The Roles of Antisociality and Neurodevelopmental Problems in Criminal Violence and Clinical Outcomes Among Male Forensic Inpatients

Abstract

Research has identified developmental pathways linking childhood abuse and poor parenting to cognitive impairment or general antisociality in adolescence. We examined pathways in adult offenders, aiming to reproduce pathways identified in the limited existing research with male forensic inpatients and testing their robustness and relation to clinical outcomes. Using structural equation modeling (SEM), we examined antisociality, neurodevelopmental problems, and antisocial parenting as statistical predictors of criminal violence, in 638 cases. Using exploratory factor analysis and SEM, we developed and tested similar models for health status on admission and institutional outcomes (symptoms and aggressive behaviors) in subsamples of 269 and 335 cases. A three-factor measurement model was obtained, replicating previous work and supporting antisociality and neurodevelopmental problems as pathways to criminal violence and poor adult health. Models of institutional outcomes were not well supported. Findings indicate robustness of antisocial development and neurodevelopmental pathways.

Keywords

pathways offending forensic antisociality neurodevelopmental

Developmental pathways to crime describe a lifetime sequence of events and characteristics, delineate the steps toward an outcome, and outline a process by which risk (and protective) factors might be linked to the outcome (e.g., Day, Wanklyn, & Yessine, 2014; Kazdin, Kraemer, Kessler, Kupfer, & Offord, 1997; Loeber & Burke, 2011). Research into developmental pathways among offenders helps identify antecedents and mechanisms that may influence criminal offending (e.g., Kazdin et al., 2007). Pathways research has implications for not only etiology and prognosis but also treatment (e.g., Moffitt, 1993). That is, for a population with any given outcome, there can be evidence for multiple pathways and different interventions may be needed to target the problems faced by a subtype of offenders—a group of individuals following one path rather than another (e.g., Kazdin et al., 1997; Moffitt & Caspi, 2001; Simpson, Grimbos, Chan, & Penney, 2015). Factor analysis or other structural statistical techniques that identify associations among variables indicative of subtypes can be interpreted as one form of evidence for theorized developmental pathways among offenders (e.g., Harris, Rice, & Lalumière, 2001; Jones, Brown, Wanamaker, & Greiner, 2014). In the present study, we examined two developmental pathways previously identified among adult offenders in secure forensic care—antisociality and neurodevelopmental problems—and their relation to criminal violence. We also tested the generalizability of these constructs, and their clinical relevance, by examining their relation to adult health and to clinical symptoms and institutional behaviors.

Developmental Pathways to Offending

Sheldon and Eleanor Glueck first identified factors among adolescents leading to chronic offending, although long-term follow-up suggests the process of desistance from crime might weaken the explanatory power of adolescent-identified trajectories over the course of adulthood (e.g., Sampson & Laub, 2003). Patterson, DeBaryshe, and Ramsay (1989) and Moffitt (1993) further proposed that different youth follow distinct hypothesized pathways to delinquency. Since then, there has been extensive research consistently identifying an early onset, life-course-persistent offender type, often distinguished from youth with late onset and adolescent-limited offending. The life-course-persistent pathway appears to reflect a unique etiology of offending and, compared with the adolescent-limited pathway, may be characterized by more risk factors, increased recidivism, criminal versatility, and relative unresponsiveness to judicial interventions (e.g., Augustyn, 2015; Nisbet, Smallbone, & Wortley, 2010). In the life-course-persistent pathway, early risk factors such as perinatal problems and early childhood abuse are thought to have broad detrimental effects that increase the risk of poor adolescent adjustment, which in turn is a risk factor for adult criminal behavior (e.g., Day et al., 2014). Childhood physical abuse increases the risk of neurological impairment (e.g., Teicher & Samson, 2016). In addition, early onset delinquent youth report more exposure to “dysfunctional” parenting, including poor supervision, inconsistent discipline, and corporal punishment (e.g., Dandreau & Frick, 2009).

Among adolescent offenders on an early onset trajectory, further pathways have been distinguished that differ according to affective and interpersonal styles; that is, on the basis of the presence of callous and unemotional traits (e.g., Frick et al., 2003). For example, research into pathways among youth who commit sex offenses includes those exhibiting sexual deviance and impaired cognitive functioning and those identified as following a more antisocial path of social deviance and general delinquency (Hunter, Figueredo, & Malamuth, 2010). Among male youth, the antisocial pathway appears more common, whereas females’ development is characterized by disadvantages such as abuse, poverty, and mental illness, as well as disruptive behaviors (e.g., Jones et al., 2014). This line of research has contributed to knowledge of the etiology of criminal behavior, especially with respect to the relative importance of early behavioral and adverse psychosocial events. It has also informed the development and evaluation of early interventions for delinquent youth (e.g., Stickle & Frick, 2002).

Thus, research with male adolescent offenders has identified possible pathways stemming from early childhood abuse and poor parenting, and expressed in cognitive impairment or general antisociality. Recent studies that followed young offenders into adulthood identified antisocial behavior and traits, and psychopathy in particular, as a factor in several pathways to offending, especially chronic violent offending (e.g., Corrado, McCuisha, Hart, & DeLisi, 2015; Diamantopoulou, Verhulst, & van der Ende, 2010; Piquero et al., 2012). In turn, a broad measure of problems in the familial environment that included harsh discipline, parental criminal record, and 13 other variables was found to be related to psychopathy (e.g., Piquero et al., 2012), although specific adverse early childhood events, such as physical and sexual abuse, did not distinguish among offender types (e.g., Corrado et al., 2015). There have been calls to converge the criminological study of offending careers with concepts from psychopathy research among adult offenders (e.g., Fox, Jennings, & Farrington, 2015). However, pathways research has not yet been well connected with the study of adult offenders. Whereas extensive research among adolescents has elucidated distinct risk factors associated with different trajectories of offending, the present study examined developmental pathways identifiable in adulthood.

The Study of Developmental Pathways Among Adult Offenders

The presentation of risk factors in adulthood might differ from that in adolescence because of social milestones and neurocognitive maturation (e.g., Corrado et al., 2015). Research with adult populations can increase our understanding of not only the etiologies of chronic criminal offending among adults but also the interventions that might be more or less beneficial for different offender types. For example, adult offenders might represent those who are difficult to treat, for whom early intervention was unsuccessful, or those exposed to new risk factors in late adolescence or emerging adulthood, such as the onset of mental disorder.

Whereas there is an extensive body of research validating the construct of psychopathy among offenders and its substantial contribution to violence risk assessment (e.g., Harris, Rice, Quinsey, & Cormier, 2015; Skilling, Harris, Rice, & Quinsey, 2002), there remains controversy as to whether development that is characterized by persistent antisocial behavior indicates a dysfunctional disorder involving neurological deficits or whether it can be theoretically and empirically distinguished from a developmental pathway marked by neurodevelopmental problems and adult psychopathology (e.g., Santana, 2016). In a study of 868 adult offenders assessed in a secure forensic psychiatric hospital in 1975-1981, researchers found evidence for distinct pathways to persistent criminal violence (Harris et al., 2001). Harris and colleagues (2001) drew information from the clinical and criminal histories contained in patients’ medical records and also obtained criminal records for offenses taking place after release from hospital. Using structural equation modeling (SEM), they identified two independent paths indicative of violent offender subtypes. One subtype had been on a pathway they labeled Psychopathy, characterized by antisocial development since childhood, and the other subtype had suffered problems since birth and infancy indicative of neurodevelopmental damage, which they labeled Neurodevelopmental Insults. The two paths were unrelated and independently led to violent offending in adulthood. The paths were also both linked with what Harris and colleagues labeled antisocial parenting: childhood physical abuse, psychological abuse or neglect, witnessing family violence, parental criminality, and parental substance abuse, events which have also been collectively called adverse childhood experiences (e.g., Finkelhor, Shattuck, Turner, & Hamby, 2015). The two theoretical paths to adult offending bear different treatment implications. For example, Harris and colleagues argued that violence perpetrated by offenders on the psychopathy pathway would not be considered pathological; rather, it is more likely to be instrumental than violence committed by offenders with neurodevelopmental damage and would perhaps be more responsive to contingencies of the institutional environment.

