiovascular disease and diabetes. Other studies that have compared the health of offenders with nonoffenders have also found this pattern of relationships. For example, in a study that compared chronic health conditions among persons in prison and jail with the general population, Binswanger, Krueger, and Steiner (2009) found that correctional populations were at higher odds of hepatitis, but like our study not diabetes or cardiovascular disease, and a lower prevalence of obesity. However, this study used data on a confined population whereas the present study comparisons were made with criminal-justice-system-involved and drug-involved persons outside of the correctional system.

The pattern of associations between health and antisocial behaviors indicates similarities between drug-involved and criminal-justice-system-involved classes and similarities across sex. Also of note in the finding is the relatively low percentage of females (2.51%) compared with males (6.88%) in the criminal-justice-system-involved classes. Despite these similarities, there were several differences worth noting. For example, among women, lung cancer and cirrhosis reached large effect sizes, but these were nonsignificant in males. Furthermore, possession of a dual diagnosis was larger among females. Conversely, among criminal-justice-involved males suicide ideation and alcohol dependence were significant, but not in females. The dual diagnosis finding is consistent with psychiatric epidemiology investigations that have found that co-occurring disorders are common among populations who experience contact with the criminal justice system (Teplin, Abram, McClelland, Dulcan, & Mericle, 2002; Vaughn et al., 2012; Vaughn, Freedenthal, Jenson, & Howard, 2007).

There are several implications that are directly tied to the present study findings. First, the current study indicates that persons with criminal justice system contact are linked to Objective 3 of the Second Chance Act of 2007, “to encourage the development and support of, and to expand the availability of, evidence-based programs that enhance public safety and reduce recidivism, such as substance abuse treatment, alternatives to incarceration, and comprehensive reentry services” (United States. The House of Representatives, 2008, p. 658). Relatedly, as noted by other scholars (Grisso, 2004; Robins et al., 2010), the Department of Juvenile Justice has a “moral and legal obligation to provide for the medical needs of adolescents in their care” (Odgers et al., 2010, p. 429). Thus, forging a stronger connection between health systems and the criminal justice system is needed to not only achieve public safety but also provide cost-effective treatments to advance public health. Criminal justice and public health systems can collaborate for mutual benefit.

Second, given the strong effects found for males and females with respect to co-occurring mental health and substance use disorder diagnoses, effective modalities that have been experimentally validated would be useful for both antisocial latent classes. Modified therapeutic communities are one such modality that can be delivered in community or prison settings. In one recent randomized clinical trial testing this approach, offenders who were treated were substantially less likely (19% compared with 38%) to be arrested over a 12-month evaluation period (Sacks, Chaple, Sacks, McKendrick, & Cleland, 2012). Third, findings suggest that gender-specific health practices may not be universally needed with antisocial behavior and criminal justice populations. General evidence-supported prevention and treatment protocols targeting behaviorally oriented health risks, such as sexually transmitted diseases and hepatitis, may be applied with equal effectiveness for females and males. Yet, additional research should be undertaken because there may be some nuanced differences regarding the differential exposure of risk factors across males and females—especially during the childhood and adolescent years.

The small but seemingly active antisocial subgroups found in the three-class solution evidenced by a large normative group and a criminally involved pathological group that is approximately 5% of the sample among males and females is consonant with criminal career models. Studies have repeatedly identified a small pathological group who are characterized by pervasive antisocial conduct including substance abuse (DeLisi, 2005; DeLisi & Piquero, 2011; Jennings & Reingle, 2012; Vaughn, DeLisi, et al., 2011; Welsh & Farrington, 2012). The behavioral problems of this group similarly contribute to a host of adjustment problems (e.g., relationship problems, frequent unemployment, interpersonal conflicts) that together negatively affect an individual’s mental and physical health.

