According to the World Health Organization (2014), there are approximately 800,000 deaths that occur annually due to suicide. Moreover, suicide has been reported to be one of the major causes of death among people aged 15 to 44 years and the second highest cause of death among individuals aged 10 to 24 years. Notably, these statistics do not include the number of attempted suicide cases.
The World Health Organization (2014) also reports that suicide accounts for about 1.3 percent of the diseases affecting individuals on a global scale. In Europe and America, mental health disorders have been reported to be major predisposing factors to suicide. The aim of the current article is to assess the effect of mental health disorders on homicide.
In the current article, the authors support the fact that mental disorders are risk factors for accidental death and suicide, but there is limited data to suggest their association with homicide (Crump et al. 2013, p. 1). The authors recognize that individuals presenting with mental illnesses could be at risk of homicidal death. The mentally ill patients are more likely to live in areas characterized by high substance use. Substance use has been reported to have an association with violence and criminal activities.
Furthermore, the mentally ill patients are more likely to socialize with other mentally ill patients. The population often views the mentally ill patients as vulnerable or dangerous and hence these patients could be subjected to homicide (World Health Organization 2014). The introduction in the current article is well outlined and the authors have clearly stated the objective of the research (Crump et al. 2013, p. 1).
The objective of the study is to determine the risk factors that predispose the mentally ill patients in Sweden to homicidal death. The authors are clear in justifying why they carried out the current research. They identify the fact that previous studies have reported that mentally ill patients have a six fold risk of homicidal death compared to healthy individuals.
However, the authors point out that the studies carried out in the past had two major limitations and they include; inadequate sample size, and selection biases. Furthermore, the introduction recognizes how the authors overcame the limitations identified in previous studies.
In an effort to deal with potential confounding factors, the authors used nationwide data from all the healthcare facilities; this ensured that all the mental illness cases had the same probability of being included in the study. In summary, the introduction is well outlined and clear and hence easy to understand what the study is about.
The study was carried out through a cohort study design that involved a national wide population of individuals aged 17 years and over (Crump et al. 2013, p. 2). The study population consisted of individuals that were living in Sweden in 2001. A total of 253, 516 subjects were included in the cohort. Although cohort studies are not the gold standard in epidemiological studies, they are effective in addressing multiple exposures (Euser et al. 2009, p. 214). This is true in the current studies as multiple risk factors are being assessed.
The population was recruited and followed up for the occurrence of homicidal deaths. Follow up studies such as the cohort study designs enable the investigators to establish the causal inference of the association. In the current study, it was possible for the investigator to confirm that the exposure (mental health) preceded the outcome (homicidal death). Cohort studies are used in the cases where randomization is not a feasible option as in the case of the current study.
Moreover, Euser et al. (2009, p. 216) indicate that cohort studies have higher efficiency and accuracy. In the introduction section of the article, the authors specify that the study was subject to confounders; the cohort study design enabled the researchers to control for confounding at the design and analysis stages of the study. Moreover, the investigators were able to mitigate selection bias. Additionally, the sample size was large enough to ensure generalizability of the research findings.
Based on the current research article, the study population was followed for eight years from January 1, 2001 to December 31, 2008 (Crump et al. 2013, p. 2). Homicide was the outcome of interest and it was determined through the International Classification of Diseases. The death registry in Sweden was used to establish the proportion of homicidal deaths in the population under study.
Moreover, deaths due to undetermined causes were excluded from the study. On the other hand, the exposure status in the study was mental disorder and it was ascertained through the registry system in both inpatient and outpatient hospitals in Sweden. Both primary and secondary diagnosis procedures were used to ascertain the mental disorders in the population under study.
In the introduction section, the authors identify limitation in previous studies with regard to the ascertainment of the exposure and outcome status of the study population. According to the authors the registries in the Swedish hospitals are 84 percent complete and this increases the accuracy of the information obtained. Using both primary and secondary diagnosis procedures prevented misclassification of the exposure status. Euser et al. (2009, p.217), reports that exposure misclassification provides results that are biased toward the null.
Therefore, investigators must ensure that the exposure and the outcome status are well defined. Furthermore, misclassification affects the relative risk between groups and biases the results toward the null value. This can result in spurious results associated with confounding factors. In summary, misclassification can affect the generalizability of the research findings.
