Background
There is a well-recognised but poorly understood association between schizophrenia and homicide offending. Meta-analyses of available studies show a fourfold increased risk of homicide offending among men with schizophrenia and an eightfold increased risk among women compared with rates of homicide offending by people without schizophrenia,Reference Fazel, Gulati and Linsell1 and that a pooled estimate of 6.5% of homicide offenders have schizophrenia.Reference Large, Smith and Nielssen2 The traditional view has been that the elevated homicide risk among people with schizophrenia is because homicide risk is intrinsic to clinical manifestations of schizophrenia, particularly persecutory delusional beliefs.Reference Tanay3 This view has contributed to an emphasis on the treatment of psychotic symptoms and on measures to contain people with schizophrenia who are perceived as potentially violent.Reference Tanay3 This view was supported by studies suggesting that homicide rates by people with severe mental illness would be more or less fixed over time and between regions, in line with the stability in prevalence of severe mental illness.Reference Coid4,Reference Taylor and Gunn5 However, a 2009 meta-analysis of 18 available population-based studies found that rates of homicide by people with schizophrenia are heterogeneous and are strongly correlated to rates of other homicide. This study contradicts earlier views about the stability and causes of rates of homicide by people with schizophrenia and suggests that other explanations, including the possibility that rather than being intrinsic to the psychopathology of schizophrenia, people with schizophrenia might be more vulnerable to the same risk factors that determine homicide rates in the general community.Reference Large, Smith and Nielssen2 The implications of this finding include that homicide by people with schizophrenia might best be prevented by limiting the exposure to factors known to be associated with violence in the community, such as access to firearms and limiting alcohol and other drug use, as much as by treatment and hospital admission. One limitation to the 2009 meta-analysisReference Large, Smith and Nielssen2 was that it examined aggregated homicide rates over the duration of the included studies and did not examine changes in rates or in the relationships between the rates of homicide by people with and without schizophrenia over time.
Several studies have examined the trends in homicide committed by a broader class of people considered to have a mental disorder. In England and Wales, rates of homicide by people with mental illness (defined as the total of those judged to have diminished responsibility, infanticide, or as being unfit to stand trial or not guilty due to mental illness) rose steeply with rates of other homicide after World War Two, before falling just as steeply, while the rate of other homicides continued to rise.Reference Large, Smith and Swinson6 Another pattern was observed in New Zealand between 1970 and 2000 where the total homicide rate rose annually between 1970 and 1990 before falling in the last decade of the study, whereas the proportion of homicide by people defined as ‘mentally abnormal’ (not guilty due to mental illness, infanticide and those unfit to stand trial) fell throughout the study, initially because homicide by people who were not mentally ill was rising and later because mental illness verdicts were falling faster than other court outcomes.Reference Simpson, McKenna and Moskowitz7
In Denmark, homicide by people with psychosis (defined using ICD-8 criteria) were reported to be stable during a period of rising homicide rates between 1956 and 1970, but then increased in parallel with total homicide rates between 1970 and the mid-1980s.Reference Gottlieb, Gabrielsen and Kramp8 In Iceland homicide by people with mental illness (defined as schizophrenia, manic depression and morbid jealousy) rose more quickly than rapidly rising rates of other intentional homicides in the period 1970 to 1984.Reference Gudjonsson and Petursson9 More recently, a study from Ontario, Canada found a non-significant fall in homicide offending by people who were mentally ill during a period when the rate of other homicide fell more.Reference Penney, Prosser and Grimbos10 To date the only study that focused on changes in rates of homicide by people with schizophrenia found a rise in the absolute number of homicide offenders with schizophrenia during a period in which other homicide rates fell in England and Wales.Reference Swinson, Flynn and While11 The major limitation in interpreting this group of studies as an indicator of the association between homicide offending by people who are mentally ill compared with those who are not mentally ill is the differences between studies in the way mental illness is diagnosed and defined in law between jurisdictions and over time.Reference Large, Smith and Nielssen2,Reference Flynn, Ibrahim and Kapur12
Aims
We have previously reported on the characteristics of offenders with schizophrenia and other mental disorders from The Chuvash Republic of the Russian Federation, a region with both a high total homicide rate and a high rate of homicide by people with schizophrenia.Reference Golenkov, Tsymbalova, Large and Nielssen13–Reference Golenkov, Large, Nielssen and Tsymbalova15 Previously we found that almost half of all homicide offenders had at least one mental disorder, including alcohol dependence experienced by 16% of offenders and schizophrenia experienced by over 4% of offenders.Reference Golenkov, Large, Nielssen and Tsymbalova15 In this paper we report on the trends in rates of homicide by people with schizophrenia and without schizophrenia (other homicides) during a period of 40 years when the rate of total homicide rose and fell dramatically. We hypothesised that if rates of homicide by people with schizophrenia were fixed over time that they would not be correlated with the rates of homicide by people without schizophrenia and that the proportion of homicides by people with schizophrenia would fall while other homicide rates were rising and would rise when other homicide rates were falling. In other words, the rate of homicide by people with schizophrenia did not change when the overall rate of homicide changed .
