Preventing suicide: what will work and what will not

Stephen J Rosenman

Population, not individual, approaches to suicide prevention

MJA 1998; 169: 100-102
See also Dudley et al
 

Suicide is the second most common cause of death among young men in the Western world.1 The emotional impact of suicide, especially of young people, on the community has produced a "do something, do anything" flurry of preventive activity, directed mostly at identifying and intervening with people at high risk.2,3 Much of the activity is ineffective4 and some may even make things worse.5-7 Here, I argue that focusing on people at "high risk" squanders resources, and that we may have to abandon this approach to suicide prevention as an immediate health target.   Can we identify and assist high-risk individuals? Focusing prevention on people at "high risk" of suicide appears efficient because it seems to concentrate on those most likely to benefit. But can we identify high-risk individuals? Dimensional risk factors, such as hypertension or depression, have a linear relationship to mortality; only an arbitrary cut-off separates high risk from low risk.8 For conditions with multiple risk factors, such as coronary heart disease and also suicide, each factor adds a little to the risk, but often only when it interacts with other factors. No single predictor or combination of predictors is present in every individual, and membership of the high-risk group changes from moment to moment. Half a bottle of whisky may create a high suicide risk within an hour. As we learn to specify risk more precisely, the identified high-risk group gets smaller and -- while the individual risk of each member increases -- as a group they contribute a diminishing proportion of all suicides. Many more suicides come from the very much larger low-risk population. In a hypothetical population of 100 000, we may identify 500 people with a high suicide risk of, say, 10% (the 10-year suicide risk in schizophrenia, for example) and the remaining 99 500 with a low risk of, say, 0.1% (the approximate 10-year suicide risk in the population). The high-risk group will produce 50 suicides, while 100 will come from the low-risk group -- prevention targeted at high-risk people will miss most of the suicides. Another problem appears once we identify high-risk individuals. They then need advice and treatment that is effective, available and acceptable to them. How do we measure effectiveness if we can count only the failures? If we look at population suicide rates, agencies such as telephone suicide crisis services seem ineffective.4 Intervention after suicide attempts has not only been of little effect,9 but is also irregular in its availability, and frequently ignored by those it is intended to benefit.10,11 Those deemed at high risk, such as those discharged from hospital after a suicide attempt, already receive intensive follow-up, and further improvements will show diminishing returns of people saved.12 However, the insuperable problem in the high-risk approach arises when we try to use common characteristics of history and personality to identify individuals at risk of events that are uncommon in statistical terms. For events as infrequent as suicide, even highly specific and sensitive predictions will have high error rates.13 Theoretically, suicide predictions with 99% specificity will predict 1000 suicides in a population of 100 000; but the current suicide rate of around 15 per 100 000 population means that, in a year, fewer than 20 will be correct predictions. In the real world, where 85% is "good" specificity, those who will ultimately die are hidden in a haystack of people "at risk".   If a high-risk approach is ineffective, is any other strategy better? Newer approaches and paradigms for prevention in public health have not fully penetrated policymaking for mental disorders. William Haddon, a former Director of the United States National Highway Safety Bureau, propounded principles for injury prevention which widened the focus from the victim to the array of environmental interactions amenable to change (see below).14 More radically, Geoffrey Rose, an epidemiologist at the London School of Hygiene, switched the spotlight away from individuals at high risk of suicide and turned it onto whole populations.15 He argued that mortality falls more if we reduce the whole population's exposure to factors related to suicide than if we identify and treat high-risk people.   Time to consider a population-based approach? The principle of the "population-based" approach is to apply preventive strategies to the whole population or subpopulation in order to shift the distribution of risk, so that at any time fewer people are over the threshold that leads to completed suicide16 (see Figure). We do not have to identify the individuals within the population to save them. This is demonstrated by the successful reduction in mortality from road trauma and from cardiovascular disease. These successes should now be models for suicide prevention. Road trauma deaths in Australia decreased after the apparently inexorable rise seen in the 1960s,17 and this decline clearly followed the introduction of measures that applied to the whole population. These measures -- seat belts, random breath tests for alcohol, speed-limit enforcement, and improvements to