Understanding the Developmental Course and Severity of Criminality Among Individuals with SMI

Anne G. Crocker, Université de Montréal and Institut Philippe-Pinel de Montréal
Michael S. Martin, University of Ottawa
Marichelle C. Leclair, Institut Philippe-Pinel de Montréal and McGill University
Tonia L. Nicholls, University of British Columbia and BC Mental Health and
Substance Use Services, Vancouver, British Columbia, Canada
Michael C. Seto, Royal Ottawa Health Care Group, Ottawa, Ontario, Canada

The early and late starter model provides one of the most enduring frameworks for understanding the developmental course and severity of violence and criminality among individuals with severe mental illness. We expanded the model to account for differences in the age of onset of criminal behavior and added a group with no prior contact with the justice or mental health systems. We sampled 1,800 men and women found Not Criminally Responsible on account of Mental Disorder in 3 Canadian provinces. Using a retrospective file-based study, we explored differences in criminal, health, demographic, and social functioning characteristics, processing through the forensic psychiatric system and recidivism outcomes of 5 groups. We replicated prior research, finding more typical criminogenic needs among those with early onset crime. Those with crime onset after mental illness were more likely to show fewer criminogenic needs and to have better outcomes upon release than those who had crime onset during adulthood, before mental illness. Individuals with no prior contact with mental health or criminal justice had higher functioning prior to their crime and had a lower risk of reoffending. Given little information is needed to identify the groups, computing the distribution of these groups within forensic mental health services or across services can provide estimates of potential intensity or duration of services that might be needed. This study suggests that distinguishing subgroups of forensic clients based on the sequence of onset of mental illness and criminal behavior and on the age of onset of criminal behavior may be useful to identify criminogenic needs and predict outcomes upon release. This updated framework can be useful for planning organization of services, understanding case mix, as well as patient flow in forensic services and flow of mentally disordered offenders in correctional services.

typology, recidivism, violence, criminality, mental illness

Background

“The criminal justice system has become a common gateway to mental health care for individuals with serious mental illness (SMI). Both forensic and civil psychiatric services provide care to growing numbers of individuals with SMI who have come into conflict with the law. With this increase comes a much more heterogeneous population of mentally ill individuals who are stigmatized with combined psychiatric and forensic labels. The variability in mental health and criminogenic needs in this evolving population has direct implications for the
organization of services in terms of intensity and breadth of services, resource allocation, as well as safety and security of patients, care providers and the community” (p. 84).

“The early and late starter model is one of the earliest and most enduring for understanding criminality among individuals with SMI . . . In the most recent version of the model, Hodgins (2008) posited three trajectories: (a) individuals who exhibit antisocial behavior during adolescence, usually prior to mental illness onset, and persist into adulthood; (b) individuals who first exhibit antisocial behavior in adulthood, after the onset of their mental illness; and (c) individuals who suddenly engage in serious violence, later in life, sometimes many years after the onset of mental illness. This revised model divided the late starter group into two groups: individuals with an atypical onset of criminal activity in mid- or late-life and those whose criminality began during the more typical periods of adolescence or early adulthood. This new group accounts for the important distinction in terms of typical versus atypical timing of onset of criminal behavior. However, to date there has been little consideration of atypical onset of mental illness” (p. 84).

“The early and late starter model suggests that mental illness may be an important risk factor for criminality for those individuals whose criminality occurs around or after the onset of mental illness. For individuals whose criminal onset precedes the illness, traditional criminogenic factors such as substance abuse or criminal associates are thought to be more important than clinical factors. Studies have found that people in the early starter trajectory display more violent behavior, more versatile criminal behavior, are more likely to have substance use problems, and have higher scores on psychopathy and antisocial personality scales compared with those who fall in the late starter trajectory. Many authors have continued to hypothesize that symptoms of mental illness would be more important drivers of violence among late compared with early starters, despite variability observed across studies. The early and late starter model can help define subgroups of individuals differing in etiologies, needs, and risk for future mental health and criminal justice involvement” (p. 84).

Current Study

“We sought to expand the testing of the early and late starter model to explore its utility across the full spectrum of severe mental disorders found in the forensic population. Our specific objective was to compare different starter groups according to their onset of criminality and mental illness on their sociodemographic, mental health and criminological characteristics, as well as their pathways through the forensic system . . . Data from this study were extracted from the National Trajectory Project, a longitudinal study of individuals found NCRMD in the three largest provinces of Canada—Québec, Ontario, and British Columbia . . . The sample included 1,800 adults found NCRMD between 2000 and 2005” (p. 85).