Method
Sample
The study examined a data-set of all homicides in Chuvash Republic of the Russian Federation between January 1981 and December 2020. The Chuvash Republic has a population of over 1.2 million living in an area of around 18 000 km2, centred in the city of Cheboksary on the Volga River about 650 km east of Moscow. A third of the population live in Cheboksary and 40% are considered to be rural. During the period of the study, and as in the rest of Russia, people in the Chuvash Republic who are charged with a homicide offence are referred for psychiatric evaluation under Section 21 of the Russian Penal Code.Reference Golenkov, Tsymbalova, Large and Nielssen13–Reference Bukhanovsky and Gleyzer16 Referral for judicial psychiatric examination by a panel of three psychiatrists is at the discretion of the investigators and the court. Some accused are not referred because there are no doubts about their sanity, and some individuals who are accused refused to be examined. However, most accused are examined, and the sample includes the majority of offenders referred for a formal evaluation, including all offenders suspected of having a psychiatric disorder. Data was also available about homicides that were immediately followed by the suicide of offenders known to mental health services.
Case identification and data extraction
Demographic, clinical and criminological data were extracted from the case files, which included the judicial psychiatric report, hospital records, out-patient notes and legal documents. About one-third of the offenders had been personally examined by either A.G. or A.T. The demographic data extracted were age, gender, occupation, marital status and years of education and the clinical data included psychiatric diagnosis, age at onset, previous treatment, pattern of symptoms, history of brain injury, and substance misuse, and the criminological data included previous convictions, method of homicide and judicial outcome. The psychiatric diagnoses were made using the version of the ICD that was current at the time of the judicial examinations. For the purpose of this study the psychiatric diagnoses in each individual was made by retrospective chart review using ICD-10 criteria.17
Statistical analysis
Non-age-adjusted homicide rates were calculated using the number of annual homicide incidents and the annual population of Chuvashia, as measured by the census or imputed using linear assumptions in the intervening years. The association between rates of homicide by people with schizophrenia and other homicide and the association between the rates of other homicide and the proportion of all homicides committed by people with schizophrenia were examined with a Pearson correlation coefficient using untransformed homicide rates.
The changes in homicide rates by people with and without schizophrenia and changes in the proportion of homicides (homicides by people with schizophrenia/total homicides) committed by people with schizophrenia over time were examined using Kendall's Tau rank correlation coefficient over the whole period of the study and over two equal periods (1981–2000 and 2001–2020) in which the homicide rates were, respectively, rising and falling. In order to illustrate the association between homicide in schizophrenia and other homicide, rates were displayed as Log10 (1 + homicide rate per million population per annum) so as to avoid the scale minimisation effects that can arise because of the low base rate of homicide offending by people with mental illness.Reference Large, Smith and Nielssen18 Possible differences in the characteristics of male and female homicide offenders with schizophrenia were examined using t-test and chi-square (χ2). When one or several cells contained five individuals or less, a Fisher's Exact test was used. All tests were performed in a two-tailed form. The data were analysed using SPSS Version 27 (SPSS Inc. Chicago, IL, USA).
Ethical approval
Approval for the research was obtained from the Ethics Committee of the Medical Faculty of Chuvash State University, and the Chuvash Association of Psychiatrists, Narcologists, Psychotherapists and Medical Psychologists (a branch of the Russian Society of Psychiatrists). As the study only uses statistical information, no oral or written consent was taken from the individuals of the study.