roads and cars -- succeeded where strategies to identify the "high-risk" individual -- the dangerous driver, "the nut holding the steering wheel" -- had failed or aggravated the problem. Each measure, ranging from those most proximate to the crash (seat belts, collapsible steering columns) through the antecedent conditions (alcohol, speed) to the most nebulous antecedents, such as public attitudes to road safety,18 contributed its bit to the sum of prevention, and mortality fell without anyone identifying a high-risk individual or tackling the chain of causes which explain an individual death. Indeed, by 1968, Haddon had already challenged myths of accident prevention: ". . . because drivers cause most accidents [we assume that] programs correspondingly must be concerned with drivers. In the real world there is no basis for making this assumption [which] leads to demonstrably false conclusions."19 The decline in cardiovascular disease mortality is not explained by the vigorous treatment of high-risk people. Established risk habits such as smoking and diet were strikingly resistant to change in high-risk individuals.20 More importantly, most cardiac deaths occur among the large numbers of lower-risk people. Overall cardiovascular mortality fell with the change in community-wide attitudes to cigarettes, diet, inactivity and stress.21 At first sight, suicide prevention in these terms seems much more difficult because the antecedents of suicide appear harder to discern and change. However, if we take what we do know about suicide risk and apply it to the population rather than to defining the individual at risk, reducing suicide may become feasible.4,12   How do we put a population-based approach into practice? Restricting access to the means of suicide (eg, by detoxification of domestic and vehicle exhaust gases, gun regulation, control of dangerous medication) will tackle the most proximate antecedents. Close antecedents susceptible to reduction include drug and alcohol abuse; the population load of mood disturbances, including clinical and subclinical depression;22-24 conduct disorders;22 and media modelling of suicide.25 The more remote antecedents in all age groups include socioeconomic deprivation26 and, to a surprisingly uncertain degree, unemployment.4,27 In younger populations, family breakdown and alienation, relationship losses, school failure and suspension may portend suicide,28 while in older populations loneliness and physical morbidity may do so.29 In Aboriginal populations, suicide is prefigured by substance abuse and high incarceration rates, which are problems for the whole community to address, not just the problems of high-risk individuals.30,31   What to do and where to start? Haddon and Baker14 assert that we should first do those things we can, rather than struggling with things we cannot define or alter. Attacks on access to means of suicide, such as firearms32 and car exhausts,33 are within our present reach. By a large margin, they will have the strongest and most immediate effect in reducing suicide rates,4 even though such measures may not strike at what might seem the most "important" emotional antecedents. The problem with social antecedents, such as alienation or drug abuse, is that they demand difficult social engineering. It is unpromising to try to reconstruct families after they have broken down. Reducing the incidence of family breakdown and other such antecedents in the community may not be easier, but may be more profitable34 in terms of prevention, and for this medical efforts must harmonise with other social and educational initiatives. Of course, these antecedent problems (and others such as mental illness) warrant action on their own merits, not just as targets in suicide prevention initiatives. We cannot comfortably assume that education campaigns will always help, and there are good arguments against tackling prevention through "suicide prevention" or "education" campaigns. In the past, school-based driver education in the United States was associated with unexpectedly higher road accidents and deaths.35 Suicide education campaigns in schools not only miss excluded students, but have also shown complex and possibly harmful effects on attitudes to suicide6,10 and on completed suicide.7   The final analysis Firstly, identifying and treating high-risk individuals is unlikely to result in lower suicide rates. The refinement of individual assessment has passed the point of diminishing returns, and the obsessive study of suicidal individuals will not uncover the Holy Grail of perfect prediction. Clearly, we will not stop doing our best for distressed and suicidal people, even though this will not reduce population suicide rates. But this is not prevention. Secondly, pessimism about suicide may not be justified, despite the rising rates. The turnaround in the rising road toll and in cardiovascular disease teaches what can be done. The lesson is that measures which reduce overall risk in the whole population will reduce the number of people above the fatal threshold, and we do not need to identify the high-risk people individually. Consequently, we may have to abandon the frontal assault on suicide. We cannot justify prevention campaigns driven by the suicidal risk of individuals, despite the intuitive and political appeal of such measures. Indeed, to the degree that they drive resources into ineffective strategies, current approaches to "suicide prevention" may impede suicide prevention. Instead, we must "bite the bullet" in restricting access to means of suicide, the most proximate factor. Beyond that, we need the diligent, unspectacular work in the population which mitigates those factors which lead, among other things, to suicide. For medicine, it is to treat the ill,36 whether or not they are suicidal, and, from a public health pulpit, to address the social ills which produce morbidity, whether or not they lead to suicide.   