Harris and colleagues (2001) offered evidence that separate types of violent offenders could be distinguished from each other by, among other variables, the extent of neurological impairment in their developmental pathways. Nestor, Kimble, Berman, and Haycock (2002) obtained findings broadly consistent with this conclusion in their study of 26 homicide offenders in a secure forensic hospital. They reported that cluster analysis of scores on psychological assessments of cognitive functioning and psychopathy distinguished two independent groups which they termed psychotic and psychopathic. Subsequently, Gilligan and Lennings (2010) reported similar findings among 26 homicide or attempted homicide offenders undergoing forensic assessment using a decision tree analysis of a larger set of cognitive and neuropsychological tests and offender risk assessments. The two offender groups, labeled psychopathic and neuropathic, differed according to frontal executive and temporal lobe functioning, with psychopathic offenders overall showing no impairment in the absence of antecedent factors such as antisocial parenting. These two small studies have shed light on some of the clinical implications of pathways to violent offending.

The Relation of Developmental Pathways to Clinical Outcomes

There has been little subsequent work on developmental pathways among adult offenders or their potential clinical implications. In other research on institutional treatment, forensic patients have been identified as being at risk of physical health problems (e.g., Hilton, Ham, Lang, & Harris, 2015). They are thought to be especially vulnerable to ill health due to extensive time in secure custody, which limits their ability to exercise or make healthy food choices (e.g., Hui, Middleton, & Völlm, 2013). There is some evidence that cardiovascular and other health problems can be predicted from early childhood variables, just as criminal behavior can (Moffitt et al., 2011), suggesting that offenders’ physical health may be related to their developmental pathway. This outcome might be more likely among offenders on a neurodevelopmental damage pathway rather than a predominantly antisocial pathway, because of the link between neurodevelopmental damage and serious mental illness, now widely accepted based on epidemiological, imaging, and genetic research (e.g., Rapoport, Giedd, & Gogtay, 2012).

Connecting the work on clinical outcomes with research on developmental pathways could benefit both theory and practice. Theoretically, a finding that antisocial adult offenders can be distinguished by a lack of neurodevelopmental problems and adult health problems would support the concept of psychopathy as a nonpathological condition. Clinically, such research could inform early identification of adult offenders at risk of poorer clinical outcome such as weight gain and related problems. Replication of the Harris and colleagues (2001) two-pathway model with an outcome other than criminal violence would provide additional evidence for the hypothesized paths among adults.

The Present Study

There has been limited further work on the roles of antisocial development and neurodevelopmental damage as pathways to criminal violence among adult offenders. Some evidence that adolescent-identified pathways lose explanatory power over the course of adulthood due to desistance (Sampson & Laub, 2003) indicates the need to study this phenomenon in adults rather than relying on extrapolation from adolescent research, and the general climate of concern about the reproducibility of much psychological research (e.g., Open Science Collaboration, 2015) increases the need for replication. In the present study, we examined developmental pathways among men recently admitted to an inpatient psychiatric program for forensic assessment.

Our first objective was an attempt to reproduce, in a different sample, the Harris and colleagues (2001) model showing the association of antisocial and neurodevelopmental measures with a constellation of adverse childhood events indicative of antisocial parenting, and their independent pathways to criminal violence as measured by preadmission charges for violent offenses. We hypothesized the following:

Hypothesis 1: Measures of antisocial behavior since childhood and neurodevelopmental problems would yield separate factors and independent pathways to criminal violence.

Our second objective was to examine the robustness and potential clinical value of these pathways by testing their relation to health status on admission and to symptoms and behaviors in the institution. We tested a model of health as a way to confirm evidence for distinct pathways of antisociality and neurodevelopmental problems detectable in adulthood. We hypothesized the following:

Hypothesis 2: Neurodevelopmental problems would be associated with poorer adult health on admission, and antisociality would be associated with relatively good health.

We tested a model that included measures of clinical concern in the forensic institution, in terms of symptoms of mental illness and institutional assaults and threats, in order to advance understanding of the implications of developmental pathways for the clinical treatment of forensic patients. We hypothesized the following:

Hypothesis 3: Antisociality would be associated with aggressive behavior, and neurodevelopmental problems would be associated with mental illness.

Method

Cases

The sample comprised patients admitted to a forensic assessment program in the Province of Ontario, Canada, from January 2009 to December 2012. The total sample was 638 inpatients, including 291 previously reported in an analysis of weight gain during hospitalization (Hilton et al., 2015). All were male and most were admitted from court under an assessment order for fitness to stand trial (n = 253, 40%) or eligibility for a defense of Not Criminally Responsible (NCR) on account of mental disorder (n = 151, 44%); the remainder were transferred from other institutions or directly from court having already been found unfit or NCR. The mean length of stay, measured at March 2013, was 152 days (SD = 261) and more than 80% were discharged within 1 year.

Procedure

The authors’ institutional research ethics board reviewed and approved the use of medical records without participant consent, based on the grounds for waiver of consent in the Tri-Council Policy Statement-2 (Canadian Institute of Health Research, Natural Sciences and Engineering Research Council of Canada, and Social Sciences and Humanities Research Council of Canada, 2014). All variables were coded from hospital records that contained clinical information from the current and any previous admissions, including a psychosocial report completed by social workers who typically gathered information about the patients’ family, childhood, and medical and psychiatric history. Records also contained information from other sources such as previous adult or adolescent assessments and the record of all criminal charges.

All variables except those pertaining to institutional outcomes were coded shortly after admission. Most cases were coded by a research psychometrist with extensive experience reliably quantifying the information in these records. Approximately one third of cases were coded by the second author after training and interrater reliability tests with the main coder in masked coding of 10 cases. A small number of cases were coded by a research student, also following training and reliability tests, under close supervision by the main coder. Reliability coefficients are reported below with each variable. Variables yielding reliability coefficients below r = .80 were modified to improve agreement and occasionally reviewed for consensus. Institutional outcome variables were coded at least 1 year after admission. Most cases were coded by a research psychometrist and approximately one fifth were coded by one of three research students.

Variables

We coded variables intended to replicate as much as possible the Harris et al. (2001) measures of psychopathy, neurodevelopmental problems, and antisocial parenting. Most variables were coded zero if there was no clear evidence for meeting the criteria, with exceptions reported per variable below.

Neurodevelopmental Problems

Exposure to prenatal, perinatal, and early childhood problems has a well-established association with risk of neurodevelopmental impairments in laboratory and naturalistic research, including risk of mental illness (e.g., Branchi & Cirulli, 2014; Mwaniki, Atieno, Lawn, & Newton, 2012; Rapoport et al., 2012). As in previous research, we sampled a range of conditions and counted the number of conditions exposed to in order to construct a sensitive proxy measure of neurodevelopmental problems.

Obstetrical complications were coded for their presence or absence in the medical record as in Harris et al. (2001), adding one point for each of the following: maternal malnutrition, toxemia, exposure to toxic substances, high fever, injury from accident or abuse, prescribed medications, alcohol or street drug use, addiction to alcohol or drugs, extreme psychological stress or abuse, mental illness, or other pregnancy difficulties reported for the mother during pregnancy; low birth weight, forceps delivery, cesarean section, prolonged labor or delivery, abnormal fetal position, umbilical or placental abnormality, asphyxia, anoxia, fetal distress, premature birth, Rh blood factor incompatibility, infections, or other perinatal problems. The possible range for the obstetrical complications score was 0 to 23. Where the medical record indicated no perinatal or birth problems, based on a reliable source, this item was scored 0. Interrater reliability in the present study was r = .95 (student coder r = .83).

