Introduction
Worldwide, male suicide rates are three times higher than those of women, with most high-income countries reporting the highest suicide rates in mid-life (45–55 years).1 2 In the UK, the USA, Australia, Canada and many other high-income countries, middle-aged men have the highest suicide rates,2–6 and are a major priority group for suicide prevention.1 They account for around a quarter of all deaths by suicide in England, Scotland and Wales, with over 1500 deaths each year.3–7 In the late 1990s, the highest suicide rates were in young men aged 20–34.3 Since 2010, there has been a shift in the pattern of age-related suicide rates; men aged 40–54 years are the group at highest risk.3 There are two main reasons for this change. First, as a direct effect of the 2008 global economic recession—when rising unemployment contributed to excess suicide deaths particularly in men, in England8 and across Europe and the USA.9 Second, it may represent a ‘cohort effect’; the impact of the earlier recession in the 1980s, when suicide rates increased substantially in young men, being carried with this same group into mid-life.10
Known factors associated with suicide in men include financial and employment stressors, limited social support including being unmarried, physical and mental ill-health (particularly depression), and drug and alcohol misuse.11–15 Richardson and colleagues’ 2019 systematic review emphasised the complexity of suicide in men and highlighted the need to understand how factors associated with suicide interact across the lifetime.15 Internationally, economic uncertainty has been linked to an increase in suicide rates, particularly in men8 16 17 and disadvantaged groups, such as those with existing mental health problems.18 A relationship between debt and physical and mental health, including suicidal behaviour, has also been established,19 with more severe debt being related to worse health outcomes.20 Studies have also cited that men may be less likely than women to talk about or report mental health problems, seek out beneficial support,21 or engage with services—especially in the context of employment, financial or relationship difficulties13 22 possibly due to stigma, pride or shame.18 21
Reducing and preventing suicide in men, particularly men in mid-life, is a UK-wide suicide prevention priority.23 National investment in suicide prevention through NHS England’s suicide reduction programme24 has seen local areas develop interventions for men, including awareness campaigns, mental health first-aid training, and short-term practical and emotional support aimed at tackling stigma and reducing suicide. Although it is too soon to evaluate the long-term impact of these projects on reducing suicide in men, previous campaigns have potentially increased public awareness and reduced stigma associated with mental health problems in general.25
We established a national study of suicide in middle-aged men to understand the stresses they faced prior to death that may be contributing to their high suicide rates and to record contacts with services as this could play a part in prevention. We also wanted to examine particular subgroups—men experiencing economic adversity, men with physical ill-health or a history of alcohol and/or drug misuse, men who had been bereaved, and men in contact with multiple agencies to investigate whether there were certain characteristics or antecedents of suicide that were more commonly associated with particular groups of men in mid-life. In this paper, we present findings from our examination of a range of investigations by official bodies that can occur when someone dies by suicide. These reports, including from coroners, the police and the NHS, provide detailed accounts from families, friends and professionals of the adversities being faced prior to death. Using these data, we aimed to: (1) examine the antecedents of suicide in middle-aged men aged 40–54, including the frequency of recent life events or clinical factors, and (2) investigate (a) the assumption that few middle-aged men would have been in contact with any support services or agencies in the context of the existing narrative that men do not seek help for their problems and (b) the presence of a combination of multiple stresses in particular subgroups of middle-aged men who died by suicide, using latent class analysis (LCA).
Materials and methods
Study population
We collected data about men aged 40–54 years (referred to as ‘middle-aged men’) who died by suicide (including probable suicide) between 1 January 2017 and 31 December 2017 in England, Scotland and Wales. Since 2010, men aged 45–49 have had the highest age-specific suicide rate, followed by those aged 40–44 and aged 50–54.3 7
Suicide deaths were notified to us by the Office for National Statistics (ONS; for deaths registered in England and Wales) and National Records for Scotland (for deaths registered in Scotland). Data from Northern Ireland were not obtained due to restrictions in the sharing of person identifiable data.26 As is standard for suicide research,27 we included deaths receiving a conclusion of suicide or intentional self-harm or events of undetermined intent. Deaths coded with the International Classification of Diseases, Tenth Revision (ICD-10) codes X60–X84, Y10–Y34 (excluding Y33·9) and Y87 were included. Any deaths receiving a narrative conclusion in England and Wales were included if ONS procedures applied one of the latter ICD-10 codes.
Sample
In 2017, there were 1516 men aged 40–54 who died by suicide in England, Scotland and Wales. We aimed to investigate factors associated with suicide in a random, stratified sample of approximately 20% of these men. This is based on our previous experience of the volume of these types of data it is possible for researchers to extract and analyse within the study period.28 Sampling was based on the proportion of individual deaths from (1) each age group (40–44, 45–49 and 50–54) and (2) each country. We used the SPSS random allocator function to select a random sample of 288 (19%) middle-aged men for investigation.
Data sources
We collected data about these 288 men from a range of investigations into their deaths by official bodies. We have previously used this methodology in an investigation of almost 600 suicide deaths in young people.28 In total, data were received for 242 (84%) of these men.
