Combining individual and ecological data to determine compositional and contextual socio-economic risk factors for suicide
Introduction
Research dating back over 100 years has highlighted that individual acts of suicide could be due to forces external to individuals, reflecting the characteristics of societies (Durkheim, 1952; Morselli, 1881). Such effects have in various disciplines become known as “contextual effects” (Greenland, 2001; Joshi, 1993) and in recent years a series of ecological studies have identified factors associated with suicide rates in different geographic areas. The main factors identified are indicators of social fragmentation (anomie) and disintegration, (Ashford & Lawrence, 1976; Congdon (1996a), Congdon (2000); Saunderson & Langford, 1996; Whitley, Gunnell, Dorling, & Smith, 1999) levels of religious belief, (Neeleman, Halpern, Leon, & Lewis, 1997; Neeleman & Lewis, 1999) indicators of socio-economic position (e.g., occupation) of residents, (Congdon (1996b), Congdon (2004); Middleton et al., 2004) levels of unemployment (Congdon, 1996b; Middleton et al., 2004) and area deprivation. (Congdon (1996a), Congdon (2000); Evans, Middleton, & Gunnell, 2004)
There are two principal explanations for the associations reported in ecological studies. First, they may result from the high concentrations of at-risk individuals living in particular areas (compositional effects). For example, associations with social fragmentation may be because people at greater risk of suicide drift into these areas because of the availability of low cost housing or the anonymity these locations offer. Second, place of residence may have a real effect on suicide risk (contextual effect). Pathways for the generation of such risk may lie either in area of residence having a direct effect on the likelihood of suicidal behaviour, or in these areas being less conducive to the provision of social support for at-risk individual, both of which in turn increase suicide risk.
Area effects such as those described above may not only influence overall suicide rates in particular geographic localities, but they also may have differential effects on particular residents which depend on their personal characteristics—such as their marital status or social class. Such effects are sometimes called “cross-level effects”. (Oakes, 2004) Several sociological theories can be suggested as a firm foundation for this phenomenon, e.g. Durkheim's theory of social integration, (Durkheim, 1952) or the “Status Integration Theory” (Gibbs, 2001) formulated by Gibbs and Martin (Gibbs & Martin, 1964) which proposes that the greater the dissonance between an individual and his/her social environment, the greater their risk of suicide. A number of studies have tested the Status Integration Theory using ecological information, (Gibbs (1969), Gibbs (1982), Gibbs (2001); Gibbs & Martin, 1966; Lester (1992), Lester (1995)) but only two studies have formally examined this issue using both ecological and individual-level information in relation to suicide. (Martikainen, Maki, & Blomgren, 2004; Neeleman & Wessely, 1999) Neeleman and Wessely's London-based study of suicide and ethnicity found that as the proportion of an area's population who came from ethnic minority groups increased, so the suicide risk of people from ethnic minorities living in that area decreased whereas that for white residents increased.(Neeleman & Wessely, 1999) These findings may be interpreted as indicating that suicide risks are higher amongst people whose social and demographic characteristics differ from those of the majority of the population living in their neighbourhood—a cross-level effect. However, Martikainen's study, based on the whole of Finland, found no evidence of cross-level effects for measures of social cohesion and socio-economic position. (Martikainen et al., 2004) Furthermore, ecological associations were attenuated when controlling for individual-level variables.
Using the detailed linkage opportunities afforded by Denmark's rich series of health, social and economic registers, we investigated whether an individual's risk of suicide in relation to their marital status, employment status or income differs depending on levels of single person households, employment and income in their municipality of residence.