Childhood health problems were coded from the medical record as Harris et al. (2001) did for infancy problems, with the exception that, because these problems were previously rarely reported, we extended the age range from early childhood up to age 15 years. We added one point for each of the following: colic, serious childhood disease, high fever, serious infection, asthma, tuberculosis, serious allergies, hearing problems, serious headaches, blurred vision, stuttering, fainting or dizziness, head injury with loss of consciousness, head injury with dizziness, or other serious childhood illnesses and physical problems. The possible range for the childhood health problems score was 0 to 16. Interrater reliability in the present study was r = .86 (student coder r = .93).

We coded the highest educational grade achieved as in Harris et al. (2001), including an additional point for each year of education beyond Grade 12 high school graduation. We multiplied this value by −1 for a measure of low educational status. No evidence about educational history was reported in 19 (3%) medical records and was replaced with the sample mean of 10.7 for analysis. We attempted to code learning disorder as in Harris et al., but few symptoms of learning disorder were detailed in the medical records. Interrater reliability in the present study was r = .92 (student coder r = .99).

Antisocial Parenting

Parental crime was coded in terms of the patient’s parents having a criminal history: 0 = neither, 1 = mother, 2 = father, 3 = both. Where there was no clear evidence about parental criminal history, this item was coded 0. Parental alcoholism was measured separately for the patient’s mother and father on a 4-point scale of 0 = never drinks, 1 = used alcohol, but there were no problems associated with use, 2 = some problems, 3 = interference in life, then summed into one score. These variables, though not reported in detail by Harris et al. (2001), were operationalized in the same manner. We coded an additional variable, parental drug use, using a similar scale. Where there was no clear evidence about parental alcoholism or drug use, these variables were coded as 1. There was complete agreement in interrater reliability tests for these variables in the present study (except student coder r = .91 for parental alcoholism).

Whereas Harris et al. (2001) included separate dichotomous variables for childhood physical abuse, neglect, and witnessing family violence, we combined them into a single summed score for a continuous measure. Physical abuse was defined as nonsexual physical abuse, neglect was defined as neglect of basic emotional or physical needs, and witnessing family violence was defined as witnessing the physical or sexual abuse of the mother or other family members, as perpetrated by a parent or guardian occurring prior to the patient’s age of 15 years. This variable had a possible score range of 0 to 8, and absence of evidence for any of these types of abuse was coded 0. Interrater reliability in the present study was r = .87 (student coder r = .93).

Antisociality

Harris et al. (2001) used several indicators of antisociality, including the score from the Psychopathy Checklist–Revised (PCL-R) and the psychopathic deviance subscale of the Minnesota Multiphasic Personality Inventory (MMPI). We coded PCL-R (Hare, 2003) as documented in the medical record by clinical psychology staff but found that it was missing in 72% of our sample, so we did not include it in our analyses. We attempted to code the relevant MMPI-2 scales (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989) or similar psychological assessment measures, but they were rarely documented in the medical record. Nevertheless, we used several other measures used by Harris and colleagues, including the Child and Adolescent Taxon Scale (CATS; Harris, Rice, & Quinsey, 1994) and a count of adulthood Antisocial Personality Disorder (APD) items, both of which have shown concurrent validity with psychopathy measures in other contexts (e.g., Harris et al., 2015; Skilling et al., 2002).

The CATS is the sum of the scores on eight dichotomous (scored 0 or 1) items including elementary school maladjustment, teen alcohol problems, childhood aggression, suspended or expelled, separation from parents, parental alcoholism, arrested under age 16 years, and childhood behavior problems. CATS scores have been shown to distinguish between men who do and do not meet criteria for psychopathy (e.g., Harris et al., 1994; Skilling et al., 2002). In previous research with the CATS, the average interrater reliability coefficient was .90 (Skilling et al., 2002). We coded each CATS item according to the scoring criteria published by Harris and colleagues (2015). The total CATS score was the sum of the eight individual items, with a possible range of 0 to 8. Interrater reliability in the present study was r = .95 (student coder r = .93).

Childhood aggression was also scored separately on a 7-point scale capturing the frequency and severity of aggression perpetrated when aged less than 15 years. This item was coded as 1 = no evidence of aggression, 2 = occasional mild aggression, 3 = frequent mild aggression, 4 = occasional moderate aggression, 5 = frequent moderate aggression, 6 = occasional or frequent severe aggression, 7 = occasional or frequent extreme aggression. Where there was no clear evidence for the presence of any item, the childhood aggression score was coded as 1. Interrater reliability in the present study was r = .98 (student coder r = .88).

APD score was calculated as the total number of adulthood items for APD occurring since age 15 years, as in Harris et al. (2001) except we used the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association [APA], 1994) rather than Diagnostic and Statistical Manual of Mental Disorders (3rd ed.; DSM-III; APA, 1980). Items were failure to conform to social norms, deceitfulness, impulsivity, irritability and aggressiveness, reckless disregard for safety of self or others, consistent irresponsibility, and lack of remorse, and the possible score range was 0 to 7. Interrater reliability in the present study was initially r = .67 (student coder r = .62); coding instructions were modified and a retest 2 years later yielded r = .84.

Criminal Violence

We measured criminal violence using the Cormier-Lang Criminal History Score (Harris et al., 2015) which reflects the frequency and severity of criminal conduct in a single dependent variable. Each criminal charge is assigned a standard value, multiplied by the number of counts of that charge, summed across all charges. For example, homicide is assigned a value of 28, with decreasing weights for other offenses such as aggravated sexual assault (15), armed robbery of a bank or store (8), or assault (2). The Cormier-Lang system has shown excellent interrater reliability in previous archival research involving offenders and forensic patients (e.g., r > .80; Harris et al., 2015), and it has exhibited concurrent validity with different criminal history measures (e.g., r = .31-.60; Jung, Daniels, Friesen, & Ledi, 2012). We counted only preindex violent charges, as in Harris et al.’s (2001) history score for prior arrests for violence, which, in the Cormier-Lang system, includes homicide, all types of assault, contact sexual offenses, armed robbery and robbery with violence, threatening with a weapon, abduction, and forcible confinement. This definition of violence excludes arson, threatening, property, or narcotics offenses. We did not have data on violent recidivism after admission to hospital, so we used historic criminal violence as a single dependent measure. Where there was no clear evidence for the presence of any charges for violent offenses, Criminal Violence was coded as 0. Interrater reliability in the present study was r = .99 (student coder r = .96).

Clinical Outcomes

Data on in-hospital events during the first 1 to 12 months of admission were coded after patients’ admission anniversary date, masked to the coding completed for admission and preadmission variables. We coded health statistics, symptoms, and problem behaviors exhibited during this first year in the institution (or part thereof, for patients who were discharged by the anniversary date). This coding was completed only for 335 men admitted in 2009, 2010, and part of 2011, due to limited resources in later years.