Results
Rates of homicide
In the 40 years from the beginning of 1981 and the end of 2020, 5741 people faced legal proceedings for a homicide offence, and 3410 homicide offenders were referred for psychiatric evaluation. Over this period there were more homicide victims than offenders because the police did not solve every case, some offenders had multiple victims, and some offenders died by suicide or died prior to being examined. A small number of offenders from other parts of Russia were examined elsewhere. The homicide clearance rates, defined as an offender being identified by the police, were more than 90% in the period of the study. Of those subjected to judicial examination, 46.7% were found to have at least one mental disorder, and 7.7% were found to be not responsible for their actions because of mental disorder according to the Russian Criminal Code.Reference Large, Smith and Nielssen18
During the study period 179 homicide offenders were identified as having schizophrenia, including 14 recidivist offenders who committed both homicides between 1981 and 2020 (Table 1). The population of Chuvashia rose slightly from 1.304 million in 1981 to 1.347 million in 1993, and subsequently fell by about 10% to 1.224 million in 2020.
Table 1 Homicide by people with schizophrenia in Chuvashia 1981–2020
a. Homicide–suicide known schizophrenia n = 2 (2011 – female, 2015 – male).
b. First homicide.
c. Victims.
During the four decades of the study the rates of homicide by people without schizophrenia rose and fell:
(a) 1981–1990: average 8.9 per 100 000 person-years;
(b) 1991–2000: average 16.4 per 100 000 person-years;
(c) 2001–2010: average 12.5 per 100 000 person-years;
2011–2020: average 6.1 per 100 000 person-years.
Rates of homicide by people with schizophrenia also rose and fell in the same periods:
(a) 1981–1990: average 0.28 per 100 000 person-years;
(b) 1991–2000: average 0.41 per 100 000 person-years;
(c) 2001–2010: average 0.47 per 100 000 person-years;
(d) 2011–2020: average 0.21 per 100 000 person-years.
The annual rate of homicide by people with schizophrenia was strongly correlated with the rate of other homicide (Pearson's r = 0.503, r 2 = 0.253, F = 12.9, d.f. = 38, P = 0.001) (Fig. 1). There was no significant association between the rates of other homicide and the proportion of all homicides committed by people with schizophrenia (Pearson's r = −0.21, r 2 = 0.045, P = 0.19).
Fig. 1 Homicide rates by people with schizophrenia and other homicide in Chuvashia 1981 to 2020.
In the period 1981–2000 rates of homicide by people with schizophrenia rose significantly (tau (τ) = 0.33 P = 0.04) as did those without schizophrenia (τ = .67 P = <0.0001). In the period 2001–2020 rates of homicide by people with schizophrenia fell significantly (τ = −0.35 P = 0.03) as did those by people without schizophrenia (τ = −0.86 P < 0.0001) (Fig. 1).
The proportion of homicides by people with schizophrenia did not change over the 40 years of the study (τ = 0.15, P = 0.18), during the period 1981–2000 when rates of homicide were rising (τ = −0.05 P = 0.75) or in the period 2001–2020 when rates of homicide were falling (τ = 0.06 P = 0.7)
Characteristics of homicides by people with schizophrenia
In total 160 homicide incidents were committed by men with schizophrenia and 19 (10.7%) by women with schizophrenia. Adolescents (defined as individuals aged between 12 and 19 years) with schizophrenia committed 11 (6.2%) homicide offences (Table 1). There were 18 recidivist homicides by people with schizophrenia, 17 of which took place in rural areas, including 14 who committed their first and second homicide in the period of the study and 4 whose initial homicide was before 1981. There were 15 homicides with multiple victims, including 3 by women, and 2 homicides followed by suicide by people who were known to have schizophrenia. Thirty-one offenders were sentenced to indefinite hospital care, and the remainder received fixed terms (Table 2).
Table 2 Characteristics of homicide offenders with schizophrenia 1981–2020
a. Victims were children (39.13% v. 4.52% in males; d.f. = 1; χ2 = 31.351; P < 0.001).
b. Victims were acquaintances (neighbours) (43.50% v. 8.70% in females; d.f. = 1; χ2 = 8.914; P = 0.0028 as amended by Yeats).
