Elsevier

Health & Place

Volume 14, Issue 3, September 2008, Pages 492-506
Health & Place

An atlas of suicide mortality: England and Wales, 1988–1994

https://doi.org/10.1016/j.healthplace.2007.09.007Get rights and content

Abstract

The incidence of suicide exhibits marked geographic variability; however, documentation of features in its spatial distribution, or the magnitude of differences, is limited. Standardised mortality ratios, commonly presented in maps, are calculated in each area independently and incorporate no information about heterogeneity or clustering. Bayesian hierarchical models with random effects for between-area and local variability in neighbouring areas were used to map age- and sex-specific estimates of rate ratios of suicide across wards in England and Wales. Differences were greater than expected due to random variation alone. Although the geography of suicide differed across age/sex groups, some common patterns emerged e.g. high rates in (a) central parts of cities and (b) remote and coastal areas. Some features were common to all, while others appeared male specific or specific to the younger age group. Suicide prevention strategies can be informed by an understanding and addressing the geography of suicide.

Introduction

In the recent years there have been marked changes in the age patterning of suicide mortality in Britain (Gunnell et al., 2003) as well as in much of the industrialised world (Cantor, 2000). As the most unfavourable trends have generally occurred in young people, suicide ranks as one of the principal causes of premature mortality (Gunnell, 2000). In their classic work, Morselli (1881) and Durkheim (1952) presented maps of national and sub-national suicide rates across Europe. For over a century now, a striking feature of the epidemiology of suicide has been the wide geographical variation in its occurrence i.e. more than 10-fold differences in rates across Europe (Cantor, 2000). Differences of similar magnitude have been reported within Britain (Bunting and Kelly, 1998) and even across smaller areas within particular cities (Sainsbury, 1955; Gunnell et al., 1995; Congdon, 1996). Since Morselli's and Durkheim's maps were used as an aetiological tool in the 19th century, there have been several quantitative investigations into area-based associations of suicide with indicators of their socio-economic characteristics. However, few studies have used maps to explore and represent the geographical variability and patterning of suicide, either in Britain or elsewhere.

The UK government has recently recognised suicide (and mental health) as an important contributor to area health inequalities (Department of Health, 2001). However, the geography of suicide is still poorly understood. Part of the problem is that standardised mortality ratios (SMRs), commonly presented in maps, are calculated in each area independently and, as such, incorporate no information about the amount of heterogeneity or clustering across the map (Clayton and Kaldor, 1987). Inference from SMR maps is particularly problematic at finer levels of geographical aggregation not only because a large proportion of areas might record no deaths but also because areas with small populations are more likely to produce extreme and unreliable estimates. Yet, exploring these issues at a smaller area level, and across different age/sex groups, is central to the investigation of possible area influences, and their magnitude, upon their residents.

Whether descriptive, or ecological assessments, previous studies have generally: (i) focused on relatively large areas or smaller areas within a defined locality e.g. single city or district and/or (ii) not investigated whether the geography of suicide varies across different age and sex groups—as recent changes in its age patterning might suggest and/or (iii) used simple regression approaches that either treat geographical areas as independent or investigated evidence of heterogeneity but not clustering across the map. In a previous analysis, we documented the geographical distribution of suicide amongst 15–44-year old males living in England and Wales (Middleton et al., 2006). We found high rates of suicide clustered in inner-city and coastal areas. Here, we extend our investigation to examine the geographical distribution of suicide in older males and females of all ages. Using geographically and statistically appropriate methodology, we investigated the magnitude and spatial patterning of suicide mortality across 9265 wards in England and Wales and addressed the extent to which there exists a common geography across different age/sex groups.