Section snippets
Source of data and population-based registers
Data were obtained by linking Danish population-based registers using the unique personal identification number, assigned to all persons living in Denmark and used across all registration systems. (Malig, 1996) The Danish Medical Register on Vital Statistics contains dates and causes of all deaths in Denmark recorded from the Cause-of-Death Certificates since 1976 and for suicide since 1970. (Sundhedsstyrelsen (The Danish National Board of Health), 1992) Suicide was defined as ICD8 (World
Results
In total, 9011 cases (2867 females) and 180,220 controls were identified. The median municipality percentages of females and males living alone were 22% (range, 4–48%) and 25% (range, 11–52%), respectively. The median percentage of employed individuals in the municipalities was 77.8% (range, 43.8%–91.2%) and the median gross income of the municipality residents was US $39605 (range US $32308 to US $68048). In Fig. 1 the geographical distribution of the quartiles of the municipality measures is
Discussion
This study analysed the effect of area (contextual) characteristics on suicide taking account of the characteristics of the individuals who committed suicide in each area (compositional effects). In keeping with the previous literature, (Ashford & Lawrence, 1976; Congdon (1996a), Congdon (1996b), Congdon (2004); Dorling & Gunnell, 2003; Lester, 1993; Middleton et al., 2004; Neeleman et al., 1997; Neeleman & Lewis, 1999; Whitley et al., 1999) we found associations of suicide with all three
Conclusions
Whilst there has been considerable and longstanding speculation (Durkheim, 1952; Morselli, 1881) concerning the effect of place of residence (context) on suicide risk—our analysis suggests that the ecological associations are due primarily to the proportion of high-risk individuals living in particular areas and only limited contextual effect of the characteristics of the areas themselves upon suicide risk was found. Our findings suggest health policy might best focus on people within places
Acknowledgements
The study was financially supported by the Stanley Medical Research Institute. Psychiatric epidemiological research at the National Centre for Register-based Research is in part funded through a collaborative agreement with the Centre for Basic Psychiatric Research, Psychiatric Hospital in Aarhus. The authors would like to thank Sarah Curtis for providing helpful comments and suggestions on this paper.
References (53)
Bayesian models for spatial incidence: A case study of suicide using the BUGS program
Health & Place
(1997)Estimating neighborhood health effects: The challenges of causal inference in a complex world
Social Science & Medicine
(2004)Combining ecological and individual variables to reduce confounding by indication: Case study—subarachnoid hemorrhage treatment
Journal of Clinical Epidemiology
(2000)The (mis)estimation of neighborhood effects: Causal inference for a practicable social epidemiology
Social Science & Medicine
(2004)- et al.
A study of the geographical distribution of suicide rates in England and Wales 1989–92 using empirical Bayes estimates
Social Science & Medicine
(1996) - et al.
Measuring the built environment: Validity of a site survey instrument for use in urban settings
Health & Place
(2001) - et al.
Bias due to aggregation of individual covariates in the Cox regression model
American Journal of Epidemiology
(2004) Risk of suicide and spouse's psychiatric illness or suicide: Nested case–control study
British Medical Journal
(2003)The effect of psychiatric illness and labour market status on suicide: A healthy worker effect?
Journal of Epidemiology and Community Health
(2005)The effect of psychiatric illness and labour market status on suicide: A healthy worker effect?
Journal of Epidemiology and Community Health
(2005)
Midlife suicide risk, partner's psychiatric illness, spouse and child bereavement by suicide or other modes of death: A gender specific study
Journal of Epidemiology and Community Health
Aspects of the epidemiology of suicide in England and Wales
International Journal of Epidemiology
Nonparametric estimation of relative mortality from nested case–control studies
Biometrics
Incidence of schizophrenia in ethnic minorities in London: Ecological study into interactions with environment
British Medical Journal
Statistical models in epidemiology
Suicide and parasuicide in London: A small-area study
Urban Studies
The epidemiology of suicide in London
Journal of the Royal Statistical Society. Series A: Statistics in Society
Monitoring suicide mortality: A Bayesian Approach
