Missed preventive opportunities

There is a paucity of research on the timing and utilisation of mental health services prior to the commission of homicide among individuals with SMI. To the extent that some of these homicides occur in the context of first episodes of psychosis or suboptimally treated psychosis, the provision of earlier and more intensive services could have significant harm reduction potential. Still, it is notable that improvements to the quality and availability of community mental health services in many industrialised nations after deinstitutionalisation have not affected the rate of SMI-associated homicide as would have been expected.^{Reference Simpson, Penney and Jones14,Reference Simpson, Mckenna, Moskowitz, Skipworth and Barry-Walsh19,Reference Large, Smith, Swinson, Shaw and Nielssen33–Reference Taylor and Gunn35} Similarly, the more widespread availability of community care and first-episode services would have been expected to reduce a substantial proportion of homicides related to first-episode events. This has also not materialised.^{Reference Flynn, Ibrahim, Kapur, Appleby and Shaw36,Reference Nielssen, Lyons, Oldfield, Johnson, Dean and Large37} At present, population-based data are required to better link the effects of these early and more assertive community interventions to homicide rates among those with SMI. Systems- and other macro-level variables (e.g. the relative restrictiveness of the mental health legislation in effect) need to contextualise these analyses.

Earlier meta-analytic work by Nielssen & Large^{Reference Nielssen and Large5} suggested that approximately 40% of homicides committed by people with a psychotic illness occur during the first episode of psychosis and prior to treatment initiation. Results from our study suggested that a minority of homicides occurred in the context of a first episode of psychosis, since approximately three-quarters (75.7%) of the sample had prior (lifetime) contact with mental health services and a similar proportion (74.2%) had one or more in-patient admission preceding the index offence. Although this contact may have been for other clinical concerns unrelated to psychosis and/or risk of violence, they nevertheless represent potential missed opportunities to intervene more assertively and prevent harm – particularly for those contacts with services occurring in close temporal proximity to the homicide offence.

In this respect, we found that half (48.1%) of the individuals in our sample had a relationship with a mental health service provider in the year before the index homicide offence (including 28.4% with an in-patient admission), but that most of this contact was infrequent or sporadic. Unfortunately, the exact reason(s) as to why this contact may have been insufficient (e.g. in terms of frequency, intensity or consistency) could not be determined from the data available. A large proportion of this sample presented with ongoing substance use concerns, which may have contributed to a loss of continuity and/or adherence to care in the community prior to forensic tenure, and emphasises the need for multimodal interventions that can address the needs of those presenting with significant comorbid substance use.

Current findings align with prior studies conducted in samples of homicide offenders with schizophrenia (but not necessarily forensically dispositioned, i.e. NCRMD or UST) showing elevated rates of missed contacts with service providers in the time preceding the offence, as well as non-adherence to treatment plans.^{Reference Oram, Flynn, Shaw, Appleby and Howard15,Reference Baird, Webb, Hunt, Appleby and Shaw18} Further, despite the finding that many individuals in this sample had been ill for many years – on average, 8 years – prior to the commission of the index offence and that the majority had one or more prior psychiatric admissions, just 38.3% had a prescription for medication at the time of the homicide. Of this group, less than one-third were deemed to be sufficiently adherent. Golenkov et al,^{Reference Golenkov, Large, Nielssen and Tsymbalova38} in their sample of 133 homicide offenders with schizophrenia in the Chuvash Republic of Russia, similarly observed many cases of missed appointments with treating agencies and non-adherence with antipsychotic medication in the period before the homicide.