Section snippets

Data sources

All suicide and undetermined deaths in the 7-year period 1988–1994 were obtained from the Office for National Statistics, ONS (ICD codes E950.0–E959.9 and E980.0–E989.9 excluding E988.8). The inclusion of undetermined deaths has been common practice in suicide research as there is evidence to suggest that most deaths given an open verdict using legal criteria are suicides (Linsley et al., 2001). Deaths coded E988.8 were excluded from the analysis because this code is used to accelerate death

Results

There were 27,839 adult (aged 15+) male and 10,241 adult female suicides in 1988–1994. Of these, 215 (0.6%) of all deaths with missing or incorrect postcodes were excluded from the analyses since they could not be placed geographically. The number of such deaths was not only small but there also is no reason to believe that the error introduced is not geographically random. Even after excluding the 5% most extreme rates on either side of the distribution, there remained 5-fold differences in

Main findings

We improved upon current evidence and practice to produce an atlas of suicide mortality in England and Wales across different age/sex groups and at a smaller area level than documented previously. With more than 9000 small areas across England and Wales, this is the finest geographical scale ever used to examine the national geography of suicide in Britain or elsewhere. Maps presented here incorporated evidence of heterogeneity and clustering in the estimation procedure and revealed patterning

Conclusion

While there is no prior knowledge as to what geographical scale might be important in assessing common area influences upon a population's suicide experience or mental health, large areas are likely to include smaller localities with large differences in their socio-economic characteristics. Thus, the level of geographical aggregation at which patterns and associations are explored is central to understanding the geography of suicide and possible area influences, and their strength, upon a

Contribution of authors

This study forms part of an original proposal put together by D.G., J.S. and N.M. for N.M.'s Ph.D. funded jointly by the University of Bristol and the Overseas Research Students (ORS) Award Scheme during which D.G. and J.S. were N.M.'s Ph.D. supervisors (2000–2004). D.G. and N.M. collected the data. All analyses were performed by N.M. J.S. provided statistical advice. A series of maps of suicide mortality and residual variation after accounting for socio-economic area characteristics appear in

Competing interest statement

All authors declare that there are no competing interests.

Acknowledgements

We wish to acknowledge the Office for National Statistics, and Claire Griffiths in particular, for providing us with the mortality data. Also, Danny Dorling (University of Sheffield) for all the socio-economic, geographic- and mapping-related data e.g. ward boundaries polygon files and postcode look-up files, and Peter Congdon, Tony Ades, Lesley Wood and Ben Wheeler for helpful discussions.

References (57)

  • J.R. Ashford et al.

    Aspects of the epidemiology of suicide in England and Wales

    International Journal of Epidemiology

    (1976)
  • R.D. Baller et al.

    Social integration, imitation, and the geographic patterning of suicide

    American Sociological Review

    (2002)
  • L. Bernardinelli et al.

    Empirical Bayes versus fully Bayesian analysis of geographical variation in disease risk

    Statistics in Medicine

    (1992)
  • L. Bernardinelli et al.

    Bayesian estimates of disease maps: how important are priors?

    Statistics in Medicine

    (2001)
  • C.A. Brewer

    Guidelines for selecting colors for diverging schemes on maps

    The Cartographic Journal

    (1996)
  • C.A. Brewer et al.

    Mapping mortality: evaluating color schemes for choropleth maps

    Annals of the Association of American Geographers

    (1997)
  • S.P. Brooks et al.

    General methods for monitoring convergence of iterative simulations

    Journal of Computational and Graphical Statistics

    (1998)
  • J. Bunting et al.

    Geographic variations in suicide mortality, 1982–1996

    Population Trends

    (1998)
  • I.H. Burnley

    The ecology of suicide in an Australian: the case of Sydney

    Australian Journal of Social Issues

    (1978)
  • C.H. Cantor

    Suicide in the western world

  • R.S. Cavan

    Suicide

    (1928)
  • D. Clayton et al.

    Empirical Bayes estimates of age-standardised relative risks for use in disease mapping

    Biometrics

    (1987)
  • P. Congdon

    Suicide and parasuicide in London: a small-area study

    Urban Studies

    (1996)
  • P. Congdon

    Monitoring suicide mortality: a Bayesian approach

    European Journal of Population

    (2000)
  • P. Congdon

    A multivariate model for spatio-temporal health outcomes with an application to suicide mortality

    Geographical Analysis

    (2004)
  • The National Health Inequalities Targets

    (2001)
  • National Suicide Prevention Strategy for England

    (2002)
  • E. Durkheim

    Suicide: a Study in Sociology

    (1952)
  • Cited by (49)

    • Natural environments and suicide

      2018, The Lancet Planetary Health
    View all citing articles on Scopus
    View full text