Study Assets and Limitations

Among the assets of this study was the unique scope of the study sample, which involved well over 100,000 nationally representative participants. This sheer sample size enabled us to empirically explore subgroups without compromising statistical power. In addition, the data contained multiple measures of antisocial behaviors and health conditions. Despite these strengths, findings from this study should be interpreted in light of several limitations. First and foremost among them is that study data—a series of large cross-sections—are not temporally ordered and causal conclusions regarding the relationship between antisocial behavior and health cannot be drawn. Thus, it should be kept in mind that results are strictly associational and correlational in nature. An additional limitation is the survey instrument and methods of data collection that relies on self-reporting and retrospective recall. Finally, the theorized relationship between antisocial behavior and health whether consistent with self-control or allostatic load theories is suggestive rather than direct due to a lack of explicit assessments of these constructs. Whereas prior studies have directly shown that low self-control is associated with diverse health risks including victimization (C. Gibson, Schreck, & Miller, 2004; Higgins, Jennings, Tewksbury, & Gibson, 2009), the current investigators were unable to directly assess this relationship.

Conclusion

To our knowledge, this is the largest, nationally representative sample ever utilized for the study of the association between antisocial behavior and health. Two major findings are noteworthy. First, chronic health conditions that are more closely associated with longer term medical problems and perhaps cumulative stress such as heart disease and diabetes are not linked to criminally engaged or drug-involved offenders. However, medical problems that are more closely related to the immediate behaviors aligned with an antisocial lifestyle such as sexually transmitted diseases, pancreatitis, and hepatitis were found to be more prevalent among antisocial classes in this sample. Second, the relative pattern of health and antisocial behavior is fairly similar for males and females, though the proportion of males in the criminally active class is substantially larger. Results of this investigation suggest that substantially greater research and preventive and treatment attention pertaining to the overlap between behavioral health and antisocial behavior are badly needed.

Footnotes

Notes

Michael G. Vaughn, PhD, is a professor in the School of Social Work, College for Public Health and Social Justice at Saint Louis University. He has contributed more than 200 scholarly publications across a wide range of areas. His research interests include biosocial criminology, youth violence, adolescent psychopathy, and drug use and abuse in relation to antisocial behavior over the life course. Current projects include a cell-to-society approach to the study of human behavior, youth violence prevention, school dropout, and drug use epidemiology.

Christopher P. Salas-Wright, PhD, is an assistant professor in the School of Social Work at the University of Texas at Austin and a faculty affiliate at the Lozano Long Institute of Latin American Studies (LLILAS) at the University of Texas at Austin. His research interests include youth problem behavior, religiosity and spirituality, and adolescent substance abuse and HIV prevention, particularly among Latino youth in the United States and in Latin America.

Matt DeLisi, PhD, is a professor in the Department of Sociology, coordinator of Criminal Justice Studies, and affiliate with the Center for the Study of Violence at Iowa State University. He is a Fellow of the Academy of Criminal Justice Sciences. His research interests include biosocial criminology and criminal careers, and he has published more than 200 scholarly books, chapters, and articles in a range of social and behavioral science journals.

Alex R. Piquero, PhD, is Ashbel Smith Professor of Criminology at the University of Texas at Dallas; adjunct professor, Key Centre for Ethics, Law, Justice, and Governance, Griffith University Australia; faculty affiliate, Center for Violence and Injury Prevention George Warren Brown School of Social Work Washington University in St. Louis; and coeditor, Journal of Quantitative Criminology. His research interests include criminological theory, criminal careers, and quantitative research methods. He has received several research, teaching, and service awards and is fellow of the American Society of Criminology and the Academy of Criminal Justice Sciences.

References

Agnew

(1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30, 47-87.

Agnew

(2006). Pressured into crime: An overview of general strain theory. Los Angeles, CA: Roxbury.

Akers

T. A.

Lanier

(2009). Epidemiological criminology: Coming full circle. American Journal of Public Health, 99, 397-402.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (DSM-IV) (4th ed.). Washington, DC: Author.

Binswanger

I. A.

Krueger

P. M.

Steiner

J. F.

(2009). Prevalence of chronic medical conditions among jail and prison inmates in the USA compared with the general population. Journal of Epidemiology & Community Health, 63, 912-919.

Bond

B. J.

Gittell

J. H.

(2010). Cross-agency coordination of offender reentry: Testing collaboration outcomes. Journal of Criminal Justice, 38, 118-129.

Broidy

Agnew

(1997). Gender and crime: A general strain theory perspective. Journal of Research in Crime & Delinquency, 34, 275-306.