Patient and victim types

The current sample was comprised almost exclusively of young men with psychotic illness. Women comprise approximately 15% of the forensic patient population in Canada, but our sample contained just 7 women (6%). This reflects that one of our data collection sites was a male-only facility, but is also consistent with the finding that the most serious forms of violence still tend to be dominated by men. Individuals in this sample otherwise showed rates of personality and substance use disorders, as well as rates of prior criminal justice involvement, comparable to those for the larger and more heterogeneous population of forensic service users in Canada^{Reference Chaimowitz, Moulden, Upfold, Mullally and Mamak39–Reference Penney, Seto, Crocker, Nicholls, Grimbos and Darby41} (see Supplementary Table 1, available at https://doi.org/10.1192/bjo.2023.567). Perhaps indicative of better premorbid adjustment, individuals in this sample were less likely to be homeless or without fixed address at the time of the offence and more likely to be living with family members. Individuals in the current sample were found to have offended significantly more frequently against family members or someone known to them (e.g. a roommate or neighbour), and less frequently against strangers, compared with these larger studies of Canadian forensic service users, which have included all offence types.

Current findings regarding victim type echo results from other jurisdictions, for example in an updated sample of Russian homicide offenders with schizophrenia^{Reference Golenkov, Large, Nielssen and Tsymbalova25} (n = 179) where most victims were familial (57.0%) or known (39.5%) to the individual and where just 3.5% of victims were strangers. Similarly, in a representative sample of Belgian NGRI acquittees, those who had committed homicide or attempted homicide (n = 105) were more likely to have a family member victim (59.8%) and less likely to have a stranger victim (15.5%) compared with NGRI acquittees who had committed less severe violence (n = 571; 32.3 and 32.8% for family and stranger victims respectively).^{Reference Jeandarme, Vandenbosch, Groenhuijsen, Oei and Bogaerts26} Family members are also injured more seriously than other types of victim,^{Reference Crocker, Nicholls, Seto, Charette, Côté and Caulet40} a finding that may reflect a higher threshold for reporting crimes within the family and stronger emotional ties resulting in more unrestrained violence.^{Reference Nordström and Kullgren42}

Persons with SMI who are in contact with the criminal justice system often have smaller social networks than the general population and those networks consist mainly of family members.^{Reference Ter Haar-Pomp, Spreen, Bogaerts and Volker43} Thus, family members – the majority of whom are parents – are more likely to be victimised because they have the most contact with their relative and are also more likely to be in a caregiving role.^{Reference Jeandarme, Vandenbosch, Groenhuijsen, Oei and Bogaerts26} There may also be poor acceptance of mental illness among both patients and their family members, resulting in inadequate monitoring and treatment of symptoms.^{Reference Raymond, Léger and Lachaux44} Providing better supports and psychoeducation to family members caring for a relative with SMI thus appears critical to manage risk, reduce rates of intra-familial violence and alleviate the isolation and emotional burden often felt by these caregivers.^{Reference Rowaert, Vandevelde, Audenaert and Lemmens45,Reference Rowaert, Vandevelde, Lemmens and Audenaert46}

Age at onset of symptoms and at first offence

Most patients in this sample first experienced symptoms of SMI prior to, or contemporaneously with, their first point of contact with the criminal justice system. This included those for whom admission to forensic care for the homicide offence was their first contact with both the mental health and criminal justice systems. In contrast, less than one-quarter of the sample had criminal justice contacts before symptom onset. The closest comparator study is of 109 Finnish persons with schizophrenia who were accused of homicide between 1983 and 2002,^{Reference Laajasalo and Hakkanen30} where similar proportions of early starters and late starters were observed. When compared with other studies of forensic patients in which varying proportions have committed homicide (Table 1), our percentage of early starters is relatively lower and our percentage of late starters relatively higher, as hypothesised. The proportion of late late starters in this sample is comparable with the two existing studies investigating this subgroup.^{Reference Simpson, Grimbos, Chan and Penney23,Reference van Dongen, Buck and van Marle24} In drawing these comparisons, it is also important to be mindful of the methods by which early-starter and late-starter groups were defined. Consistent with the method adopted in this study, Crocker et al,^{Reference Crocker, Martin, Leclair, Nicholls and Seto20} Jones et al,^{Reference Jones, Van den Bree, Ferriter and Taylor47} Simpson et al^{Reference Simpson, Grimbos, Chan and Penney23} and van Dongen et al^{Reference van Dongen, Buck and van Marle24} constructed subgroups using indicators of illness onset in relation to offending onset. In contrast, other studies^{Reference Laajasalo and Hakkanen30–Reference Tengström, Hodgins and Kullgren32} have restricted their categorisation of early-starter and late-starter groups to the onset of behavioural problems (e.g. presence of conduct disorder, age at first conviction).