Adult Health Status

A variety of health indices were coded from the medical record. Three health variables were coded from the vital statistics measured and documented by nursing staff on admission (or within a month of admission, if patients had not at first been amenable to assessment). Systolic blood pressure is a risk factor related to cardiovascular disease and other diseases (e.g., Alberti et al., 2009) and can be measured with some precision and objectivity. Therefore, the primary inclusion criterion for cases in the adult health status analyses was having a blood pressure recorded on admission; 269 cases (76% of the sample with in-hospital data coded) had blood pressure recorded within their first month. We coded systolic blood pressure measured in millimeters of mercury (mm Hg) based on the first documented measurement. Body mass index (BMI) is a ratio of weight and height and is related to risk for cardiovascular disease, diabetes, and other diseases (e.g., Jolliffe & Jansen, 2007). We coded BMI as the quotient of weight in kilograms by height in square meters. Information necessary for coding BMI was absent in 55 cases overall (26 or 10% of these 269 cases) and replaced with the sample mean, M = 26.06. Health problem was coded as 0 = absent or 1 = present if the patient had cardiovascular disorder, diabetes, hepatitis, HIV, or similar infectious disease or health problem requiring medication and monitoring on admission. Information about health problems was missing in 39 cases overall (11 or 4% of these 269 cases) and coded as 0. Mean interrater reliability coefficients for blood pressure, height, weight, and health problem were r = .99, 1.0, .78, and .76, respectively.

Three further health variables were based on the digital head-and-shoulders wallet-sized photograph of each patient on his admission report and were assessed before sections of the medical record pertaining to patient history were read. The first was physical attractiveness, which has previously been associated with overall health and longevity (e.g., Nedelec & Beaver, 2014; Re et al., 2015). This variable was rated in terms of facial attractiveness on a scale of 1 to 10, a method used in previous research with excellent interrater reliability (e.g., r > .80; Hilton & Simmons, 2001). In eight cases, this variable was not coded due to poor photograph quality and was replaced with the sample mean, M = 5.16. Interrater reliability for attractiveness was initially r = .40. To increase objectivity of attractiveness ratings, coders were instructed to code attractiveness after a brief first glance at the photograph and before coding other variables (to avoid influence of the patient’s criminal history, etc.), and to set a value of 7 for an “average looking” man. This value was set by consensus among coders, with the expectation that the physical appearance of forensic inpatients on admission fell below that of other men and, while not especially scientific, this approach met with some success in that the mean rated physical attractiveness in this study was 5 (Table 1). A subsequent test of interrater reliability showed improved results, r = .68. This variable was occasionally reviewed for consensus.

Table 1:

Sample Characteristics, Shown as Percentage or Mean and Standard Deviation and Correlation Coefficients With Preadmission Criminal Violence

Characteristic	%	M	SD	r criminal violence
Age at admission		35.2	12.3	.083, p = .036
Diagnosis
Schizophrenia	56			−.004, p = .920
Other psychosis	21			.019, p = .639
Antisocial personality disorder	11			.208, p < .001
Substance related disorder	24			.096, p =.015
Neurodevelopmental problems
Obstetrical problems (pregnancy and perinatal)	14	0.28	0.85	−.008, p = .861
Childhood physical and medical problems	27	0.37	0.73	−.026, p = .566
Highest grade achieved		10.7	2.94	−.181, p < .001
Antisocial parenting
Parental crime	8			.136, p < .001
Parental alcoholism	26			.075, p = .098
Parental drug use	9			.155, p < .001
Experienced childhood abuse or neglect or witnessed family violence	25			.111, p = .005
Antisociality
Childhood aggression score		1.95	1.61	.086, p = .030
Child and Adolescent Taxon Scale		2.38	2.37	.154, p < .001
Antisocial Personality Disorder scale^a		3.84	1.70	.373, p < .001
Criminal violence
Cormier-Lang History Score for violent charges		6.90	7.46
Adult Health Status (n = 269)
Systolic blood pressure		130	17.5	.065, p = .267
Body mass index		26.2	5.41	.082, p = .181
Health problem	18			.046, p = .450
Physical attractiveness		5.06	1.45	−.186, p =.002
Age appearance category				.014, p = .814
Looks younger than stated age	10
Looks older than stated age	14
Facial width-to-height ratio		63.1	4.82	.089, p = .144
Institutional outcomes (n = 335)
Assaults	16	0.50	2.40	.239, p < .001
Threats	29	1.36	3.12	.121, p = .096
Psychotic behaviors		2.33	2.00	−.044, p = .543
Mood problems		1.50	1.34	.005, p = .944

Note. n = 638 except where stated.

Correlation of antisocial personality disorder (APD) and criminal violence may be inflated because criminal history was considered in the APD items.

The second health variable coded from the photograph was the appearance of the patient’s age compared with his actual age (“age appearance”). Perceived age is widely used as a general indicator of overall health, and appearing older than one’s age based on a facial photograph is empirically associated with mortality (e.g., Christensen et al., 2009). This variable was coded by estimating the patient’s age from the photograph, then checking the patient’s actual age, then coding 0 = looks younger than his actual age, 1 = patient looks his actual age, and 2 = looks older than his actual age. Interrater reliability for age appearance was initially r = .36. To increase agreement, additional instructions were added to use values of 0 or 2 only if at least mildly surprised to read the patient’s age after viewing the photograph. A subsequent test of interrater reliability showed improved results, r = 1.0. In seven cases, this variable was not coded due to poor photograph quality and was replaced with a value of 1.

The final health variable coded from the photograph was facial width-to-height ratio (fWHR). This ratio is considered another cue to overall health in men because of its relation to testosterone (e.g., Lefevre, Lewis, Perrett, & Penke, 2013). Facial shape is related to physical strength, independent of attractiveness (e.g., Windhager, Schaefer, & Fink, 2011). When controlling for BMI, higher testosterone is associated with faces being wider and shorter (e.g., Lefevre et al., 2013). Our measure of fWHR was calculated as the quotient of facial width at its widest by facial height at the center of the head from forehead to lower point of the jaw. Interrater reliability for fWHR was r = .81. Poor photo quality, indirect angle, or features being covered by hair meant that this variable was not coded in 56 cases and was replaced with the sample mean, M = 63.08.

Institutional Outcomes

We counted the number of assaults committed during the year in which forceful physical contact was made by the patient against another person. We also counted the number of threats committed during the year in which an unambiguous verbal threat to harm another person was made. Because of the propensity for a small number of patients to make a large number of such threats, we counted up to a maximum of 10 threats per patient in the year, based on previous research with similar populations (e.g., Rice, Harris, & Cormier, 1992) in which 10 threats represented the 75th percentile (G. T. Harris, personal communication, June 15, 2009). The absence of any information about assaults or threats was coded as 0. The mean interrater reliability coefficients for assaults and threat were r = .93 and .88, respectively.

We also coded variables pertaining to events and presentation in hospital using the file version of the Problem Identification Checklist (PIC; Harris et al., 2015). The PIC is used to code symptoms and behaviors exhibited and documented during a 6-month period; we used clinical information from the 6-month period that ended on patients’ anniversary date, or the total time in hospital if less than 6 months. The PIC has four scales, derived from factor analytic research (e.g., Rice, Harris, & Quinsey, 1996), that have shown internal consistencies of .68 to .90 (Quinsey, Coleman, Jones, & Altrows, 1997) and have been coded with interrater reliability coefficients typically .80 or over (e.g., Harris, Rice, & Quinsey, 1993; with a lower reliability of .63 reported by Quinsey et al., 1997, for a “skills deficits” scale that was dropped from the final version published in Harris et al., 2015). Each of the scales has been associated with violent incidents among forensic patients (Quinsey, Jones, Book, & Barr, 2006). For this study, we used the two scales that pertain to symptoms of mental illness. Six items related to psychosis: unusual thought content, hallucinatory behavior, conceptual disorganization, psychotic actions (e.g., stereotypes, bizarre mannerisms, obviously inappropriate laughter), inappropriate suspicion, and grandiosity. Seven items concerned mood: agitation, anxiety, mania, anger, blunted affect, depression, and excessive feelings of guilt. These measures of psychosis and mood problems yielded internal reliability coefficients of α = .74 and .34 in our sample. We counted the number of items present for psychosis and mood problems such that each variable had a possible range of 0 to 6 or 7. The absence of any information indicating these clinical problems was coded as 0. The mean interrater reliability coefficients for psychosis and mood item counts were r = .63 and .51, respectively. To improve reliability, we introduced additional coding rules and examples to aid with the identification of PIC items in the medical record, and routinely coded these items by consensus, which for student coders included at least weekly case review with the first and/or second author. We coded, but did not use for this study, other PIC sections that capture problems of social withdrawal and institutional management (ranging from poor housekeeping and self-care to impulsivity and obnoxious verbal behavior).