For deaths in England and Wales, we obtained audio copies of coroner inquest hearings (or copy statements or depositions, where unavailable). For deaths in Scotland, redacted Crown Office and Procurator Fiscal Service police death reports were requested. These present a concise summary of the circumstances leading up to death but are less detailed than coroner inquest hearings. We obtained information from coroner inquest hearings or police death reports for 228 (79%) deaths (194 and 34 deaths, respectively). For 12 deaths, the coroner (or equivalent) was unable (n=5) or did not wish (n=7) to provide data, and for 34 deaths, data were not returned within the study period. The content of the coroner inquest hearings varied depending on the complexity of the individual case but typically provided detailed personal testimony (ie, witness statements) from family and friends, police, postmortem and toxicology reports and, if relevant, medical evidence pertinent to the cause and circumstances of a death from primary and secondary care services.
The National Confidential Inquiry into Suicide and Safety in Mental Health (NCISH) collects data on a UK-wide case series of people who die by suicide within 12 months of being seen by mental health services. A full description of the NCISH methodology has been previously published.29 Information from the NCISH database was obtained for 86 (30%) men identified as mental health patients.
We obtained 68 internal serious incident investigation reports from NHS services on patients who had contact within 12 months of death; 12 deaths had not met the criteria for review. For six deaths, data were not returned within the study period.
In England and Wales, the Prison and Probations Ombudsman (PPO) independently investigates the deaths of people in detention, residents in approved premises, and immigration detainees; these reports are published on the PPO website. In Scotland, deaths in prison are investigated at Fatal Accident Inquiries (FAI). The judgements of FAIs are published by the Scottish Courts and Tribunals Service. We identified three published criminal justice reports which matched men sampled in the study.
Procedure
A pro-forma, developed with advice from people with lived experience and based on our previous study,28 was used to extract information from the described data sources for aggregate analysis. Information was collected about demographic, medical, psychiatric, long-term and short-term (within 3 months of death) events that may have contributed to suicide risk. The last known contacts with general practitioner (GP), accident and emergency department (A&E) and secondary care services were also recorded. We recorded factors referred to in any of the data sources as present at any time or specifically in the 3 months prior to death (described as ‘recent’); presence suggests that a factor was relevant to the death but not necessarily causal. Variable definitions are shown in the online supplemental table 1.
A sample of 30 (12%) cases was reviewed by three researchers (CR, JG, SGT) and a Kappa Fliess interrater reliability test was performed, with initial agreement for individual items of 58–100%. Where there was disagreement, information was discussed and re-evaluated increasing concordance to 100%. Any areas of uncertainty in the remaining 212 cases were discussed with the team to ensure a consistent reviewing approach.
Statistical analysis
The denominator in all estimates was the total number of men where at least one data source was obtained (ie, 242 individuals), unless otherwise stated. Items of information not recorded in any data source were assumed to be absent or not relevant to the individual’s death. Pearson’s χ2 tests or Fisher’s exact tests (for cell counts less than five) were used to determine the association between particular subgroups of middle-aged men (including (1) men experiencing economic adversity (unemployment and/or recent financial and/or accommodation problems), (2) men with a physical health condition, (3) men with a history of alcohol and/or drug misuse, (4) men who had been bereaved, and (5) men in contact with multiple agencies (contact with three or more of primary care, justice system, mental health, or other services)), and key sociodemographic and clinical characteristics (such as marital and employment status, living circumstances, internet use, mental health diagnosis, self-harm, and service contact; 33 characteristics in total). The subgroups were based on prior knowledge and previous literature11–15 19 20 and compared with all other men not in that particular subgroup. We used the Bonferroni correction method when reporting statistical significance.30 This was at the 0.15% level (p≤0.0015) to account for the 33 multiple risk factors that were examined for each subgroup (0.05/33). Exploratory LCA was used to identify distinct classes (subgroups) of middle-aged men who died by suicide. We did this to determine potentially informative risk profiles based on the confluence of multiple risk factors which might inform specific prevention initiatives. Using logit LCA, we compared models that allowed for one to five latent classes and determined the best model fit using the likelihood ratio (G2) test and comparing Akaike’s information criterion (AIC) and Bayesian information criterion (BIC) values.31 The criteria for the best fitted model were when we failed to reject the null hypothesis for G2, in combination with lower values of AIC and BIC. Results from the LCA with p values less than 5% (p≤0.05) were considered statistically significant.
When reporting results we have suppressed cell counts under three, including zero, in accordance with ONS guidance on disclosure control to protect confidentiality. All analyses were undertaken using Stata software V.16·1.
Patient and public involvement
Three members of Mutual Support for Mental Health Research, a group of people with experience of self-harm, suicidality, or mental illness as either patients or carers, were involved in the research process by providing advice during the design of the study. They shared their insights and thoughts on the design of the proforma and data items. The group were not involved in the conduct or reporting of the study but were asked for feedback on key messages derived from the study’s findings.
Role of the funding source
The funders played no part in the design, data collection and analysis, or interpretation and writing up of the study.
Data availability statement
Data from this study are not available due to the sensitive nature of the research and information governance restrictions in place to protect confidentiality. A request to access data can be made to the Healthcare Quality Improvement Partnership: https://www.hqip.org.uk/national-programmes/accessing-ncapop-data/%23.ZEAOsOzMKAM