Akihiro Nishio, MD, Department of Psychiatry, Gifu University Graduate School of Medicine, 1-1 Yanagido, Gifu 501-1194, Japan. Email: firstname.lastname@example.org
Aims: The relationship between suicide and disaster is an important problem but it's not clear. We conducted this study to determine whether a natural disaster affects suicide rates.
Methods: We collected data on suicides during the 84 months before and the 60 months after the earthquake and compared the suicide rate in Kobe to that in Japan as a whole. We also examined what groups were significantly affected.
Results: Compared with Japan as a whole, the suicide rates in Kobe significantly decreased in the 2 years after the earthquake.
Conclusions: An influence on suicide rate after the disaster clearly appeared in middle-aged men.
THE RELATIONSHIP BETWEEN suicide and disaster is an important problem.1 Since the English translation of Le Suicide in 1952, David Emile Durkheim has stood preeminently during the last half of the 20th century as the presumed founder of scientific suicidology.2 Durkheim's taxonomy of suicide flowed from the central theme about the regulative role of social control.3 The central hypothesis of Le Suicide is that, when social conditions fail to provide people with the necessary regulation of their lives, their psychological health is compromised and the most vulnerable among them commit suicide.3 He differentiated three types of suicide: anomic, egoistic and obligatory (or altruistic) forms, and suggested that a common external threat decreased anomie and hence the suicide rate,4 but it has not been proven.
In this study, we prospectively investigated the relationship between the suicide rate and a severe natural disaster, the 1995 Great Hanshin-Awaji earthquake. We addressed three issues: (i) the effects of the disaster on the suicide rate; (ii) how long the suicide rate was influenced after the disaster; and (iii) as Imamura (1995) suggested that the elderly and men have a tendency to kill themselves after a disaster,5 we also examined which groups were significantly affected.
MATERIAL AND METHODS
We set a 7-year baseline period before and 5-year follow-up period after 1995 when the earthquake occurred and calculated the number of suicide victims monthly and yearly during these 12 years (1988–1999). We added up the total number of suicide victims for each month. The differences between the pre- and the post-disaster suicide rates were also calculated according to gender and age (young: 20–39 years, middle: 40–59 years, and elderly: 60–79 years).
The monthly populations of totals, male, female and aged-groups in Japan were obtained from the Journal of Health and Welfare Statistics. Kobe-specific monthly population totals, males and females were compiled from the Statistical Data of Kobe published by the Planning and Coordination Bureau, General Planning and Research Division, Kobe City Office. As for age group populations in Kobe, the information from the 1990 national census was used before the earthquake (1988–1994 years), and the information from the 1995 and 2000 national censuses was estimated by a linear approximation for five years after the earthquake (1995–1999 years).
For comparison purposes, we made similar calculations for Japan as a whole using the monthly reports on the Vital Statistics of Japan published by the Statistics and Information Department, Minister's Secretariat, Ministry of Health, Labor and Welfare in Japan.6
Multiple regression analysis was performed to relate the trend of suicide rates to the corresponding years. The dependent variable was the difference between the monthly suicide rates (per one million) of Kobe and Japan and the candidates of the independent variables consisted of the twelve dummy variables Zi (i = 1, ----, 12) indicating each year of 1988 through 1999. Forward stepwise regression methods were performed to select the best subset of year variables associated with the change in suicide rates.7 Partial F tests were used to test whether the year variable enhances the prediction of the dependent (the change in suicide rates) variable by entering into the regression model. The F-to-enter and F-to-remove limits were defined as 3.92 and 3.90, respectively. The data were analyzed using the statistical software BMDP 9R.8
The present study was approved by the ethical committee of Niigata University Graduate School of Medical and Dental Sciences.
Comparisons of suicide rates between Kobe and Japan as a whole (1988–1999)
The upper part of Fig. 1 shows the annual suicide rates (per 100 000) in Kobe and Japan as a whole during the 7 years (1988–1994) preceding the earthquake and also between 1995 and 1999. The suicide rate of 1995 in the total population of Kobe was the lowest in the last 54 years (1946–1999), that is, since the conclusion of World War II. The regression analysis disclosed that the two years of 1995 and 1996 showed significantly lower suicide rates in Kobe than those in Japan as a whole.
The lower part of the figure represents the monthly changes in the suicide rate in Kobe and in Japan as a whole covering the 12 years. During 1995 the suicide rate gradually decreased between August and September and increased in December again. As for 1996, the minimum and maximum rates were in May and October, respectively, although it is difficult to find a tendency of seasonal variation because of the marked fluctuation.
The age groups whose suicide risk decreased after an earthquake
For years that showed significant post-disaster decreases in the suicide rates, we examined whether these decreases were greater for a particular gender or age group and how long, if any, the post-disaster decreases continued. We compared the annual changes in suicide rates in each group pre- and post-earthquake in Kobe as well as in Japan as a whole, between 1988 and 1999. Our study shows that middle-aged people had a significant decrease in the suicide rate at one year (1995 year) after the earthquake, but young and elderly people did not. We also compared the changes in the male and female suicide rates before and after the earthquake in Kobe with Japan as a whole, between 1988 and 1999, and obtained the following results: the male suicide rates significantly decreased in the 24 months (1995 to 1996) following the earthquake, while the female suicide rates did not change significantly. In particular, the middle-aged and elderly male groups were the most severely affected for the two years following the earthquake.
The major findings of our study were the following two results: (i) The 1995 Hanshin-Awaji earthquake led to decreases in the total suicide rate for the two years following the disaster. This result appears to support Durkheim's thesis. (ii) The middle-aged males groups were the most severely affected by the suicide decline.
In 2004, Shoaf et al. examined the suicide rate in Los Angeles County in relation to the Northridge earthquake and found that the rates of suicide were significantly lower in the three years following the earthquake than they were in the three years prior to the earthquake.9 On the contrary, two studies indicated increasing suicide rates after the Taiwan earthquake.10,11 As for the suicide rate however, they compared victims to non-victims, or affected counties to control counties. On the other hand, Shoaf's study and ours used a comparison between annual suicide rates before and after the earthquake in the affected area. In addition, there are some differences in sample size, severities of earthquake, and ethnicity among these studies. Our study appears to demonstrate Durkheim's thesis.4
In the present study, middle-aged men were at a decreased risk of suicide after the earthquake. However, there is a possibility that middle-aged men left Kobe in search of work and died of a disease.
In summary, we found that the suicide rate was lower for the 2 years following a devastating natural disaster, and that an influence on the suicide rate after the disaster clearly appeared in middle-aged men who were the high-risk group for suicide.
This study was supported by grants from the Japan Society for the Promotion of Science (JSPS) (No. 19203031) and the Research Center for Natural Hazards and Disaster Recovery, Niigata University, Japan. We wish to thank Siu Wa Tang, MD, PhD and Yoshitugu Tatsuno, MD, for their special contribution in advancing this study.