Analysis Plan

We examined data for univariate outliers, and an unrealistically high BMI in two cases led us to check and correct the data. We examined the data for multivariate outliers and multivariate normality and found that several scales were positively skewed, particularly those derived from summing values on categorical variables. Consequently, we applied bootstrapping to our statistical analysis. We used a step approach to estimation as suggested by Anderson and Gerbig (1988), using confirmatory factor analysis to test the measurement model and then estimating the full structural model. In deriving models, we used maximum likelihood estimation with bootstrapping, which is considered suitable for theory testing and robust for moderate violations of the normality assumptions (Anderson & Gerbig, 1988). We used the jigsaw method (Bollen, 2000) to determine the best fitting measurement and structural models. All structural equation analyses were conducted in AMOS Version 21. Within models variable multicollinearity was below r = .85. We illustrated the model results in a series of figures, in which latent variables (those not directly observed, but rather inferred through a mathematical model) are represented by ovals and observed (directly measured) variables are indicated by rectangles.

Criminal Violence

We first tested Harris et al.’s (2001) measurement model for the latent variables of Antisociality, Neurodevelopment Problems, and Antisocial Parenting, in the full sample of 638 cases. We then tested the structural model predicting criminal violence, with our single dependent measure of preindex violent charges, again with bootstrapping.

Clinical Outcomes

We first conducted an exploratory factor analysis (EFA) using SPSS Version 22 to identify possible latent factors among the adult health status variables, in the subsample of 269 cases that had admission blood pressure data. We used principal components analysis, Varimax rotation with Kaiser normalization, identifying factors based on eigenvalues > 1 rather than setting a number of factors. Next, we confirmed the result in a measurement model for adult health status and then examined a structural model using antisociality and neurodevelopmental problems with adult health status as the outcome. Similarly, we conducted an EFA to identify possible latent factors among the behavioral and symptomatic variables in the subsample of 335 cases that had institutional data for the first year in hospital. Next, we confirmed the result in a measurement model and then examined a structural model using antisociality and neurodevelopmental problems to predict institutional outcomes.

Results

Sample characteristics are shown in Table 1, along with their bivariate associations with criminal violence. Our sample appears broadly similar to that reported by Harris and colleagues (2001) in rates of criminal violence as well as such characteristics as obstetrical problems, CATS score, and criminal violence, although there appears to be a somewhat higher rate of childhood problems (likely because we used an older age cut off) and lower rate of parental alcoholism (perhaps because we used a lower score cut off, but we are unable to verify this possibility).

Measurement Models

Developmental Pathways

Our attempt to replicate the three-factor structure was successful, with the addition of a covariance between error terms for obstetrical complications and childhood problems. The measurement model pictured in Figure 1 yielded a good fit to the data as shown by the indices in Table 2. Antisociality, Neurodevelopmental Problems, and Antisocial Parenting were positively related to each other and, as in Harris et al. (2001), it appeared that Antisociality was better instantiated than the other latent variables.

Figure 1:

Developmental Pathways Measurement Model

Table 2:

Goodness of Fit Indices for Developmental Pathways and Structural Models

Goodness of fit index	Model 1: Measurement model, developmental pathways	Model 2: Measurement model, adult health	Model 3: Measurement model, institutional outcomes	Model 4: Structural model, criminal violence	Model 5: Structural model, adult health status	Model 6: Structural model, institutional outcomes
χ²	58.742	19.215	1.124	239.027	129.942	71.859
df	31	8	1	41	98	72
χ²/df	1.895	2.402	1.124	5.830	1.326	0.998
GFI	0.982	0.977	0.998	0.940	0.944	0.971
TLI	0.969	0.764	0.992	0.811	0.947	1.00
CFI	0.979	0.874	0.999	0.859	0.957	1.00
RMSEA	0.037	0.072	0.019	0.087	0.035	0.000
90% CI	[0.022, 0.052]	[0.031, 0.114]	[0.000, 0.148]	[0.077, 0.098]	[0.016, 0.050]	[0.000, 0.032]
n	638	269	335	638	269	335

Note. Indices within range of good fit indicate a well-fitting model (Byrne, 2001). GFI = goodness of fit index, good fit values approaching 1.00; TLI = Tucker–Lewis index, good fit values ≥ .95; CFI = comparative fit index, good fit values ≥ .95; RMSEA = root mean square error of approximation, good fit values < .05; CI = confidential interval.

Adult Health Status

The EFA yielded two factors in a rotated solution. Variables loading on the first factor were age appearance, health problem, and physical attractiveness, which we labeled Youthfulness. Variables loading on the second factor were blood pressure, BMI, and fWHR, which we labeled Obesity. A measurement model with Youthfulness and Obesity as the latent measures yielded a reasonably good fit as shown by the indices in Table 2. This model is illustrated in Figure 2.

Figure 2:

Adult Health Measurement Model

Institutional Outcomes

EFA indicated two factors. One was comprised of assaults and threats, which we labeled Aggressive Behaviors. The other included symptoms of psychosis and mood disorder, which we labeled Mental Illness. This yielded a well-fitting measurement model (Table 2), illustrated in Figure 3.

Figure 3:

Institutional Outcomes Measurement Model

Structural Models

Criminal Violence

Our attempt to replicate the Harris et al. (2001) structural model of the two hypothetical pathways to criminal violence was limited to a single dependent variable, history of criminal violence. We replicated this structural model (Table 2), pictured in Figure 4. All paths and covariances illustrated were significant.

Figure 4:

Criminal Violence Structural Model

Adult Health Status

The structural model of Antisociality and Neurodevelopmental Problems with Youthfulness and Obesity outcomes was specified and tested (Table 2). Regression weights were large and significant. Weights were positive between Antisociality and Youthfulness and negative between Antisociality and Obesity, with the opposite for Neurodevelopmental Problems (Figure 5).

Figure 5:

Adult Health Status Structural Model

Institutional Outcomes

The structural model of Antisociality and Neurodevelopmental Problems with Mental Illness and Aggressive Behaviors outcomes was specified and tested (Table 2). Regression weights were significant; however, some negative variances were observed, suggesting the model was not identified. Modifications to this model that yielded an identified model failed to produce a good fit.

Discussion

In this sample of men admitted for assessment in a forensic psychiatric hospital, we attempted to replicate the psychopathy and neurodevelopmental pathways obtained by Harris et al. (2001) and extend them to models of not only criminal violence but also adult health status and institutional outcomes. The three-factor structure of latent variables that Harris and colleagues labeled Psychopathy, Neurodevelopmental Problems, and Antisocial Parenting was successfully replicated, albeit without use of the PCL-R (hence our label Antisociality in lieu of Psychopathy), and the measurement model yielded a good fit to the data. Consistent with Hypothesis 1, we obtained a structural model supporting antisociality and neurodevelopmental problems as pathways to criminal violence.

