Radical
Statistics
Researching
suicidal behaviour
Sarah Payne and Rachel Lart1
Introduction – gender and suicidal behaviour
Each year in the UK over 5,000 people commit suicide, whilst between 50,000 and 100,000 people will attempt suicide (OPCS Mortality Data, 1996; Gunnell et al, 1995). Suicide risks however are not evenly spread between men and women: completed suicide is more common amongst men, whilst women are more likely to commit acts of deliberate self-harm. For example, in 1992 male suicide produced four and a half times as many working years lost as female suicides (67,300 compared with 14,700 years lost) (OPCS Mortality Data, 1992).
However, recent statistics in the UK suggest a change in these trends in suicidal behaviour for men and women: whilst completed suicide has increased substantially for men in the past few years, particularly amongst younger age groups, male rates of parasuicide or deliberate self-harm have also increased (Hawton et al, 1996). During this time suicide mortality rates for women have decreased, whilst female rates for parasuicide have largely remained stable. Thus the gender gap in completed suicide has widened, while the gender gap in attempted suicide has narrowed.
This paper discusses methodological issues relating to statistics on suicidal behaviour, and the question of suicide and gender. The basis of this discussion is a recent literature review on suicidal behaviour and gender, comprising a search of four major computerised databases identified as the most relevant for this subject matter.2 Our interest was the extent to which gendered differences in suicidal behaviour had been explored in different disciplinary areas, in particular medical, anthropological and sociological disciplines, and what were seen as the major explanations for both the difference between men and women and the direction of recent trends.
Methodological issues in suicide research
Research into both suicide and parasuicide is problematic for a number of reasons. Firstly, difficulties arise with both suicide and parasuicide in terms of the meaning ascribed to either of these acts and the question of intent. Thus figures for suicide mortality may be unreliable, in that suicide deaths may be recorded as suicide, or as ‘undetermined deaths’, or in some circumstances as accidental death.
A general issue in suicide research, then, is the extent to which official figures reflect real patterns of actions, and the extent to which the figures are socially constructed. This is reflected in popular culture in the gendered stereotype that whilst men mean to die, women’s suicidal acts are more likely to be seen as a plea for help (Canetto, 1992-3).
Similar problems arise with the source of statistics on parasuicide or deliberate self-harm. Not all acts of deliberate self-harm are recorded, and statistics tend to count only those parasuicides which come to notice. This means that research on parasuicide is inevitably biased towards some acts. The vast majority of people recorded as having committed an act of deliberate self-harm are those who have attempted suicide though self-poisoning, who are most likely to be admitted to hospital (Platt, 1992). Whilst people who harm themselves using other methods will also be treated in hospital where appropriate, this is less common. Others will remain outside the hospital and their acts are unrecorded. Statistics on parasuicide, then, will always be incomplete, and particular methods of self-harm are unlikely to be counted.
In addition to the problems of defining suicidal behaviour, and related issues of measurement, further difficulties arise in research which aims to assess the underlying motivations or causes of suicidal behaviour. Two major approaches to research in this area have been used. One method explores relationships between different sets of data - for example, national or regional statistics on suicide may be compared with figures for the same area for marital breakdown, unemployment, indicators of poverty, or female labour force participation (Congdon, 1996; Diekstra, 1989; Stack, 1996; Gunnell et al, 1995). Such research may suggest relationships - for example, Diekstra (1989) found that worldwide, high unemployment rates were predictive of an increase in suicide mortality - but cannot prove that those individuals who commit suicide or who engage in deliberate self-harm are also those who are divorced, unemployed, poor or who have difficulties as a result of increased female employment. This problem - the ‘ecological fallacy’ - is less open to more complex statistical approaches such as multi-level modelling due to the relatively small numbers involved - particularly with completed suicide.
The second major approach - more commonly used in research within medical disciplines - focuses on the individual, and either, in the case of completed suicide, attempts to reconstruct a profile of the person before the suicidal act, or, with deliberate self harm, gathers information on that individual (e.g. Charlton, 1995). Sometimes called a ‘psychological autopsy’, such research utilises a range of records, including medical notes, the Coroner’s report, and interviews with ‘survivors’ - those who had a close relationship with the person who has completed suicide. Collecting data at the level of the individual allows a closer analysis of the factors that may have led to the act itself, but suffers from other problems. Data relating to factors such as employment status and occupational class, marital status and age for example are relatively straightforward. However, more complex issues - including the impact of unemployment or divorce, as well as questions of mental health - are largely reported retrospectively, and are inevitably open to influence by the suicidal act itself. Studies of the mental health of completed suicides, for example, largely find, unsurprisingly, that the person was depressed before they committed suicide (e.g. Foster et al, 1997). Interviews, both with those who knew the person committing suicide and those with health and social care professionals who may have had a relationship with the individual, are likely to be affected by the knowledge of the suicide. However, the extent and direction of this effect cannot be determined. Similarly, reliance on evidence collected from coroners’ reports is affected by retrospective data collection, and evidence collected from other records completed prior to the suicidal act - such as medical notes - is rarely complete.
