Re: Impact of 2008 global economic crisis on suicide: time trend study in 54 countries
Chang et al’s study  provides challenging information on the impact of the 2008 economic crisis on suicide mortality in a very large sample of countries. It is the first assessment of this impact with a constant method across different countries all over the world. Nevertheless, we would like to draw readers’ attention to some methodological issues.
Firstly, according to what is explained by Chang et al in their paper, we can deduce that the confidence intervals of the rate ratios of suicides in 2009 were estimated on the basis of the variance of the expected number of suicides in 2009 if 2000-2007 trends had continued. However, the variance of the rate ratio of suicides depends not only on the variance of the expected number of suicides, but also on the variance of the observed number of suicides. Under the null hypothesis that there was no crisis effect, the observed number of suicides in 2009 is supposed to follow, as in previous years, an overdispersed Poisson distribution, and hence, to have a variance of larger magnitude than the expected number of suicides. Not taking into account this part of the variance could be, in a way, comparable to not taking into account the number of cases in the estimation of the confidence interval of an odds ratio in a case-control study. We compared Chang et al’s results with a more conventional method on a sample of eight EU countries (Austria, Finland, France, Germany, the Netherlands, Spain, Sweden, and the UK). In the Table, we present:
A - the results published in Chang et al’s paper (we found closely comparable results when we reproduced their model);
B - the results of the most conventional way to estimate the rate ratio and its confidence interval, with a negative binomial log-linear model including a dummy variable for 2009.
The 95% confidence intervals of the rate ratios obtained with the model used in (B) take into account the variance of the observed number of suicides. In the eight analysed countries, when Chang et al found significant associations in men in all countries, and in Austrian, Finnish, and Dutch women, we found no significant association in women, and a significant association only in German, Spanish and British men. The confidence intervals found by Chang et al may therefore be too small, and may conduce to spurious conclusions.
Secondly, we found that the results seem to be very sensitive to the model used. In particular, they vary according to the interaction terms included in the model (e.g. a specific time trend by age). It seems difficult to find the model that fits better to the observed data of mortality, and that takes the necessary interactions into account. The inconsistency of the association between suicide mortality and the 2008 economic crisis according to the statistical methodology was already described in England: divergent associations were found by Barr et al  and Saurina et al . Saurina concluded that the inconsistency of the estimates may result from a lack of control for potential confounders.
Thirdly and finally, Chang et al suggested that their results were in favour of a dose-response effect of unemployment on suicide mortality. The individual unemployment-suicide association has already been well documented: the risk of suicide is increased among the unemployed, but this may be a combination of causal and self-selection processes (with notably the impact of psychiatric disorders) . Ecological studies on mortality databases, as Chang et al's study, are not designed to assess an individual link, but the combination between an individual and a contextual (macro-social) effect. In 1990, Crombie hypothesized that unemployment does not explain all the rise in suicide mortality during increasing unemployment periods, and that the unexplained part could result from the more generalized effects of economic recession . Then he concluded that an increase in suicide rates is not an inevitable consequence of increased unemployment.
Thus, if unemployment explains probably a part of suicide increase during the 2008 economic crisis, it remains to identify the other factors explaining this increase, in order to implement efficient prevention policies targeting these factors. One of these factors could be the prevalence of psychiatric disorders and the quality and effectiveness of their treatment . Another explanation could be the impact of suicide reporting in the media, with an imitative effect (also called Werther effect) which has been largely reported . The World Health Organization published guidelines for media professionals to limit these side effects . In opposition, a protective effect (called Papageno effect) of reporting positive behaviour in adverse circumstances has also been reported .
The association between the 2008 crisis and the increase in suicide rates has to be considered with care. Multiple causes may be involved, not only the rises in unemployment rates, and these factors can have an impact at an individual and/or a contextual scale. We have to be particularly careful on the statistical method used, as finding too small confidence intervals may lead to premature conclusions.
1. Chang S-S, Stuckler D, Yip P, Gunnell D. Impact of 2008 global economic crisis on suicide: time trend study in 54 countries. BMJ. 2013;347:f5239–f5239.
2. Barr B, Taylor-Robinson D, Scott-Samuel A, McKee M, Stuckler D. Suicides associated with the 2008-10 economic recession in England: time trend analysis. BMJ. 2012;345:e5142.
3. Saurina C, Bragulat B, Saez M, López-Casasnovas G. A conditional model for estimating the increase in suicides associated with the 2008-2010 economic recession in England. J. Epidemiol. Community Health. 2013;
4. Hawton K, van Heeringen K. The International Handbook of Suicide and Attempted Suicide. Chichester, England: John Wiley & Sons, Ltd; 2000.
5. Crombie IK. Can changes in the unemployment rates explain the recent changes in suicide rates in developed countries? Int. J. Epidemiol. 1990;19:412–6.
6. Gusmão R, Quintão S, McDaid D, Arensman E, Van Audenhove C, Coffey C, et al. Antidepressant Utilization and Suicide in Europe: An Ecological Multi-National Study. PloS One. 2013;8:e66455.
7. Preventing Suicide: A Resource for Media Professionals. WHO; 2008.
8. Niederkrotenthaler T, Voracek M, Herberth A, Till B, Strauss M, Etzersdorfer E, et al. Media and suicide. Papageno v Werther effect. BMJ. 2010;341:c5841.
Competing interests: No competing interests