Whereas patients in Harris and colleagues’ (2001) study were assessed in 1975-1981, our sample pertained to admissions in 2009-2012. This sample represents a new generation of forensic patients which differed in some clinical characteristics such as higher rates of substance-related disorders and higher risk of violent recidivism (but no greater prevalence of serious mental illness; Hilton, Ham, & Lang, 2011) and likely greater ethnic diversity, reflecting demographic changes in Ontario during this time (Statistics Canada, 2008). The current findings confirm previous work with adult offenders and thus indicate some robustness of antisocial development and neurodevelopmental damage as two pathways to criminal violence. It is consistent with the theory that one type of offender exhibits pathological development and behavior, whereas another type, associated with psychopathy, exhibits more adaptive development and nonpathological violence (e.g., Harris, Rice, Hilton, Lalumière, & Quinsey, 2007; Harris et al., 2001). However, we did not find that the two paths were independent. It may be that legislative changes or criminal justice practices over the past 30 years resulted in a population of forensic assessees that is more generally antisocial so that both pathways represent variations of the life-course-persistent antisocial pathway, or perhaps the wider use of substances is indicative of vulnerability to mental illness in both offender types (e.g., Ksir & Hart, 2016).

We examined clinical data in order to test whether the theoretical pathways generalized beyond criminal violence and to consider the implications for clinical practice. We obtained well-fitting measurement models of adult health status, including indicators of youthfulness and obesity factors, and institutional outcomes including aggressive behaviors and mental illness. Consistent with Hypothesis 2, antisociality and neurodevelopmental problems statistically predicted youthfulness and obesity outcomes, respectively, lending support to the notion of these two paths as developmental trajectories relevant to lifestyle outcomes beyond criminal behavior. Paths to the obesity outcomes were particularly large and indicated a strong positive association with neurodevelopmental problems. Measured obesity is a component of the metabolic syndrome (MetS), a condition that also includes hypertension, hyperglycemia, and hyperlipidemia and that carries a risk of increased mortality, particularly among the seriously mentally ill (e.g., De Hert et al., 2011). Forensic inpatients appear vulnerable to obesity associated with substantial weight gains during their hospitalization (e.g., Hilton et al., 2015). The current findings suggest that vulnerability to obesity and related health risks could be linked to precursors of mental illness identifiable in early childhood, raising the potential for early identification of risk for adverse health outcomes in the forensic patient population.

Contrary to Hypothesis 3, institutional outcomes were not well predicted in the structural model using mental illness and aggressive behaviors as the dependent factors. Our inability to produce good model fit might be partly attributable to the PIC scales of psychosis and mood disorders being used in combination. In previous research with the PIC, the scale evaluating mood problems yielded lower internal reliability and item loadings than other scales (e.g., Quinsey et al., 1997; Rice et al., 1996), which was also true in our study. Mood problem constituent items such as anger and agitation might be less suited to the concept we intended to instantiate in our latent variable of mental illness.

Limitations

Whereas Harris and colleagues (2001) included both preadmission and recidivism offenses in their definition of criminal violence, we used only preadmission violent charges as a single measured dependent variable. Therefore, we did not fully replicate their model using the same latent variable. We did, however, measure violence in the same manner using the Cormier-Lang Criminal History score and extended their model to adult health outcomes, suggesting that the latent structure of antisociality and neurodevelopmental problems bears some generalizability beyond the construct of criminal violence. Future research might include recidivism and other posthospitalization outcomes such as symptom management, avoidance of readmission, employment, and other indicators of treatment and rehabilitation. Although many men assessed in the forensic system may go on to be convicted, and there is much overlap of the correctional and forensic populations, the generalizability of the current results to correctional inmates cannot be assumed; similar studies to replicate findings in the correctional population could also examine institutional adjustment, infractions, and postrelease recidivism.

We did not have sufficient cases with a clinician-scored PCL-R assessment to be able to use this standard measure of psychopathy. We also omitted relevant MMPI measures. As a result, our construct included only three of the five variables that Harris and colleagues (2001) used to instantiate psychopathy, and our results may not be directly comparable. This limitation applies to neurodevelopmental problems and antisocial parenting as well, in that we did not use exactly the same variables in either case, and in other variables we had some missing values for which we used mean replacement, but it is most notable for PCL-R which is a widely used diagnostic and research tool for measuring psychopathy. We named the latent variable “Antisociality” in light of the absence of PCL-R, although we recognize that there is no single measure of the concept of psychopathy.

We did not obtain a successful structural model for institutional outcomes, and we have noted the possible limitations of our measures using the PIC. Future research examining a broader range of clinically relevant events (e.g., treatment compliance and response, seclusion, progression through security levels) could explore whether the antisociality and neurodevelopmental problems pathways bear any implication for clinical interventions.

Our measure of physical attractiveness, an indicator of youthfulness, was limited to ratings of a nonstandardized head-and-shoulders photograph, and it was inversely associated with criminal violence. In some previous research, facial attractiveness was not significantly related to physical fitness whereas body attractiveness was (Hönekopp et al., 2007). Replication of the current results using full-body photographs might be of value.

Research and Clinical Implications

Our study was restricted to the available cross-sectional data, and we coded most data at a single time point based on records that had been gathered over time and placed in an essentially static record. Existing research following youth at risk for criminal behavior into adulthood (e.g., Corrado et al., 2015; Piquero et al., 2012) has been important to understanding offender pathways and can help in the early identification of youth on the riskiest paths before they emerge as chronic offenders. Similar research delineating pathways among adult offenders would benefit from longitudinal data collection; however, such studies are extraordinarily resource intensive. Furthermore, there is potential clinical utility in identifying offender types among adults based on data gathered upon admission to a correctional or treatment facility. Cross-sectional research with postrelease follow-up may be the most feasible way to do this.

Evidence for separate pathways involving antisocial development or neurodevelopmental problems underscores the need for assessment of adult offenders that covers not only criminal history but also medical history and psychosocial development. With respect to the former, both pathways were positively related to criminal violence, consistent with the need to treat criminogenic needs. Pathways assessment could help guide different treatment approaches according to offender subtype. Harris and colleagues (2001), for example, suggested that offenders with neurodevelopmental damage, compared with those with a history indicating psychopathy, might be more responsive to environmental contingencies. Behavioral token economies or therapeutic milieus could be valuable in such cases, as well as attending to environmental or psychopathological triggers of violent behavior. In contrast, offenders on the antisociality pathway might be more likely to commit instrumental violence and may be more amenable to aggression replacement interventions. The present study was unable to distinguish between instrumental and pathological or reactive aggression, and this is a needed next step.

The present study design cannot elucidate causal relationships, but it is plausible that patients on the pathway of neurodevelopmental damage rather than antisociality may be more prone to physical ill health characterized by obesity. In the few existing studies of forensic psychiatric samples, most patients were overweight or obese (Haw & Rowell, 2011; Hilton et al., 2015; Ojala et al., 2008; Vasudev, Thakkar, & Mitcheson, 2012). They had more coronary heart disease than the surrounding communities (e.g., Ivbijaro, Kolkiewicz, McGee, & Gikunoo, 2008) and a higher than normal rate of diabetes mellitus (Kelbrick, Muthu-Veloe, & Picchioni, 2011). Awareness of patients’ developmental pathway on admission could help identify those who would most benefit from preventative interventions for such illnesses.

Summary and Conclusion

This study shows that the antisociality and neurodevelopmental pathways identified among adult offenders are replicable, although perhaps not as independent as previously thought, and may have implications for clinical outcomes such as physical health among forensic patients. The antisociality path is commensurate with existing research with adolescents showing an early onset of antisocial conduct, related to dysfunctional parenting in the life-course-persistent pathway. The neurodevelopmental path suggests new lines of research that could be extended into adolescent work, particularly in large cohort samples, in which medical events and poor educational performance might yield stronger evidence for a neurodevelopmental pathway.

Footnotes

Acknowledgements

We thank the late Grant T. Harris for sharing research coding manuals to permit comparable coding of information in the present study. We are indebted to Carol Lang for her extensive contribution to the coding manuals and data collection in the present study, and Sonja Dey, Jenna Rutherford, and Chelsea Turan for research assistance. We thank Terri Newman, Louise Moreau, Waypoint Provincial Forensic Programs staff, and Waypoint Clinical Information Services staff for assistance and feedback. Martin Lalumière, Alecia Dretzkat, and Jon Bridekirk provided helpful comments on an earlier manuscript.