The difficulty is further complicated by the fact that the direction of any effect of retrospective analysis of the individual’s state of mind and reasons for their action cannot be assumed. People reporting the death by suicide of a family member, for example, are likely to have a number of conflicting, and very difficult, emotions, which may increase the likelihood of attributing the act to preceding depression or a factor such as loss of paid employment or the breakdown of a relationship. Alternatively, however, they may wish to put less emphasis on factors such as relationships and more on other factors which can be seen as immutable and somehow less blameworthy for the survivor. Thus long term depression may seem a more acceptable explanation than recent relationship breakdown.
Similarly, research at the level of the individual which explores deliberate self-harm or parasuicide may suffer from the problem of retrospective reporting, although in this case the person committing the act of self-harm is alive and able to speak for themselves. Again, this is less of a problem in studies which focus on more objective or practical aspects of the individual’s circumstances prior to the suicidal act - such as employment status, marital status, age and address. The research of the Oxford Monitoring System, for example, has been able to collect data over more than twenty years, particularly on factors like employment status, age and marital status of people attempting suicide (e.g. Hawton et al, 1996) and has contributed significantly to our understanding of who commits such acts of deliberate self-harm.
However, more complex issues relating to motivation are likely to be difficult to measure. This may be the result, again, of a diversity of factors, including the individual’s opportunity to reflect on and revise their feelings in relation to the attempt, and also the individual’s desire to present themselves in a particular way - which may result in a downplaying of the significance of particular factors or the act itself.
Methodological issues and gender
We turn now to look at the implications of these methodological issues in suicide research in terms of their impact on one particular question - the role of gender in suicidal behaviour.
Firstly, the matter of intent and measurement: do figures of reported suicide and parasuicide accurately reflect gendered patterns of behaviour? With suicide, we must ask whether cultural stereotypes influence the assignment of particular deaths to the category of suicide, and if so are these assumptions gendered? If suicide is more or less acceptable for men and women in any given society, there might be gendered differences in the willingness to record any death as suicide. For example, a society which sees suicide as an honourable course of action for men in particular circumstances, but not for women, will be more willing to register male deaths as suicide. Such cultural values may also impact on actual patterns of deaths, as men would be more likely to choose suicide in such a culture - so that the impact of this cultural difference on measurement is difficult to determine. Canetto (1992-3) suggests that in developed countries there are such gendered notions of suicidal behaviour. She challenges the notion of classifying suicidal behaviour in terms of intent i.e. as either 'attempted' or 'successful' suicide. These traditional classifications are inherently gendered; women are seen as 'attempting' suicide more often, but failing more often, than men. This, she argues, is linked to the underlying explanations for suicidal behaviour attributed to each sex; that women attempt suicide because of problems in their personal relationships, while men do so because of threats to their public persona. Women's suicides are seen as irrational and emotional responses, while men's are seen as rational choices. Canetto argues that although there is evidence to refute these stereotypes, much research on suicide starts from these assumptions and therefore reinforces them.
Even if suicide rates reflect real patterns of actions, a further problem is whether some methods of suicide are more or less likely to be classified as suicides. Men and women tend to choose different patterns of methods, with higher male suicide mortality rates for methods involving hanging and motor vehicle exhaust fumes, and higher female rates for poisoning. This can also affect suicide statistics, as some methods may more readily be classified as accidental. Self-poisoning involving prescription medicines, for example, can be more readily classified as accidental than suicide involving car exhaust fumes.
These problems of classification raise a particular question in the use of suicide statistics: are gendered patterns real? A study by Platt, Backett and Kreitman (1988) found no support for the social construction thesis in their comparison of suicides and undetermined deaths among deaths in Edinburgh 1968-1983. Their main findings were that '[s]uicide and undetermined deaths could not be differentiated by gender, marital status, previous psychiatric contact, age or social class' (p220). The only significant discriminator was method, with deaths from 'active' methods (i.e. hanging, jumping, firearms and cutting) much more likely to be classified as suicide than deaths from 'passive' methods (poisoning, gassing etc.). However, once adjusted for gender differences in method, there was no difference in verdict.
Turning to deliberate self-harm, there is a gendered effect in the way statistics are collected and the reliance on data on medical treatment, as women and men also use different methods for uncompleted suicide (Platt et al, 1988) (although this may be changing - Hawton et al, 1996). The effect of this bias is that it is likely that figures for parasuicide are better able to capture women than men committing this act. This may explain to some extent the over-representation of women in this act, whilst the increasing use by men of self-poisoning as a method in recent years may also explain the apparent growth in male rates of parasuicide.