Discussion
Main findings
This study examines the rates of homicide by people with schizophrenia over a period of 40 years in a region that had a dramatic rise and equally dramatic decline in the rate of homicide. The results confirm the well-recognised over-representation of people with schizophrenia among homicide offenders, but also suggest that the rate of homicide by those with schizophrenia is correlated with changes in the rate of other homicide and that the proportion of homicides committed by people with schizophrenia might not fall when other homicide rates are higher and might not rise as other homicides become less common.
Interpretation of our findings
Although the reasons for the five- to tenfold increase in homicide risk associated with schizophrenia remains uncertain, the findings of this study suggest that the increased rate is at least partly because of a greater exposure or greater sensitivity of people with schizophrenia to more general risk factors for homicide, including factors such as substance use and social disadvantage, rather than an intrinsic or inevitable propensity to lethal violence associated with the illness.
In this study the r 2 statistic indicated that about one-quarter of the variance in the rate of homicide associated with schizophrenia was explained by the changes in the overall rate of homicide. This can be compared with the results of a 2009 meta-analysis that found that three-quarters of the variance in rates of homicide in schizophrenia between jurisdictions was explained by the rates of other homicide.Reference Large, Smith and Nielssen2 Hence both cross-sectional and longitudinal studies suggest that homicide rates by people with schizophrenia are not fixed over time and between regions and that social factors that influence homicide by people without schizophrenia, for example, alcohol and other drug use, levels of violence in the community, and the efficiency of the police in preventing crimes influence the likelihood of a person with schizophrenia committing a homicide offence. However, the moderate R 2 value in the association between rates of homicide in people with schizophrenia and other homicide does suggest that there are other factors that influence the rate homicide in schizophrenia that operate independently to those associated with total homicide. These might include the availability and quality of psychiatric care, variation in the prevalence of schizophrenia, and the specific impact of substance use on people with schizophrenia.
The decline in the total rate of homicide in Chuvashia in the past 20 years is striking, and may be because of greater economic prosperity and the re-establishment of social order after the turbulent glasnost years. The decline may also be related to the ageing of the population, as many people of working age have moved elsewhere, and a rise in the proportion of women in the population. There was an increase in the number of people known to have schizophrenia in Chuvashia during the years of the study that might be as a result of an increase in the identification and reporting of patients with schizophrenia by a larger mental health workforce. Hence, it is also possible that an increase in availability of treatment contributed to the decline in homicide offences by people with schizophrenia, as suggested by the near elimination of recidivist homicide offences by people with schizophrenia in the past decade and the over-representation of recidivists in rural areas where services are less available. During the period of the study there were no changes to the laws governing psychiatric treatment, which has been proposed as a possible reason for changes in rates of homicide by people who are mentally ill in the USA, along with the quality of services.Reference Segal19
A decline in the consumption of alcohol in Chuvashia might also have been important. During the period of the study alcohol consumption in Chuvashia fell from being above the Russian national average to well below the national averageReference Matveev20 during a period in which the national per capita alcohol consumption fell by more than 40%.Reference Neufeld, Bunova and Gornyi21 About half of the offenders with schizophrenia (83/179, 46%) were reported to be intoxicated with alcohol at the time of the offence, a pattern that had not changed in the decade since an earlier report.Reference Golenkov, Large, Nielssen and Tsymbalova14
Strengths and limitations
Strengths of this study are the complete nature of the sample and related statistical information, consistency in the way the psychiatric examinations were recorded and an absence of changes to relevant sections of the legal code. Limitations of the study include that some homicide offenders did not receive a judicial examination, not all homicides were solved by the police, and the absence of information about the mental state of some of the homicide offenders who died by suicide before they could be examined.
Implications
A more reliable estimate of the extent to which rates of homicide by people with schizophrenia vary from other homicide rates might be obtained by meta-analysis once further longitudinal studies have been conducted in other jurisdictions. The extent of this correlation has implications for our understanding of the reasons for homicide by people with schizophrenia, and the degree to which systems of treatment for psychosis, and more general measures to reduce violence and homicide in the community might play a role in the prevention of homicide by people with emerging and established mental illness.
Data availability
The complete data-set is held by A.G. and is available for research purposes on reasonable request and ethical review.
Author contributions
A.G. conceived the study and compiled the data, M.L. performed statistical analysis and contributed to each draft, O.N. wrote the first draft of the paper, and revised each draft, and A.T. was involved in the collection and compilation of the data and reviewed the manuscript.
Funding
The study was not funded. Publication fee met by M.L.'s research account.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.