Time Sensitive Dynamic Risk Factors May Help Predict Adverse Outcomes When Releasing Forensic Patients Into The Community

Stephanie R. Penney Centre for Addiction and Mental Health, Toronto, Ontario, Canada, and University of Toronto
Lisa A. Marshall Ontario Shores Centre for Mental Health Sciences, Whitby, Ontario, Canada, and University of Toronto
Alexander I. F. Simpson Centre for Addiction and Mental Health, Toronto, Ontario, Canada, and University of Toronto

Individuals with serious mental illness (SMI; i.e., psychotic or major mood disorders) are vulnerable to experiencing multiple forms of adverse safety events in community settings, including violence perpetration and victimization. This study investigates the predictive validity and clinical utility of modifiable risk factors for violence in a sample of 87 forensic psychiatric patients found Not Criminally Responsible on Account of Mental Disorder (NCRMD) transitioning to the community. Using a repeated-measures prospective design, we assessed theoretically based dynamic risk factors (e.g., insight, psychiatric symptoms, negative affect, treatment compliance) before hospital discharge, and at 1 and 6 months postdischarge. Adverse outcomes relevant to this population (e.g., violence, victimization, hospital readmission) were measured at each community follow-up, and at 12 months postdischarge. The base rate of violence (23%) was similar to prior studies of discharged psychiatric patients, but results also highlighted elevated rates of victimization (29%) and hospital readmission (28%) characterizing this sample. Many of the dynamic risk indicators exhibited significant change across time and this change was

related to clinically relevant outcomes. Specifically, while controlling for baseline level of risk, fluctuations in dynamic risk factors predicted the likelihood of violence and hospital readmission most consistently (hazard ratios [HR] = 1.35–1.84). Results provide direct support for the utility of dynamic factors in the assessment of violence risk and other adverse community outcomes, and emphasize the importance of incorporating time-sensitive methodologies into predictive models examining dynamic risk.

community violence, dynamic risk, mental illness, risk assessment, victimization

“Several large-scale studies have investigated whether individuals with serious mental illness (SMI; i.e., psychotic or major mood disorders) are at greater risk for violence as compared with members of the general population. Much of this research suggests that persons with SMI, particularly those experiencing psychotic disorders, are at elevated risk for violence toward others, a finding that has emerged across diverse samples including incarcerated offenders, civil psychiatric patients and in population-based birth cohorts. A smaller number of prospective studies have assessed community violence among psychiatric patients following hospital discharge, reporting base rates between 17% and 38% occurring within the first two years of community tenure, and lower rates for more serious forms of violence (e.g., those resulting in a formal conviction; 3% to 10%)” (p. 374).

“Studies of the prevalence of violent reoffending among forensic patients transitioning to the community report comparable base rates, although the range is somewhat wider (3% to 36%), likely owing to the greater variability in the measurement of violence (e.g., self-report vs. official records), sample type (e.g., patients released from maximum vs. minimum secure hospitals), and lengths of community follow-ups (6 months to 15 years) that characterize these studies… Nevertheless, establishing reliable prevalence rates for community violence is a critical area of study within forensic psychiatric samples given that the perception of threat to public safety posed by such patients in the community is often higher. Furthermore, it is not often recognized that individuals with SMI, including forensic service users, are at greater risk of being victimized by crime and violence than they are of perpetrating violence” (p. 374 – 375).

“It is possible that specific symptoms of SMI such as psychosis, when active or intensifying, may cause violence directly by focusing, destabilizing, or disinhibiting behavior; however, such symptoms can also motivate violence indirectly by increasing stress or exposure to provocation and conflict. Furthermore, other symptoms of SMI (e.g., negative symptoms of psychosis) may have a nonsignificant or inverse relationship with violence… Recent literature points to the necessity of assessing changes in risk status over time to accurately conceptualize an individual’s risk and focus treatment more effectively. The measurement of change in dynamic risk indicators can also facilitate better-timed interventions, as well as help evaluate the effectiveness of already implemented risk management strategies. It may also be an important first step in advancing the study of risk factors to the identification of risk mechanisms, or even causal risk mechanisms” (p. 375).

“The study of how dynamic risk factors operate in real time has the potential to advance the field of violence risk assessment and illuminate processes through which risk factors increase the probability of clinically relevant outcomes such as violence or victimization. Few studies have examined dynamic variables in this manner, and essentially none have assessed changes in risk over key transition points such as discharge from hospital and the commencement of community living… the present study examines how modifiable risk factors for violence fluctuate over time, and how they impact an individual’s likelihood of engaging in or experiencing violence, as well as being readmitted to hospital. We further examine how dynamic risk indicators function in relation to static risk, to assess whether dynamic factors offer incremental utility over static factors, as well as examine whether the magnitude of change observed in dynamic factors varies as a function of static risk” (p. 376).