The authors classified mental disorders according to the International Classification of Diseases and these were classified as substance use disorders, any mental disorder, delusional disorders and schizophrenia, depression, mood disorders, neurotic disorders, and stress related disorders.
Confounding can either be controlled at the design or the analysis phase of any epidemiological study (Kamangar 2012, pp. 508-509). According to Kamangar (2012, p. 508), investigators should identify the confounding variables at the design stage of the study in an effort to deal with these factors earlier on in the study. In the current research, the authors identify several demographic variables as confounding factors.
In the analysis phase of the study, confounders were adjusted through the Cox Proportional Hazard Regression model (Crump et al. 2013, p. 3). Kamangar (2012, p. 511), states that the Cox Proportion Hazards Regression Model is used when the outcome is characterized by ‘time to event’. In the current study, the homicide events were characterized by different follow-up durations. The association between the different mental disorders and the various demographic factors were assessed through the Cox Proportional Hazards Regression ratios.
All prospective cohort studies require measuring the time to event as the different subjects develop the outcome at different times during the follow-up period ( Euser et al. 2009, p. 215). During the regression modelling phase of the current study, the different disorders associated with the different individuals were modelled as independent variables.
Only confounders that that had the strongest effect on the relative risk were included in the final analysis. In summary, the methodology section of the article is well explained and very clear on the methods that were used to carry out the study. However, there is no indication whether the study was prone to loss to follow-up and how this was going to be mitigated in the current study.
Descriptive analysis of the 253,516 study subjects revealed that there were 410 deaths among men and 205 deaths among women (Crump et al. 2013, p. 2). This occurred after 54.4 million years of follow up. In total, there were 141 homicidal deaths in the study population. Moreover, the study established the crude mortality rate that occurred due to homicide which was 1.1 in the population. Specifically, the crude mortality rate was 2.8 among individuals that had mental disorders.
Relative risk analysis revealed that men were two times higher at risk of homicidal death than women in the study after adjusting for confounding variables. The P value in the study was set at P<0.001; any value less than 0.001 indicated a significant association between the dependent and the independent variables.
Moreover, there was an increased risk of homicide among the Nordic and non-Nordic immigrants compared to the Sweden-born nationalities. After analysis of the homicidal deaths, the study revealed that 22.9 percent of the deaths due to homicide occurred among individuals that had mental disorders. On the other hand, there were 9.4 percent deaths due to homicide in the general population.
Since age was a confounding factor, it was adjusted in the Cox Proportional Hazard Regression model (Crump et al. 2013, p. 4). After adjusting for age, the different mental disorders were associated with a sevenfold risk of homicidal deaths. The confidence intervals for the relative risk lie between 5.98 and 8.83 among the individuals that died due to homicide. According to Kamangar (2012, p. 510), confidence intervals indicate where the values would lie in the analysis was applied to the general population.
Since the range does not contain the value 1, there is a significant association between the different mental disorders and homicide. According to the World Health Organization (2014), substance use is one of the major factors associated with homicidal deaths among people with mental health disorders. In the current study, there was a 16 fold relative risk of homicidal deaths among subjects that abused drugs.
However, the authors state that there were only two homicidal deaths that occurred due to bipolar disorders and this could have affected the associations. Apart from the bipolar disorders, the risk of homicide in all the other mental disorders was reported to be fivefold. After the model adjusted for various demographic variables, the risk of homicidal death in the different individuals rose; sevenfold among people with mental disorders and nine fold among individuals with substance-use related disorders.
The authors performed a stratified analysis on some selected variables (Crump et al. 2013, p. 4). Stratification is one of the methods used to deal with confounding factors at the analysis phase (Kamangar 2012, p. 511). Variables are stratified to specific categories that are likely to confound the association between the dependent and independent variables. In the study, the analysis was stratified into gender (male and female) and this revealed that men with mental disorders were more prone to homicidal deaths (P=0.03) compared to women.
Moreover, individuals that had high income were significantly associated with homicidal deaths (P=0.02). Generally, the result section is well analyzed and the article is clear on the different analyses that were performed by the authors. However, there is no analysis of the relative risk of the association between the mental disorders and the mentally ill patients being victims of suicide.
In summary, the research findings in the current study showed that mental disorders were significantly associated with homicidal death (Crump et al. 2013, p. 5). This was after adjusting for different variables such as age, sex and other socio-demographic factors. Relative risk analysis revealed that individuals with substance-use mental disorders were at a higher risk of homicide among the mentally ill group.