Chapple

C. L.

Vaske

Hope

T. L.

(2010). Sex differences in the causes of self-control: An examination of mediation, moderation, and gendered etiologies. Journal of Criminal Justice, 38, 1122-1131.

Chassin

Piquero

A. R.

Losoya

S. H.

Mansion

A. D.

Schubert

C. A.

(2013). Joint consideration of distal and proximal predictors of premature death among serious juvenile offenders. Journal of Adolescent Health, 52, 689-696.

10.

Cohen

(1998). The monetary value of saving a high risk youth. Journal of Quantitative Criminology, 14, 5-33.

11.

Cohen

Piquero

A. R.

(2009). New evidence on the monetary value of saving a high risk youth. Journal of Quantitative Criminology, 25, 25-49.

12.

Cross

C. P.

Copping

L. T.

Campbell

(2011). Sex differences in impulsivity: A meta-analysis. Psychological Bulletin, 137, 97-130.

13.

DeLisi

(2005). Career criminals in society. Thousand Oaks, CA: SAGE.

14.

DeLisi

(2013). Criminal psychology. San Diego, CA: Bridgepoint Education.

15.

DeLisi

Gatling

(2003). Who pays for a life of crime? An empirical assessment of the assorted victimization costs posed by career criminals. Criminal Justice Studies, 16, 283-293.

16.

DeLisi

Piquero

A. R.

(2011). New frontiers in criminal careers research, 2000-2011: A state-of-the-art review. Journal of Criminal Justice, 39, 289-301.

17.

DeLongis

Folkman

Lazarus

R. S.

(1988). The impact of daily stress on health and mood: Psychological and social resources and mediators. Journal of Personality and Social Psychology, 54, 486-495.

18.

de Ridder

D. T.

Lensvelt-Mulders

Finkenauer

Stok

F. M.

Baumeister

R. F.

(2012). Taking stock of self-control: A meta-analysis of how trait self-control relates to a wide range of behaviors. Personality and Social Psychology Review, 16, 76-99.

19.

Eme

R. F.

(2007). Sex differences in child-onset, life-course-persistent conduct disorder: A review of biological influences. Clinical Psychology Review, 27, 607-627.

20.

Eme

R. F.

(2010). Male life-course-persistent antisocial behavior. Archives of Pediatrics and Adolescent Medicine, 164, 486-487.

21.

Fagerstrom

K. O.

(1978). Measuring degree of physical dependence to tobacco smoking with reference to individualization of treatment. Addictive Behaviors, 3, 235-241.

22.

Farrington

D. P.

(1995). Crime and physical health: Illness, injuries, accidents, and offending in the Cambridge study. Criminal Behaviour and Mental Health, 5, 261-278.

23.

Gibson

C. L.

Ward

J. T.

Wright

J. P.

Beaver

K. M.

DeLisi

(2010). Where does gender fit in the measurement of self-control? Criminal Justice and Behavior, 37, 883-903.

24.

Gibson

Schreck

C. J.

Miller

J. M.

(2004). Binge drinking and negative alcohol-related behaviors: A test of self-control theory. Journal of Criminal Justice, 32, 411-420.

25.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

26.

Grisso

(2004). Double jeopardy: Adolescent offenders with mental disorders. Chicago, IL: University of Chicago Press.

27.

Hedden

S. L.

Gfroerer

J. C.

(2011). Correlates of perceiving a need for treatment among adults with substance use disorder: Results from a National Survey. Addictive Behaviors, 36, 1213-1222.

28.

Higgins

G. E.

Jennings

W. G.

Tewksbury

Gibson

C. L.

(2009). Exploring the link between low self-control and violent victimization trajectories in adolescents. Criminal Justice and Behavior, 36, 1070-1084.

29.

Jennings

W. G.

Reingle

J. M.

(2012). On the number and shape of developmental/life-course violence, aggression, and delinquency trajectories: A state-of-the-art review. Journal of Criminal Justice, 40, 472-489.

30.

Long

J. S.

Freese

(2006). Regression models for categorical dependent variables using stata (2nd ed.). College Station, TX: StataCorp Press.

31.

Lykken

D. T.

(1995). The antisocial personalities. Hillsdale, NJ: Erlbaum.

32.

McEwen

(1998). Protective and damaging effects of stress mediators. New England Journal of Medicine, 338, 171-179.

