Bulletin of the World Health Organization

Methods of suicide: international suicide patterns derived from the WHO mortality database

Vladeta Ajdacic-Gross ^a, Mitchell G Weiss ^b, Mariann Ring ^a, Urs Hepp ^c, Matthias Bopp ^d, Felix Gutzwiller ^d, Wulf Rössler ^a

Introduction

Restricting access to the means of suicide is an important component of comprehensive strategies for suicide prevention. To improve prevention efforts, better knowledge of national, regional and local suicide patterns is vital, and better understanding of underlying mechanisms is absolutely crucial.

National studies on suicide indicate that suicidal behaviour and, in particular, the preferred suicide method, varies between countries. Some patterns are well known, such as the high percentage of firearm suicides in the United States of America.¹ In addition, the role of pesticide suicide in Asian countries became apparent in the 1990s.²^,³ The emergence of a new method, charcoal-burning suicide, in Hong Kong Special Administrative Region (SAR), China and urban Taiwan, China,⁴^,⁵ has been a surprise, but serves as a warning that research and prevention efforts should be reinforced.

While numerous factors contribute to the choice of a suicide method, societal patterns of suicide can be understood from basic concepts such as the social acceptability of the method (i.e. culture and tradition) and its availability (i.e. opportunity).⁶^,⁷ International or intercultural comparisons of suicide methods help increase understanding of the interplay between these two factors and provide a basis for preventive strategies.⁸^–¹⁰

The aims of this study were:

to provide, for the first time, a comprehensive international comparison of suicide methods using WHO mortality data (i.e. data collected in a standard way);
to use correspondence analysis (a data reduction technique) to identify underlying patterns in the data;
to contribute to discussions on the impact of the availability and acceptability of suicide methods on suicide behaviour;
to draw conclusions about suicide prevention.

Methods

The data presented below were derived from the WHO mortality database (http://www.who.int/whosis/mort/en/index.html) on 17 November 2006. The database includes data gathered since 1950 and is dedicated to collecting and making available data on the underlying cause of death in all Member States using a standard format. Several countries have not reported mortality data to the WHO, or have reported data covering only a few years.¹¹ Others have reported only partial data, for example, on subpopulations or regions. However, the database also provides estimates of the completeness and coverage of mortality data for each reporting country. Data for industrialized countries are most complete, whereas those for developing countries and conflict regions are often incomplete or missing altogether.

For this study, detailed information about suicide methods was extracted from WHO mortality data only if it was encoded according to the International Classification of Diseases, 10th revision (ICD-10). Since not all countries report mortality data to this standard, the study relied on only a selection of countries. In particular, countries from Africa were absent, except South Africa. In addition, countries with small populations, and thus few suicides, were excluded. In total, 56 countries and all the years covered by ICD-10 data, 1994–2005 at most, were included in the analysis, though the time span varied slightly from country to country (Table 1, available at: http://www.who.int/bulletin/volumes/86/9/07-043489/en/index.html).

The ICD-10 codes for suicide are in the range X60–X849. We differentiated suicides due to a pesticide or an unspecified poison (X68–X699), other poisons (X60–X679), hanging (X70–X709), drowning (X71–X719), firearms and explosives (X72–X759), and jumping from a height (X80-X809), and other suicide methods. Some methods, such as hanging or firearm suicide, are reported more accurately than others, such as poisoning or drowning.¹² This should be kept in mind when interpreting frequency distributions, particularly in countries with incomplete data.

To provide an overview of suicide methods by country, we calculated the number of suicides with each method as a proportion of the total number of suicides for each sex and country. In addition, we used correspondence analysis (CA)¹³^,¹⁴ to summarize information from cross-tabulated data. In practice, CA converted data into a graphical form that grouped countries with similar suicide method patterns. Metaphorically, a new map of the world is emerging in which countries are grouped into quasi-continents on the basis of suicide methods (or groups of suicide methods). The analysis was repeated separately for European and non-European countries.

Correspondence analysis is an exploratory multivariate technique that is very similar to principal component analysis (PCA). In CA, the ‘extraction’ of dimensions is related to the |² distance, which is the measure of association (analogous to the correlation in PCA). A minimum number of dimensions is sought to account for the maximum amount of the inertia (analogous to the total variance in PCA). The inertia is a weighted measure of the total |² (that is, |²/N).

Correspondence analysis relies on standardized row or column profiles, or both. Because of standardization, the proportion of suicides by a particular method in each country contributes similarly to the CA results. Typically, the first or the first two dimensions extracted in CA are used to display the results in correspondence maps. Scores on these dimensions represent distances from the average row and column profile (i.e. the country and method profile, respectively). The distance between any two points is analogous to the weighted Euclidian distance.

Despite symmetrical standardization along both the row and column profiles used in this study, interpretations based on the cross-comparison of distances between countries and suicide methods should be avoided. It is advisable to focus on similar directions from the zero point, however, and to consider the distance from the zero point separately for each variable.

Among the problems which may arise in applying CA is the possibility that a straight sequence of objects or data points may be displayed in a curvilinear or pyramidal way. This phenomenon is called the ‘arch effect’, which derives from unimodal response curves or from the compression of distances at the ends of an axis, generally the first axis. In such cases, only the first axis, rather than two axes, is included in data interpretation. Alternatively, another CA technique (e.g. detrended CA) can be used.

Table 1. Suicide by method (% out of all suicides by country) according to the WHO mortality database (as at November 2006); countries reporting ICD-10 data
html, 50kb

Results

Hanging was the predominant method of suicide in most countries included in the analysis (Table 1). The highest proportions were around 90% in men and 80% in women, as observed in eastern Europe (i.e. Estonia, Latvia, Lithuania, Poland and Romania). There were a number of deviations from the predominant pattern. As might be expected, firearm suicide was the most common method in the United States, but was also prevalent in Argentina, Switzerland and Uruguay, although only men used this method in Switzerland. Jumping from a height (designated as falls in the figures) plays an important role in small, predominantly urban societies such as Hong Kong SAR, Luxembourg and Malta. In contrast, in rural Latin American countries (e.g. El Salvador, Nicaragua and Peru), Asian countries (e.g. the Republic of Korea and Thailand) and also in Portugal, poisoning with pesticides was a major problem, notably among women. Poisoning with drugs was common in women from Canada, the Nordic countries and the United Kingdom. It also played an important role in male suicide in these countries.

The well-known differences between men’s and women’s preferred suicide methods can also be seen in Table 1. Violent and highly lethal methods such as firearm suicide and hanging are more frequent among men, whereas women often choose poisoning or drowning, which are less violent and less lethal.

Correspondence analysis for all countries yielded the correspondence maps shown in Fig. 1 for men, and Fig. 2 for women. In the men’s data, the first two dimensions accounted for 84.5% of the inertia; in the women’s data, it accounted for 77.1%. In the figures, pesticides, hanging (and other methods) and firearms form a triangle. In addition, the countries with the highest proportions of pesticide poisoning and firearm suicide listed above also define the triangle apices. Hanging is in an intermediate position since it is more frequent than firearm suicide in almost all countries with a high proportion of pesticide suicide. Conversely, hanging is more frequent than pesticide suicide in almost all countries with a high proportion of firearm suicide. Thus, there are two slopes: one from hanging to pesticide suicide and the other from hanging to firearm suicide. This is the typical arch effect observed in CA. To be safe, we should not interpret these CAs using the second axis. The first axis should be kept in mind, on which the poles are pesticide suicide and firearm suicide (and, to a lesser extent, other poisoning), with hanging and other methods lying in between.