The study also showed that other mental disorders such as anxiety, depression, and schizophrenia resulted in an increased risk of homicidal death. These are the same mental illness identified by the World Health organization (2014) with regard to the risk of homicidal death. The authors in the current article state that prevention strategies in suicide events should target the risk factors that are identified in the analysis.
A different study in the United States revealed that 80 percent of the suicide attempts were associated with individuals with mental disorders (Nock et al. 2010, p. 871). Although the study in the United States focused on attempted suicide, the findings were similar to the current study. Similarly, the study found out that the Odds ratio of substance use in relation to attempted homicide was 2.7 and it was also statistically significant at the multivariate level.
The authors of the current study recognize that it was the first study to establish the association between the mental disorders and homicide using a nationwide population (Crump et al. 2013, p. 5). One of the strengths of the current study is the use of both inpatient and outpatient data as this ascertained the diagnosis of the various mental disorders. The results obtained in the study are thus accurate and reliable. Therefore, the results can be generalized to the entire population.
Moreover, adjustment of the various confounding variables ensures that the associations are more inclined to the truth. According to the authors, majority of the studies that had been previously conducted on the association between mental illnesses and homicide had small sample sizes and hence reduced generalizability (Crump et al. 2013, p. 5). Furthermore the previous studies did not assess the socioeconomic confounders identified in the current study.
Kamangar (2012, p. 515) indicates that a large sample size enhances the representativeness of the study group. Since the study conducted in Sweden incorporated a nationwide cohort, the sample size is representative of the Swedish population. However, the rate of homicide is very low in Sweden and hence it is hard to generalize the findings to nations characterized by high homicide levels.
In reference to the findings that substance use disorders are associated with homicide, there is the likelihood that there were confounding factors in the association. It is likely that these individuals lived in high deprivation areas as recognized by the authors (Crump et al. 2013, p. 5). However, the authors did not assess these neighborhoods in the analysis and this could have affected the findings of the study.
Moreover, due to the substance use, these individuals could have been less aware of the risk associated with the symptoms that they had. According to Nock et al. (2010, p. 872), there could be social factors associated with the neighborhoods that the mentally ill patients live and this could affect the association of the variables under study. Additionally, there are family characteristics that could also explain the patient’s behaviors such as lack of stability and abuse.
Although the current study is strong with regard to the high number of subjects that were recruited, there are several limitations that could have affected the findings. The authors have identified some of the limitations in the study (Crump et al. 2013, p. 6). There is a possibility that some individuals in the general population had mental disorders that had not been diagnosed by the time the study was carried out and this could have underestimated the true prevalence of mental disorders in the population.
The authors also recognize that there is a low rate of homicide in Sweden and this could have affected the precision of some risk factors in the study. The World Health Organization (2014) indicates that patients with mental illness have a higher predisposition to accidents and suicides but this was not assessed in the current study. Furthermore, the researchers indicate that they only obtained information on severe victimization.
Therefore, there is the likelihood that some mild forms of victimization were not reported. In the analysis phase, the researchers seem to focus on confounding factors with little assessment of effect modification. The study did not assess family related factors that could also have been linked to the behaviors exhibited by the mentally ill patients.
Crump, C., Sundquist, K., Winkleby, M. A. & Sundquist , J. 2013, ‘Mental disorders and vulnerability to homicidal death: Swedish nationwide cohort study’, BMJ, vol. 346, no.557, pp. 1-8.
Euser, A. M., Zoccali , C., Jager, K. J. & Dekker, F. W. 2009, ‘Cohort Studies: Prospective versus Retrospective’, Nephron Clinical Practice, vol.113,no. 23, p. 214–217.
Kamangar, F. 2012, ‘Confounding Variables in Epidemiologic Studies: Basics and Beyond’, Archives of Iranian Medicine, vol.15, no. 8, pp. 508-516.
Nock, M. K., Hwang, I., Sampson, N. A. & Kessler, R. C. 2010, ‘Mental Disorders, Comorbidity and Suicidal Behavior: Results from the National Comorbidity Survey Replication’, Molecular Psychiatry, vol. 15, no.11, p. 868–876.
World Health Organization 2014, Mental Health, Geneva. Web.