The finding of comparatively more late starters and late late starters in this sample suggests that many persons who are admitted to forensic care for homicide act primarily in the context of illness, with few other aggravating risk factors. This is supported by findings that individuals who commit homicide while actively unwell, compared with those without symptoms at the time, are less likely to have previous convictions for violence or histories of substance misuse.^{Reference Oram, Flynn, Shaw, Appleby and Howard15} Forensic patients who have committed homicide, compared with patients committing other types of offence, are also found to have less severe psychopathology, fewer previous compulsory admissions and criminal convictions, and generally better premorbid adjustment in terms of employment and socioeconomic status.^{Reference Fioritti, Ferriani, Rucci and Melega48} Because the management of risk in these cases will be largely focused on illness stabilisation, successful attenuation of symptoms will substantially and directly mitigate risk. In contrast, those who have been admitted to forensic care for homicide but who also have premorbid behavioural adjustment problems and legal contacts may have multiple areas of risk that are more complex to manage effectively (e.g. personality disorder, substance misuse and problems with housing, employment and relationships).

Application of an age at onset framework thus permits a greater clarification of risk management approaches that may need to be tailored according to the timing of illness vis-à-vis violence and criminality. A strategy targeting mental healthcare will probably be differentially effective for the sizeable proportion of late starters/late late starters who may otherwise be at increased risk for violence in the context of emerging psychotic symptoms. By contrast, multi-modal interventions that span illness-related and criminogenic needs will be required to address the more complex and long-standing risk issues present among the early-starter group.

Limitations

Limitations of this study included the fact that this was not a population-based study but rather a purposive sampling of patients admitted to forensic care at one of three provincial forensic hospitals in Ontario. Unlike other studies,^{Reference Flynn, Ibrahim, Kapur, Appleby and Shaw36,Reference Skipworth, Bevin, McKenna, Simpson, Brinded and Pearson49} we do not have a good measure of the proportion of homicide offenders with SMI who receive a correctional (versus forensic) disposal in Ontario or Canada. This information would have better contextualised the frequency with which the ‘not criminally responsible on account of mental disorder’ (NCRMD) defence is successfully applied in cases of homicide involving an accused person with a psychotic illness. The generalisation of current results to other jurisdictions may therefore be limited and our comparison with existing population-based studies of forensic patients must be evaluated with this in mind. This was also a predominantly White male sample, with few females and other ethnicities represented. Questions remain as to whether gender and ethnicity affect the trajectory and motivations for serious violence in the context of SMI.

Data were missing for some of the historical variables examined (i.e. prior psychiatric admissions (available cases: n = 89) and criminal justice contacts (n = 82)) and the coding of frequency of contact with mental health services was based on information available in the health record. Further, sample sizes became small when partitioned by age at onset of illness and offending. We also note that some individuals in this sample had committed their index offence many years ago (one-third of the sample had their NCRMD finding 20 or more years ago), creating the possibility that the questions examined here in relation to service utilisation and age at onset show time-based effects.

Implications

Many individuals who are admitted to forensic care for an offence of homicide have had inadequate mental healthcare leading up to this point, highlighting what might have been missed opportunities to prevent some episodes of violence through improved community mental healthcare. Effective management of violence risk conferred by mental illness should follow best practice standards in the delivery of evidence-based interventions generally, and targeted to commonly observed risk factors in this population (e.g. medication-assisted cognitive–behavioural therapy for substance use relapse prevention; evidence-based components of first-episode psychosis services such as community and family group psychoeducation, peer support and acute-phase care^{Reference Addington, McKenzie, Norman, Wang and Bond50}; specialised and supported employment and vocation services). Interventions will need to be multi-modal in nature, encompassing psychological and pharmacological therapies, in order to effectively address both illness-related and criminogenic domains.