33.

McLachlan

Peel

(2000). Finite mixture models. New York, NY: John Wiley.

34.

Mezuk

Rafferty

J. A.

Kershaw

K. N.

Hudson

Abdou

C. M.

Lee

Jackson

J. S.

(2010). Reconsidering the role of social disadvantage in physical and mental health: Stressful life events, health behaviors, race and depression. American Journal of Epidemiology, 172, 238-249.

35.

Moffitt

T. E.

Arsenault

Belsky

Dickson

Hancox

R. J.

Harrington

Caspi

(2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proceedings of the National Academy of Sciences of the United States of America, 108, 2693-2698.

36.

Nakawaki

Crano

W. D.

(2012). Predicting adolescents’ persistence, non-persistence, and recent onset of nonmedical use of opioids and stimulants. Addictive Behaviors, 37, 716-721.

37.

Odgers

C. L.

Caspi

Broadbent

J. M.

Dickson

Hancox

R. J.

Harrington

H. L.

Moffitt

T. E.

(2007). Conduct problems subtypes in males predict differential adult health burden. Archives of General Psychiatry, 64, 476-484.

38.

Odgers

C. L.

Robins

S. J.

Russell

M. A.

(2010). Morbidity and mortality risk among the “forgotten few”: Why are girls in the justice system in such poor health? Law and Human Behavior, 34, 429-444.

39.

Piquero

A. R.

Daigle

L. E.

Gibson

Leeper Piquero

Tibbetts

S. G.

(2007). Are life-course-persistent offenders at risk for adverse health outcomes? Journal of Research in Crime & Delinquency, 44, 185-207.

40.

Piquero

A. R.

Farrington

D. P.

Nagin

D. S.

Moffitt

T. E.

(2010). Trajectories of offending and their relation to life failure in late middle age: Findings from the Cambridge study in delinquent development. Journal of Research in Crime & Delinquency, 47, 151-173.

41.

Piquero

A. R.

Farrington

D. P.

Shepherd

J. P.

Auty

(2013). Offending and early death in the Cambridge Study in delinquent development. Justice Quarterly. Advance online publication. doi:10.1080/07418825.2011.641027

42.

Piquero

A. R.

Gibson

Tibbetts

(2002). Does low self-control account for the correlation between binge-drinking and alcohol-related behaviors. Criminal Behaviour and Mental Health, 12, 135-154.

43.

Piquero

A. R.

Shepherd

J. P.

Farrington

D. P.

(2011). Impact of offending trajectories on health: Disability, hospitalization, and death in middle-aged men in the Cambridge study in delinquent development. Criminal Behaviour and Mental Health, 21, 189-201.

44.

Piquero

N. L.

Sealock

M. D.

(2004). Gender and general strain theory: A preliminary test of Broidy and Agnew’s gender/GST hypotheses. Justice Quarterly, 21, 125-158.

45.

Reingle

J. M.

Jennings

W. G.

Piquero

A. R.

Maldonado-Molina

M. M.

(2013). Is violence bad for your health? An assessment of chronic disease outcomes in a nationally representative sample. Justice Quarterly. Advance online publication. doi:10.1080/07418825.2012.689315

46.

Repetti

R. L.

Taylor

S. E.

Seeman

T. E.

(2002). Risky families: Family social environments and the mental and physical health of offspring. Psychological Bulletin, 128, 330-366.

47.

Robins

S. J.

Odgers

C. L.

Russell

M. A.

(2010). Incarcerated girls’ physical health: Can the juvenile justice system help to reduce long-term health costs? Court Review, 46, 30-35.

48.

Sacks

Chaple

Sacks

J. Y.

McKendrick

Cleland

C. M.

(2012). Randomized trial of a reentry modified therapeutic community for offenders with co-occurring disorders: Crime outcomes. Journal of Substance Abuse Treatment, 42, 247-259.

49.

Schroeder

R. D.

Hill

T. D.

Haynes

S. H.

Bradley

(2011). Physical health and crime among low-income urban women: An application of general strain theory. Journal of Criminal Justice, 39, 21-29.

50.

Shepherd

Farrington

D. P.

Potts

(2004). Impact of antisocial lifestyle on health. Journal of Public Health, 26, 347-352.

