Abstract
Elmar Etzersdorfer, Furtbachkrankenhaus, Hospital for Psychiatry and Psychotherapy, Stuttgart, Germany; Martin Voracek, Department of Psychoanalysis and Psychotherapy, University of Vienna Medical School, Vienna, Austria; and Gernot Sonneck, Department of Medical Psychology, University of Vienna Medical School, Vienna, Austria:
Sensational newspaper reports about suicides may trigger additional suicides (the ‘Werther effect’ [1, 2]). However, until now, a ‘dose–response’ relationship between regional newspaper distribution and magnitude of imitational effects has not been revealed [3, 4].
In late October 1990, the owner of a famous Viennese hotel, Hotel Sacher, committed suicide by firearm. Mass-media coverage of this celebrity suicide was extensive, although mostly in a moderate way, which has become common practice among the media following an increase in subway suicides in Vienna [5]. The only newspaper that reported the Sacher suicide in a very dramatic way, including headline coverage and photographs, was Austria's leading tabloid newspaper, the Neue Kronenzeitung (NKZ). It ran a very detailed and sensationalized serial about the ‘life and death’ of the suicide victim, covering two pages per issue for five consecutive days.
The NKZ is read by almost 50% of the population, although there is substantial regional variation. This reporting, in conjunction with the distribution of the NKZ, provided the opportunity for a natural experiment. We obtained Austrian suicide data, including method and region, within a time-frame aligned ± 3 weeks around this celebrity suicide for 1990 (n = 223) and for a control period (1981–1989, n = 2051), and regional NKZ distribution data.
In the control period (1981–1989), use of firearm as suicide method declined slightly for regions both of ‘low’ and ‘high’ distribution of NKZ (from 10.8% before to 7.9% after that day for ‘low’ regions, and from 16.0% to 15.8% for ‘high’ regions, respectively). In 1990 in ‘low’ regions the proportion of suicide by firearm was 15.8% of all suicides prior to the celebrity suicide, whereas no case occurred during the 3 weeks thereafter. For the ‘high’ regions a statistically significant increase from 13.2% to 34.1% was found (OR = 3.41, 95% CI = 1.11–10.56) during this period, which represents a tripling of the numbers (from five to 15 suicides). Total suicides by firearm in Austria increased from 17 to 28 cases; however, the total number of suicides increased only slightly from 107 to 116 cases.
Further regional-level analysis revealed a strong association between NKZ distribution and an increase of suicides by firearm (log OR) after the NKZ serial (Pearson's r = 0.62, n = 9 regions, one-tailed p = 0.04). In other words, nearly 40% of the variance in changes in suicide by firearm was attributable to differential NKZ distribution. There was no such association for the control period.
This finding of a ‘dose–response’ relationship provides further evidence of the influence of media reports on suicidal behaviour from a new perspective, lending support to efforts to achieve more moderate reporting in mass media [1–5].