“Our application of the model included five groups based on two dimensions: (a) the sequence of onset of mental illness and criminal behavior, and (b) the age of onset of criminal behavior . . . Based on the first dimension, we identified two groups, which we further split based on the age of onset of criminal behavior. Preillness starters (traditionally labeled “early starters”) had a first criminal charge before their first contact with mental health services. They were divided into adolescent preillness starters (criminal onset before 18 years old) and adult preillness starters (criminal onset at 18 years old or older). Postillness starters (traditionally labeled “late starters”) had a first criminal charge after their first contact with mental health services. They were divided into younger postillness starters (criminal onset before 35 years old) and older postillness starters (criminal onset at 35 years old or older) . . . Finally, we labeled those whose NCRMD verdict was the first formal contact with both the mental health and criminal justice system as first presenters, regardless of their age” (p. 85).

Results

“Adolescent preillness starters had 0.3 times the odds of high school completion compared with younger postillness starters. They had similar primary diagnoses, but had 63% increased odds of having a comorbid substance use disorder and two times the odds of having a comorbid personality disorder or traits. They had a lower rate of prior psychiatric hospitalization per year lived, but over three times the rate of prior charges compared with younger postillness starters. Among those with at least one prior charge, adolescent preillness starters displayed greater criminal diversity than younger postillness starters” (p. 86).

“Adult preillness starters had two times the odds of being in a relationship and of earning an income at the time of the index offense compared with younger postillness starters. There were few differences in terms of diagnoses, with the exception that the relative odds of having “other” primary diagnoses rather than psychotic spectrum disorder were 85% higher in adult preillness starters than in younger postillness starters. Similar to adolescent preillness starters, adult preillness starters had under half the rate of prior psychiatric hospitalization, but an increased rate of prior charges compared with younger postillness starters. However, they displayed lower criminal diversity. Older postillness starters had three times the odds of being in a relationship compared with younger postillness starters. The relative odds of having a primary diagnosis of mood spectrum disorder rather than psychotic spectrum disorder were over 50% higher in this group than in younger postillness starters. They also had half the odds of having a comorbid substance use disorder and of having a comorbid personality disorder or traits. They had lower rates of psychiatric hospitalizations and prior charges, and those with at least one charge showed lower criminal diversity compared with younger postillness starters” (p. 87-88).

“First presenters had three times the odds of being in a relationship and six times the odds of earning their income compared with younger postillness starters. They also had half the odds of being homeless at the time of the index offense. They differed in terms of diagnosis: the relative odds of having a primary diagnosis other than mood spectrum or psychotic spectrum disorders rather than psychotic spectrum disorder were 2.7 times higher in first presenters than in younger postillness starters, and they had half the odds of having a comorbid substance use disorder” (p. 88).

“There were few differences between the groups in terms of psychiatric symptoms at the time of the index offense. Adolescent preillness starters were similar to younger postillness starters with respect to all characteristics of the index offense. Adult preillness starters had lower odds of having a mention of psychotic symptoms at the time of the index offense compared with younger postillness starters. They also had lower odds of weapon use, and 50% increased odds of victimizing an acquaintance. Older postillness starters had lower odds of drug/alcohol use at the time of the index offense compared with younger postillness starters. They also had lower relative odds of having perpetrated an index offense against a person and of having perpetrated an administrative offense rather than an “other” type of offense. First presenters were those that showed the strongest differences from younger postillness starters in terms of index offense characteristics. They had lower odds of drug/alcohol use at the time of the offense, but twice the odds of suicidal ideation. They also had 1.5 times the odds of weapon use and had almost twice the odds of victimizing a family member and of victimizing an acquaintance” (p. 88).

“We hypothesized that the postillness starters would have a shorter trajectory through the Review Board system than preillness starters. Similarly, when compared with the preillness starters, the postillness starters were expected to experience more successful community reintegration in terms of recidivism. Findings were generally consistent with these expectations. In fact, differences were observed regarding outcomes while under the purview of the Review Board and in the reoffense rates after the NCRMD verdict. Adolescent preillness starters were similar to younger postillness starters, with the exception that they were more likely to recidivate. Adult preillness starters had lower relative odds of receiving a detention order in hospital as their first Review Board disposition and higher relative odds of receiving an absolute discharge as their first Review Board disposition rather than a conditional discharge compared with younger postillness starters. They also had higher rates of absolute discharge and release from detention before the end of the follow-up. Older postillness starters and first presenters showed similar patterns in terms of outcomes compared with younger postillness starters. They both had higher relative odds of receiving an absolute discharge as their first Review Board disposition rather than a conditional discharge compared with younger postillness starters. They were both less likely to display violent behaviors, suicidal behaviors, and to use substances between Review Board hearings. Their rates of absolute discharge and release from detention were also higher than for younger postillness starters. They were also less likely to recidivate” (p. 88).