51.

Shiffman

Waters

A. J.

Hickcox

(2004). The Nicotine Dependence Syndrome Scale: A multi-dimensional measure of nicotine dependence. Nicotine & Tobacco Research, 6, 327-348.

52.

StataCorp. (2011). Stata statistical software: Release 12. College Station, TX: Author.

53.

Stogner

Gibson

C. L.

(2010). Healthy, wealthy, and wise: Incorporating health issues as a source of strain in Agnew’s general strain theory. Journal of Criminal Justice, 38, 1150-1159.

54.

Stogner

Gibson

C. L.

Miller

J. M.

(2013). Examining the reciprocal nature of the health-violence relationship: Results from a nationally representative sample. Justice Quarterly. Advance online publication. doi:10.1080/07418825.2012.723029

55.

Substance Abuse and Mental Health Services Administration (SAMHSA). (2011). Results from the 2010 National Survey on Drug Use and Health: Summary of national findings. Rockville, MD: Author.

56.

Teplin

L. A.

Abram

K. M.

McClelland

G. M.

Dulcan

M. K.

Mericle

A. A.

(2002). Psychiatric disorders in youth in juvenile detention. Archives of General Psychiatry, 59, 1133-1143.

57.

U.S. House of Representatives. (2008, March 11). 110th Congress, 2nd Session, An Act to reauthorize the grant program for reentry of offenders into the community in the Omnibus Crime Control and Safe Streets Act of 1968, to improve reentry planning and implementation, and for other purposes [passed by the U.S. House of Representatives; 11 March 2008]. 110th Cong., 2nd sess. Congressional Bills. Retrieved from http://www.govtrack.us/congress/billtext.xpd?bill=h110-1593

58.

Vaughn

M. G.

(2011). Addiction and criminal justice: Empirical findings and theory for advancing public safety and health. Journal of Criminal Justice, 39, 205-206.

59.

Vaughn

M. G.

Beaver

K. M.

DeLisi

Wright

J. P.

(2009). Identifying latent classes of behavioral risk based on early childhood manifestations of self-control. Youth Violence and Juvenile Justice, 7, 16-31.

60.

Vaughn

M. G.

DeLisi

Gunter

Beaver

K. M.

Perron

B. E.

Howard

M. O.

(2011). The severe 5%: A latent class analysis of the externalizing spectrum in the United States. Journal of Criminal Justice, 39, 75-80.

61.

Vaughn

M. G.

DeLisi

Perron

B. E.

Beaver

K. M.

Abdon

(2012). Toward a criminal justice epidemiology: Behavioral and physical health of probationers and parolees in the United States. Journal of Criminal Justice, 40, 165-173.

62.

Vaughn

M. G.

Freedenthal

Jenson

J. M.

Howard

M. O.

(2007). Psychiatric symptoms and substance use among juvenile offenders: A latent profile investigation. Criminal Justice and Behavior, 34, 1296-1312.

63.

Vaughn

M. G.

Wernet

S. J.

DeLisi

Beaver

K. M.

Perron

B. E.

Howard

M. O.

(2011). Characteristics of abstainers from substance use and antisocial behavior in the United States. Journal of Criminal Justice, 39, 212-217.

64.

Vaughn

M. G.

Howard

M. O.

(2004). Adolescent substance abuse treatment: A synthesis of controlled evaluations. Research on Social Work Practice, 14, 325-335.

65.

Vermunt

J. K.

Magidson

(2008). LG-syntax user’s guide: Manual for Latent GOLD 4.5 syntax module. Belmont, MA: Statistical Innovations.

66.

Welsh

B. C.

Farrington

D. P.

(2012). Science, politics, and crime prevention: Toward a new crime policy. Journal of Criminal Justice, 40, 128-133.

67.

World Health Organization. (2012). The World Health Organization disability assessment schedule. Retrieved from http://www.who.int/classifications/icf/whodasii/en/index.html

68.

Yang

(2006). Evaluating latent class analyses in qualitative phenotype identification. Computational Statistics & Data Analysis, 50, 1090-1104.

69.

Zhang

(1998). What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. Journal of the American Medical Association, 280, 1690-1691.