This study utilized a prospective, repeated measures design to follow 87 forensic patients as they were discharged to the community. The participants were hospitalized after having been found to be not criminally responsible due to a mental disorder (i.e. an insanity defense). “Most (82.8%) participants had been charged with one or more violent offenses as part of the predicate offense. The most frequent primary diagnosis was schizophrenia (81.6%), with more than half (57.5%) of the sample diagnosed with a comorbid substance use disorder. Mood disorders were infrequently represented in this sample, with 5.7% of the sample diagnosed with major depression and 5.7% with bipolar disorder” (p. 376).

“This prospective study utilized a repeated-measures design to investigate fluctuations in dynamic risk factors at three discrete time points. Although few episodes of serious violence occurred during the study period, dynamic risk factors were found to vary over time and be related to a number of relevant clinical outcomes, some of which may serve as proxies for violence risk and recidivism” (p. 381).

“Although no official reports of violence were recorded, self-report and health record information indicated rates of violence similar to previous studies of discharged civil and forensic patients. From a public safety perspective, it is relevant that no patient in the community engaged in serious violence resulting in serious injury or life-threatening harm to another person. Turning to victimization experiences, findings from this sample are consistent with prior studies indicating that individuals with SMI are more vulnerable to being victims of violence and crime than members of the general population, and that the experience of recent victimization may increase one’s risk of perpetrating violence in the future” (p. 382).

“In the context of a logistic regression model, dynamic risk indicators failed to incrementally enhance the variance accounted for by static variables, a finding that differs from recent studies in forensic samples. However, when the trajectory of dynamic variables was accounted for, time-dependent scores on the dynamic subscales of the HCR-20V3 were significantly related to violence and rehospitalization after controlling for static risk. Furthermore, when fluctuations in psychiatric symptoms were examined, they predicted increased victimization rates. These findings align not only with the large body of literature documenting the predictive validity of the HCR-20V3 in relation to violence, but further suggest that risk factors for rehospitalization within forensic samples may share significant overlap with risk factors for violence” (p. 383).

“Decreases in dynamic risk as measured by the HCR-20V3 were only apparent in patients with an already low level of static risk. This finding contrasts somewhat to [earlier research] in incarcerated sexual offenders, who found that higher-risk offenders exhibited greater change as compared to lower risk offenders. These investigators suggested that higher risk offenders with many needs areas have more room for improvement (e.g., to reduce their risk scores), in contrast to lower risk offenders. In the current sample of forensic patients, it may be that among those with an already low level of static risk, it is comparatively easier to effect change and improvements in dynamic risk indicators surrounding acute mental health concerns and ongoing treatment compliance. In contrast, for those individuals with a higher baseline level of static risk, effecting change may be more difficult as the dynamic risk factors are underpinned by both illness-related factors as well as more enduring criminogenic ones (e.g., a lengthier history of employment instability and substance misuse, personality disorder)” (p. 383).

“Over a quarter of participants were rehospitalized within 12 months of discharge. Rehospitalization is a clinically significant outcome for patients and an indicator of overall service effectiveness. The readmission rate reported here may be indicative of timely intervention by outpatient teams to manage a perceived increased risk for violence or some other adverse outcome. At the same time, hospital readmission is not typically viewed as a positive outcome indicator, and may signal challenges that forensic service users often face upon transitioning to the community (e.g., unstable housing, a lack of social or family support, increased stress). Furthermore, high rates of readmission to inpatient units translate into increased costs and may discourage the patient and his or her caregivers if readmission is equated with treatment failure. Although prior studies have examined risk factors for rehospitalization in general medical or psychiatric samples, there have been few extensions into forensic populations where the legal ramifications associated with a readmission are often more consequential. Results presented here suggest that risk factors originally intended to predict violence and offending have utility in predicting the likelihood of hospital readmission as well” (p. 382).

“Dynamic risk factors in the domains of psychiatric symptoms and negative and positive affect were assessed via self-report during the first six months after discharge from hospital. In contrast, dynamic variables appearing on the HCR-20V3 were assessed by researchers, who relied on the participants’ self-report as well as clinical documentation in the health record. Ratings on the HCR-20V3 appeared to have stronger predictive ability, particularly in terms of forecasting outcomes of violence perpetration and rehospitalization. In contrast, all self-report measures used in the current study failed to predict any of the adverse outcomes, with the exception of the BPRS (for which final ratings are similarly generated by researchers and based on interview data plus file information). Thus, results point to the relatively higher utility of expert rater data as compared with data from the self-report questionnaires in terms of predicting community outcomes” (p. 383).

“From a clinical practice perspective, results emphasize the importance of conducting repeated risk assessments on patients at specified intervals, and to recognize the importance of changes in risk factors, rather than their simple presence or absence, in terms of prognosis and outcome. Future studies may consider increasing the frequency of data collection intervals over a longer period of time to better map the trajectory of change among dynamic risk indicators and more accurately connect changes in dynamic variables to violent incidents or other proxies of risk and recidivism. Finally, to move closer to causal models of violence perpetration and other outcomes, studies must also seek to document that targeted interventions have the ability to influence dynamic risk factors, and that this influence ultimately reduces the likelihood of an adverse outcome” (p. 384).

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