References Pritchard C. New patterns of suicide by age and gender in the United Kingdom and the Western World 1974-1992; an indicator of social change? Soc Psychiatry Psychiatr Epidemiol 1996; 31: 227-234. CDC. From the Centers for Disease Control and Prevention. Suicide among children, adolescents and young adults -- United States 1980-1992. JAMA 1995; 274: 451-452. Commonwealth Department of Health and Family Services (Australia). Youth suicide in Australia: the National Youth Suicide Prevention Strategy. Canberra: Australian Government Printing Service, 1997. Gunnell D, Frankel S. Prevention of suicide: aspirations and evidence. BMJ 1994; 308: 1227-1233. Shaffer D, Garland A, Vieland V, et al. The impact of curriculum-based suicide prevention programs for teenagers. J Am Acad Child Adolesc Psychiatry 1191; 30: 588-596. Hazell P, King R. Arguments for and against teaching suicide prevention in schools. Aust N Z J Psychiatry 1996; 30: 633-642. Lester D. State initiatives in addressing youth suicide: evidence for their effectiveness. Soc Psychiatry Psychiatr Epidemiol 1992; 27: 75-77. Patton GC, Harris R, Carlin JB, et al. Adolescent suicidal behaviours: a population based study of risk. Psychol Med 1997; 27: 715-724. Deykin E, Chung-Chen H, Joshi N. Adolescent suicidal and self-destructive behaviours: results of an intervention study. J Adolesc Health Care 1986; 7: 88-95. Shaffer D, Garland A, Gould MS, et al. Preventing teenage suicide: A critical review. J Am Acad Child Adolesc Psychiatry 1988; 27: 675-687. Goldney RD. Out-patient follow-up of those who have attempted suicide: fact or fantasy? Aust N Z J Psychiatry 1975; 9: 111-113. Lewis G, Hawton K, Jones P. Strategies for preventing suicide. Br J Psychiatry 1997; 171: 351-354. Pokorny AD. Prediction of suicide in psychiatric patients. Arch Gen Psychiatry 1983; 40: 249-257. Haddon W, Baker SP. Injury control. In: Clark DW, McMahon B. Preventive and community medicine. Boston: Little Brown, 1981. Rose G. The strategy of preventive medicine. Oxford: Oxford University Press, 1985. Rose G. Mental disorder and the strategies of prevention. Psychol Med 1993; 23: 553-555. Federal Office of Road Safety (FORS). Trends in road safety. Monograph No 1. Canberra: FORS, 1997. Graham JD. Injuries from traffic accidents: Meeting the challenge. Annu Rev Public Health 1993; 14: 515-543. Haddon W. The changing approach to the epidemiology, prevention and amelioration of trauma: the transition to approaches etiologically rather than descriptively based. Am J Public Health 1968; 58: 1431-1438. Susser M. The tribulations of trials: Intervention in communities [editorial]. Am J Public Health 1995; 85: 156-158. Epstein FH. Cardiovascular disease epidemiology: a journey from the past into the future. Circulation 1996; 93: 1755-1764. Shaffer D, Gould MS, Fisher P, et al. Psychiatric diagnosis in child and adolescent suicide. Arch Gen Psychiatry 1996; 53: 339-348. Depression and suicide: are they preventable [editorial]? Lancet 1992; 340: 700-701. Brent DA, Perper JA, Goldstein CE, et al. Risk factors for adolescent suicide. Arch Gen Psychiatry 1988; 45: 581-585. Phillips DP, Carstensen MS. Clustering of teenage suicides after television news stories about suicide. New Engl J Med 1986; 315: 685-689. Gunnell DJ, Peters TJ, Kammerling RM, Brooks J. Relation between parasuicide, suicide psychiatric admissions and socioeconomic deprivation. BMJ 1995; 311: 226-230. Platt S. Unemployment and suicidal behaviour: a review of the literature. Soc Sci Med 1984: 19; 93-115. Gould MS, Fisher P, Parides M, et al. Psychosocial risk factors of child and adolescent completed suicide. Arch Gen Psychiatry 1996; 53: 1155-1162. Draper BM. Prevention of suicide in old age. Med J Aust 1995; 162: 533-534. Royal Commission into Aboriginal Deaths in Custody. National Report. Canberra: Australian Government Publishing Service, 1991. Hunter E. An examination of recent suicides in remote Australia: Further information from the Kimberley. Aust N Z J Psychiatry 1991; 25: 197-202. Kellerman AL, Rivara F, Somes G, et al. Suicide in the home in relation to gun ownership. New Engl J Med 1992; 327: 467-472. Rosenman SJ. Car exhaust suicide. Med J Aust 1997; 166: 288-289. Zigler EF, Taussig C, Black K. Early childhood intervention: a promising preventative for juvenile delinquency. Am J Psychol 1992; 47: 997-1006. Robertson LS. Crash involvement of teenaged drivers when driver education is eliminated from high school. Am J Public Health 1980; 70: 599-603. Wilkinson G. Can suicide be prevented? Better treatment of mental illness is more appropriate aim. BMJ 1994; 309: 860-861.

Authors' details

Reprints: Dr Stephen J Rosenman, GPO Box 610, Canberra, ACT 2601.
E-mail: sjrATatrax.net.au