“Our results replicated findings from prior studies, which point toward traditional criminogenic needs (e.g., substance abuse, personality disorder, extensive criminal history) among forensic patients with preillness onset of crime. Our findings also reinforce prior research revealing the heterogeneity among adult criminal-onset individuals that could not have been observed in studies using the original two-group model. These findings help to disentangle unique and important needs among subgroups of forensic psychiatric patients that could be targeted in prevention, rehabilitation and risk management strategies” (p. 89-90).

“As expected from previous studies, comorbid substance use disorder was less common among adult postillness starters and first presenters and more common among adolescent preillness starters. Older postillness starters and first presenters were also less likely to have used alcohol or drugs at the time of the offense. The role of mental health symptoms vis-a`-vis substance use in relation to the index offense requires further consideration for a verdict of NCRMD. While substance use may be an indicator of antisociality and moderate the relationship between mental illness and crime, it can also be a consequence of mental illness because of self-medication or increased vulnerability to substance use. Thus, consistent with prior research, the added presence of substance use disorders and personality disorders appears to be important in differentiating mentally ill individuals who are at elevated risk of criminal versatility and recidivism (i.e.,early starters/preillness offenders)” (p. 90).

“Younger postillness starters had similar risk of problem behaviors while under the Review Board to preillness starters, and were less likely than all the other groups (except for adolescent preillness starters) to be discharged by the end of the study. However, their risk of recidivism is not particularly high which bears the question regarding the justification of time spent under the Review Board purview. They seem to be getting into trouble for noncompliance and substance use issues but do not seem to pose a particularly great risk of violence. As hypothesized, the older postillness starters and the first presenters had very low risk of problem behaviors while under the Review Board, and a low risk of recidivism” (p. 90).

“Lastly, this study sheds light upon a little-known group of offenders, who had no prior contact in either the justice and mental health systems. Although Hodgins (2008) had discussed and examined “first offenders,” a group of mentally ill offenders who unexpectedly commit a very serious crime without any prior signs of antisociality, they seemed to have already had contact with mental health services, which is not the case of the first presenters. First offenders and first presenters are similar in that they both have better psychosocial functioning than the other groups. This group is of particular clinical interest and requires more investigation. This group is more likely to be found NCRMD following a tragic event involving a family crisis or in a highly emotionally distressed situation, including suicidality” (p. 90).

“These results suggest that the early and late starter model is relevant to risk management, as it is associated with violence and criminal recidivism. The new five-group model provides more refinement in potential developmental and services trajectories of this heterogeneous population. Examining potential typologies can be helpful in attempting to understand pathways to offending among individuals entering forensic services and targeting intervention strategies. For example, a clear focus on antisocial attitudes and behavior would be optimal for the adolescent preillness starters whereas more focus may be put on illness management for the older postillness starters and the first presenters. The model is associated with dynamic risk factors such as failure to comply with conditions, substance use, and medication noncompliance, which are all potential targets for ongoing intervention” (p. 90-91).

“It is an accepted principle of the risk-need-responsivity model that low-risk individuals should receive minimal or less intense services because providing more intense services is unnecessarily costly and raises the potential of iatrogenic effects. Given the costs of forensic hospitalization from an economic perspective and in terms of potential loss of employment, income, housing, and relationships, which are all protective against further crime, particular attention to duration of forensic hospitalization for the the older postillness starters and first presenters would appear warranted. Consideration should be given to the potential iatrogenic effect of undue (lengthy) hospitalization or Review Board purview if the patient responded well to medications and psychiatric symptoms have abated; particularly, if the patient does not have prior criminal justice involvement and/or other antisocial behavior. While they were more likely to be absolutely discharged by the end of the study, it is possible that forensic services are “over managing” these two groups by keeping them under the Review Board mandate longer than necessary” (p. 91).

“Although individualized risk assessments are essential to address specific needs and responsivity issues, knowledge of case-mix can potentially guide resource allocation and flag whether alternative, less expensive responses (e.g., diversion from the justice system to mental health services, even greater use of absolute or conditional discharge) might be warranted. Appropriate discharge from forensic services may be important to ensure optimal health and prevent escalation of criminal risk, reduce to the greatest extent possible the stigma associated with the forensic label and from a system-level can help to reduce the considerable back-log of beds common in forensic contexts. The model could provide a clinical-administrative tool in service planning at the organizational level” (p. 91).

As always, please join the discussion below if you have thoughts or comments to add!