Authors’ Note:

Michelle Green is now at Beaver Creek Institution, Gravenhurst, Ontario.

N. Zoe Hilton, PhD, is senior research scientist at the Waypoint Centre for Mental Health Care, associate professor of psychiatry at the University of Toronto, and a registered psychologist. She holds degrees from the University of Cambridge and the University of Toronto and she has published primarily on interpersonal violence and risk assessment.

Elke Ham, OCGC, is a research psychometrist at the Waypoint Centre for Mental Health Care. She holds a research analyst diploma from Georgian College. She is the project coordinator for ODARA 101: The Electronic Training Program. Her research interests are in program evaluation, violence risk communication, and violent offenders.

Michelle M. Green, PhD, is a consultant psychologist on the Forensic Assessment Program at the Waypoint Centre for Mental Health Care and the Chief of Mental Health Services at Beaver Creek Institution. She also maintains a private practice. She holds a PhD from Brock University.

References

Alberti

K. G.

Eckel

R. H.

Grundy

S. M.

Zimmet

P. Z.

Cleeman

J. I.

Donato

K. A.

. . . Smith

S. C.

(2009). Harmonizing the metabolic syndrome: A joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation, 120, 1640-1645.

American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author.

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

Anderson

J. C.

Gerbig

D. W.

(1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411-423.

Augustyn

M. B.

(2015). The (ir)relevance of procedural justice in the pathways to crime. Law and Human Behavior, 39, 388-401. doi:10.1037/lhb0000122

Bollen

K. A.

(2000). Modeling strategies: In search of the Holy Grail. Structural Equation Modeling: A Multidisciplinary Journal, 7, 74-81. doi:10.1207/S15328007SEM0701_03

Branchi

Cirulli

(2014). Early experiences: Building up the tools to face the challenges of adult life. Developmental Psychobiology, 56, 1661-1674. doi:10.1002/dev.21235

Butcher

J. N.

Dahlstrom

W. G.

Graham

J. R.

Tellegen

Kaemmer

(1989). MMPI-2: Minnesota Multiphasic Personality Inventory-2 manual for administration and scoring. Minneapolis: University of Minnesota Press.

Byrne

B. M.

(2001). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum.

10.

Canadian Institute of Health Research, Natural Sciences and Engineering Research Council of Canada, and Social Sciences and Humanities Research Council of Canada. (December, 2014). Tri-council policy statement: Ethical conduct for research involving humans. Retrieved from http://www.pre.ethics.gc.ca/pdf/eng/tcps2-2014/TCPS_2_FINAL_Web.pdf

11.

Christensen

Thinggaard

McGue

Rexbye

Hjelmborg

J. v. B.

. . . Vaupel

J. W.

(2009). Perceived age as clinically useful biomarker of ageing: Cohort study. British Medical Journal, 339, b5262. doi:10.1136/bmj.b5262

12.

Corrado

R. R.

McCuisha

E. C.

Hart

S. D.

DeLisi

(2015). The role of psychopathic traits and developmental risk factors on offending trajectories from early adolescence to adulthood: A prospective study of incarcerated youth. Journal of Criminal Justice, 43, 357-368. doi:10.1016/j.jcrimjus.2015.04.007

13.

Dandreau

D. M.

Frick

P. J.

(2009). Developmental pathways to conduct problems: A further test of the childhood and adolescent-onset distinction. Journal of Abnormal Child Psychology, 37, 375-385.

14.

Day

D. M.

Wanklyn

S. G.

Yessine

A. K.

(2014). A review of terminological, conceptual, and methodological issues in the developmental risk factor literature for antisocial and delinquent behavior. Child & Youth Care Forum, 43, 97-112. doi:10.1007/s10566-013-9227-9

15.

De Hert

Correll

C. U.

Bobes

Cetkovich-Bakmas

Cohen

Asai

. . . Leucht

. (2011). Physical illness in patients with severe mental disorders. I. Prevalence, impact of medications and disparities in health care. World Psychiatry, 10, 52-77.

16.

Diamantopoulou

Verhulst

F. C.

van der Ende

(2010). Testing developmental pathways to antisocial personality problems. Journal of Abnormal Child Psychology, 38, 91-103.

17.

Finkelhor

Shattuck

Turner

Hamby

(2015). A revised inventory of adverse childhood experiences. Child Abuse & Neglect, 48, 13-21.

18.

Fox

B. H.

Jennings

W. G.

Farrington

D. P.

(2015). Bringing psychopathy into developmental and life-course criminology theories and research. Journal of Criminal Justice, 43, 274-289. doi:10.1016/j.jcrimjus.2015.06.003

19.

Frick

P. J.

Cornell

A. H.

Bodin

S. D.

Dane

H. E.

Barry

C. T.

Loney

B. R.

(2003). Callous–unemotional traits and developmental pathways to severe conduct problems. Developmental Psychology, 39, 246-260. doi:10.1037/0012-1649.39.2.246

20.

Gilligan

D. G.

Lemmings

C. J.

(2010). Psychopathic and neuropathic pathways to homicide: Examination of Harris and Rice’s two-path model of criminal violence in homicide. Psychiatry, Psychology and Law, 17, 148-168. doi:10.1080/13218710903427875

21.

Hare

R. D.

(2003). Hare PCL-R (2nd ed.). Toronto, Ontario, Canada: Multi-Health Systems.

22.

Harris

G. T.

Rice

M. E.

Hilton

N. Z.

Lalumière

M. L.

Quinsey

V. L.

(2007). Coercive and precocious sexuality as a fundamental aspect of psychopathy. Journal of Personality Disorders, 21, 1-29.

23.

Harris

G. T.

Rice

M. E.

Lalumière

(2001). Criminal violence: The roles of psychopathy, neurodevelopmental insults, and antisocial parenting. Criminal Justice and Behavior, 4, 402-426.

24.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

(1993). Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Criminal Justice and Behaviour, 20, 315-335.

25.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

(1994). Psychopathy as a taxon: Evidence that psychopaths are a discrete class. Journal of Consulting and Clinical Psychology, 62, 387-397.

26.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

Cormier

C. A.

(2015). Violent offenders: Appraising and managing risk (3rd ed.). Washington, DC: American Psychological Association.

27.

Haw

Rowell

(2011). Obesity and its complications: A survey of inpatients at a secure psychiatric hospital. The British Journal of Forensic Practice, 13, 270-277.

28.

Hilton

N. Z.

Ham

Lang

(2011). Oak Ridge admissions in 2009: Characteristics and comparison with 1989. Unpublished report, Waypoint Centre for Mental Health Care.

29.

Hilton

N. Z.

Ham

Lang

Harris

G. T.

(2015). Weight gain and its correlates among forensic inpatients. Canadian Journal of Psychiatry, 60, 232-238.

30.

Hilton

N. Z.

Simmons

J. L.

(2001). The influence of actuarial risk assessment and clinical judgments in tribunal decisions about mentally disordered offenders. Law and Human Behavior, 25, 393-408.

31.

Hönekopp

Rudolph

Beier

Liebert

Müller

(2007). Physical attractiveness of face and body as indicators of physical fitness in men. Evolution and Human Behavior, 28, 106-111. doi:10.1016/j.evolhumbehav.2006.09.001

32.

Hui

Middleton

Völlm

(2013). Coercive measures in forensic settings: Findings from the literature. International Journal of Forensic Mental Health, 12, 53-67.

33.

Hunter

J. A.

Figueredo

A. J.

Malamuth

N. M.

(2010). Developmental pathways into social and sexual deviance. Journal of Family Violence, 25, 141-148. doi:10.1007/s10896-009-9277-9

34.