European Journal of Population
A multivariate model for spatio-temporal health outcomes with an application to suicide mortality
Geographical Analysis
IDA—en integreret database for arbejdsmarkedsforskning (The Integrated Database for Longitudinal Labour Market Research)
A glossary for multilevel analysis
Journal of Epidemiology and Community Health
Suicide: The spatial and social components of despair in Britain 1980–2000
Transactions of the Institute of British Geographers
Suicide (originally published 1897)
Social fragmentation, severe mental illness and suicide
Social Psychiatry and Psychiatric Epidemiology
Marital status and suicide in the United States: A special test of the status integration theory
American Journal of Sociology
Testing the theory of status integration and suicide rates
American Sociological Review
Cited by (87)
Neighborhood social context and suicide mortality: A multilevel register-based 5-year follow-up study of 2.7 million individuals
2022, Social Science and MedicineMidlife suicide: A systematic review and meta-analysis of socioeconomic, psychiatric and physical health risk factors
2022, Journal of Psychiatric ResearchCitation Excerpt :After removing duplicates and screening of titles and abstracts, 874 full-text articles were retained for further assessment. Consequently, 62 articles fulfilled the criteria of inclusion for analysis (Agerbo et al., 2007; Bjorkenstam et al., 2005; Blakely et al., 2003; Borrell et al., 2017; Charlton, 1995; Chen et al., 2009; Christensen et al., 2007; Corcoran, 2009; Corcoran and Arensman, 2010; De Leo et al., 2013; Doyle et al., 2016; Duberstein et al., 2004; Erlangsen et al., 2008, 2012; Fall et al., 2009; Fang et al., 2012; Fazel et al., 2013, 2014; Fukuchi et al., 2013; Heikkinen et al., 1995; Hempstead and Phillips, 2015; Hiyoshi et al., 2018; Im et al., 2011; Inoue et al., 2005; Isometsa et al., 1995; Kimura et al., 2016; Kolves et al., 2006; Koo et al., 2017; Kwan et al., 2005; Lageborn et al., 2017; Lee et al., 2009; Lofman et al., 2011; Lorant et al., 2005; Lundin et al., 2010; Masocco et al., 2008; Massetti et al., 2018; Moustgaard et al., 2013; Na et al., 2019; O'Neill et al., 2018; Osborn et al., 2008; Petrosky et al., 2018; Pirkola et al., 2000; Pompili et al., 2013; Qin, 2011; Rodriguez-Pulido et al., 1992; Rossow and Amundsen, 1995; Sørensen and Fenger, 1992; Stone et al., 2016, 2018; Strand et al., 2010; Strid et al., 2014; Sun et al., 2018; Tamosiunas et al., 2005; Tian et al., 2016; Wong et al., 2008; Wyder et al., 2009; Yamauchi et al., 2014, 2016; Yeh et al., 2008; Yip et al., 2012; Yip and Thorburn, 2004; Zaheer et al., 2018), of which 48 reported the distribution of exposures and 41 the suicide risk associated with exposure. Nine studies reported the estimates for factors in two domains (De Leo et al., 2013; Hempstead and Phillips, 2015; Im et al., 2011; Inoue et al., 2005; Isometsa et al., 1995; Koo et al., 2017; Na et al., 2019; O'Neill et al., 2018; Stone et al., 2018), while one study reported the estimates related to factors in all three domains (Wong et al., 2008).
Longitudinal effects of physical and social neighbourhood change on suicide mortality: A full population cohort study among movers and non-movers in the Netherlands
2022, Social Science and MedicineCitation Excerpt :Associations have been found between higher suicide risk and, for example, rural residence (Hirsch and Cukrowicz, 2014), socioeconomically deprived neighbourhoods (Cairns et al., 2017; Martikainen et al., 2004; Rehkopf and Buka, 2005), high social fragmentation (Agerbo et al., 2007; Congdon, 2013; Hagedoorn et al., 2020; Martikainen et al., 2004), and low social capital (Kunst et al., 2013). However, associations with neighbourhood socioeconomic characteristics (Agerbo et al., 2007), social fragmentation (Collings et al., 2009) and green space (Helbich et al., 2020) were not universally found. A possible reason for the spurious associations between suicide risk and neighbourhood characteristics is the widespread use of cross-sectional study designs (Agerbo et al., 2007; Hagedoorn et al., 2020; Helbich et al., 2020; Kunst et al., 2013; Martikainen et al., 2004) failing to represent neighbourhood change over time and neglecting that people move home throughout their lifetime (Helbich, 2018).
Why adults with psychiatric conditions request euthanasia: A qualitative interview study of life experiences, motives and preventive factors
2021, Journal of Psychiatric ResearchExamining the mechanisms by which adverse life events affect having a history of self-harm, and the protective effect of social support
2020, Journal of Affective Disorders