The problem of retrospective analysis may also be influenced by gender. With completed suicide, evidence is collected from a range of sources including coroners’ reports, medical notes and personal accounts of ‘survivors’. How likely is it that gender would affect such retrospective thoughts? No research has specifically focused on this issue, but Canetto’s research (1992-3) suggests that stereotypical pictures of male and female suicides - ‘She died for love and he for glory’ (p.1) - affect interpretations of the reasons for suicidal behaviour both by those near to the deceased and by society more widely. Women are more frequently treated for mental ill-health (Ussher, 1991) and women are more likely than men to appear in suicide statistics as having a pre-existing psychiatric problem.
Conclusion
This is a very brief review of some of the issues affecting suicide research, and some of the ways in which these methodological considerations may affect suicide statistics and understanding of gender differences in suicidal behaviour. Much of the vast amount of research carried out in this field does not specifically address gender as a dimension. However, whether gender is a focus or not, it is critical that researchers question the gendered nature of implicit assumptions embedded in methods used - without this recognition, differences between men and women’s suicidal behaviour remain imperfectly understood.
1. This paper arises from a project carried out by Rachel Lart, Lesley Doyal and Sarah Payne, School for Policy Studies, and David Gunnell, Department of Social Medicine, University of Bristol. The project was funded by the NHS Research and Development Programme.
2. Medline (1974-1997), Embase (1980-1997), Sociofile (1974-1996), and Psychlit (1974-1996) databases were searched on the words ‘suicid*’ and ‘parasuicid*’ or ‘self harm’, each refined by the words ‘gender’ or ‘male’ and ‘female’.
REFERENCES
Canetto, S. (1992-3), ‘She died for love and he for glory: gender myths of suicidal behaviour’, Omega, Vol. 26, No 1, pp.1-17
Charlton, J. (1995), ‘Trends and patterns in suicide in England & Wales’, International Journal of Epidemiology, Vol. 24, Sppl. 1, pp.45-52
Congdon, P. (1996), ‘Suicide and parasuicide in London: a small area study’, Urban Studies, Vol. 33, (1), pp.137-158
Diekstra, R. (1989), ‘Suicide and attempted suicide: an international perspective’, Acta Pyschiatrica Scandinavica, Vol. 80, Sppl. 87, No. 371, pp.9-20
Foster, T., Gillespie, K., McClelland, R. (1997), ‘Mental disorders and suicide in Northern Ireland’, British Journal of Psychiatry, Vol. 170, pp.447-452
Gunnell, D., Peters, T., Kammerling, M. Brooks, J. (1995), ‘Relation between parasuicide, suicide, psychiatric admissions and socioeconomic deprivation’, British Medical Journal, Vol. 311, pp.226-230
Hawton, K., Fagg, J., Simkin, S., Mills, J. (1993), Attempted suicide in Oxford 1991 & 1992, University Department of Psychiatry, Warneford Hospital and Barnes Unit, Department of Psychological Medicine, John Radcliffe Hospital, Oxford
O’Carroll, P.W. (1989), ‘A consideration of the validity and reliability of suicide mortality data’, Suicide and Life threatening behaviour, Vol. 19, No. 1, pp.1-16
OPCS (1992), Mortality statistics, HMSO: London
OPCS (1996) Mortality statistics, HMSO: London
Platt, S. (1992), ‘Epidemiology of suicide and parasuicide’, Journal of Psychopharmacology, Vol. 6, No. 2, Sppl., pp.291-299
Platt, S., Beckett, S. & Kreitman, N. (1988), ‘Social construction or causal ascription: distinguishing suicide from undetermined deaths’, Social Psychiatry and Psychiatric Epidemiology, Vol. 23, pp.217-221
Platt, S., Hawton, K., Kreitman, N., Fagg, J., Foster, J. (1988), ‘Recent clinical and epidemiological trends in parasuicide in Edinburgh and Oxford: a tale of two cities’, Psychological Medicine, Vol. 18, pp.405-418
Stack, S. (1996), ‘Effect of female labor force participation on female suicide attitudes’, Death Studies, Vol. 20, (3), pp.285-291
Ussher, J. (1991), Women's madness: misogyny or mental illness? Harvester Wheatsheaf: New York
Sarah
Payne and Rachel Lart
Centre
for Health and Social Care
School for Policy
Studies
University
of Bristol
8
Priory Road
Bristol
BS8 1TZ
Tel:
(0117) 9546750/9546775
E-mail:
sarah.payne@bristol.ac.uk
E-Mail:
r.lart@bristol.ac.uk