Ivbijaro

Kolkiewicz

L. A.

McGee

L. S.

Gikunoo

(2008). Addressing long-term physical healthcare needs in a forensic mental health inpatient population using the UK primary care Quality and Outcomes Framework (QOF): An audit. Mental Health in Family Medicine, 5, 51-60.

35.

Jolliffe

C. J.

Jansen

(2007). Development of age-specific adolescent metabolic syndrome criteria that are linked to the adult treatment panel III and International Diabetes Federation criteria. Journal of the American College of Cardiology, 49, 891-898.

36.

Jones

N. J.

Brown

S. L.

Wanamaker

K. A.

Greiner

L. E.

(2014). A quantitative exploration of gendered pathways to crime in a sample of male and female juvenile offenders. Feminist Criminology, 9, 113-136. doi:10.1177/1557085113501850

37.

Jung

Daniels

M. K.

Friesen

Ledi

(2012). An examination of convergent constructs among level of service measures and other measures. The Journal of Forensic Psychiatry & Psychology, 23, 601-619.

38.

Kazdin

A. E.

Kraemer

H. C.

Kessler

R. C.

Kupfer

D. J.

Offord

D. R.

(1997). Contributions of risk-factor research to developmental psychopathology. Clinical Psychology Review, 17, 375-406.

39.

Kelbrick

Muthu-Veloe

Picchioni

(2011). An audit of diabetes mellitus management within a specialist secure psychiatric hospital. Journal of Psychiatric Intensive Care, 8, 88-95. doi:10.1017/S1742646411000239

40.

Ksir

Hart

(2016). Cannabis and psychosis: A critical overview of the relationship. Current Psychiatry Reports, 18, 1-12. doi:10.1007/s11920-015-0657-y

41.

Lefevre

C. E.

Lewis

G. J.

Perrett

D. I.

Penke

(2013). Telling facial metrics: Facial width is associated with testosterone levels in men. Evolution and Human Behavior, 34, 273-279. doi:10.1016/j.evolhumbehav.2013.03.005

42.

Loeber

Burke

J. D.

(2011). Developmental pathways in juvenile externalizing and internalizing problems. Journal of Research on Adolescence, 21, 34-46. doi:10.1111/j.1532-7795.2010.00713.x

43.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

44.

Moffitt

T. E.

Arseneault

Belskya

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, 108, 2693-2698. doi:10.1073/pnas.1010076108

45.

Moffitt

T. E.

Caspi

(2001). Childhood predictors differentiate life-course persistent and adolescence-limited antisocial pathways among males and females. Development and Psychopathology, 13, 355-375.

46.

Mwaniki

M. K.

Atieno

Lawn

J. E.

Newton

C. R. J. C

. (2012). Long-term neurodevelopmental outcomes after intrauterine and neonatal insults: A systematic review. The Lancet, 379, 445-452. doi:10.1016/S0140-6736(11)61577-8

47.

Nedelec

J. L.

Beaver

K. M.

(2014). Physical attractiveness as a phenotypic marker of health: An assessment using a nationally representative sample of American adults. Evolution & Human Behavior, 35, 456-463. doi:10.1016/j.evolhumbehav.2014.06.004

48.

Nestor

P. G.

Kimble

Berman

Haycock

(2002). Psychosis, psychopathy, and homicide: A preliminary neuropsychological inquiry. American Journal of Psychiatry, 159, 138-140.

49.

Nisbet

I. A.

Smallbone

S. W.

Wortley

(2010). Developmental, individual and family characteristics of specialist, versatile, and short duration adolescent sex offenders. Sexual Abuse in Australia and New Zealand, 2, 85-96.

50.

Ojala

Niskanen

Tiihonen

Paavola

Putkonen

Repo-Tiihonen

(2008). Characterization of metabolic syndrome among forensic psychiatric patients. The Journal of Forensic Psychiatry and Psychology, 191, 33-51.

51.

Open Science Collaboration. (2015). Estimating the reproducibility of psychological science. Science, 349, aac4716-aac47168. doi:10.1126/science.aac4716

52.

Patterson

G. R.

DeBaryshe

B. D.

Ramsey

(1989). A developmental perspective on antisocial behavior. American Psychologist, 44, 329-335.

53.

Piquero

A. R.

Farrington

D. P.

Fontaine

N. M. G.

Vincent

Coid

Ullrich

(2012). Childhood risk, offending trajectories, and psychopathy at age 48 years in the Cambridge Study in Delinquent Development. Psychology, Public Policy, and Law, 18, 577-598. doi:10.1037/a0027061

54.

Quinsey

V. L.

Coleman

Jones

G. B.

Altrows

I. F.

(1997). Proximal antecedents of eloping and reoffending among supervised mentally disordered offenders. Journal of Interpersonal Violence, 12, 794-813. doi:10.1177/088626097012006002

55.

Quinsey

V. L.

Jones

G. B.

Book

A. S.

Barr

K. N.

(2006). The dynamic prediction of antisocial behavior among forensic psychiatric patients: A prospective field study. Journal of Interpersonal Violence, 21, 1539-1565. doi:10.1177/0886260506294238

56.

Rapoport

J. L.

Giedd

J. N.

Gogtay

(2012). Neurodevelopmental model of schizophrenia: Update 2012. Molecular Psychiatry, 17, 1228-1238. doi:10.1038/mp.2012.23

57.

D. E.

Tskhay

K. O.

Tong

M.-O.

Wilson

J. P.

Zhong

C.-B.

Rule

N. O.

(2015). Facing fate: Estimates of longevity from facial appearance and their underlying cues. Archives of Scientific Psychology, 3, 30-36. doi:10.1037/arc0000015

58.

Rice

M. E.

Harris

G. T.

Cormier

C. A.

(1992). An evaluation of a maximum security therapeutic community for psychopaths and other mentally disordered offenders. Law & Human Behavior, 16, 399-412.

59.

Rice

M. E.

Harris

G. T.

Quinsey

V. L.

(1996). Treatment for forensic patients. In Sales

B. D.

Shah

S. A.

(Eds.), Mental health and law: Research, policy, and services (pp. 141-189). Durham, NC: Carolina Academic Press.

60.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555-592. doi:10.1111/j.1745-9125.2003.tb00997.x

61.

Santana

E. J.

(2016). The brain of the psychopath: A systematic review of structural neuroimaging studies. Psychology & Neuroscience, 9, 420-443. doi:10.1037/pne0000069

62.

Simpson

A. I.

Grimbos

Chan

Penney

S. R.

(2015). Developmental typologies of serious mental illness and violence: Evidence from a forensic psychiatric setting. Australian & New Zealand Journal of Psychiatry, 49, 1048-1059. doi:10.1177/0004867415587745

63.

Skilling

T. A.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

(2002). Identifying persistently antisocial offenders using the Hare Psychopathy Checklist and DSM Antisocial Personality Disorder criteria. Psychological Assessment, 14, 27-38.

64.

Statistics Canada. (2008). Canada’s ethnocultural mosaic, 2006 census (Catalogue No. 97-562-X). Retrieved from http://www12.statcan.ca/census-recensement/2006/as-sa/97-562/pdf/97-562-XIE2006001.pdf

65.

Stickle

T. R.

Frick

P. J.

(2002). Developmental pathways to severe antisocial behavior: Interventions for youth with callous–unemotional traits. Expert Review of Neurotherapeutics, 2, 511-522.

66.

Teicher

M. H.

Samson

J. A.

(2016). Enduring neurobiological effects of childhood abuse and neglect. Journal of Child Psychology and Psychiatry, 57, 241-266. doi:10.1111/jcpp.12507

67.

Vasudev

Thakkar

P. B.

Mitcheson

(2012). Physical health of patients with severe mental illness. International Journal of Health Care Quality Assurance, 25, 363-370.