In suicide prevention one of the recognised public health approaches is responsible media reporting on suicidal behaviours [1,2]. Notions about suicidal contagion after reporting cases of suicide in newspapers go back to 19th century medical literature [3]. In the literature, suicide contagion is also referred to as imitative, copycat or mass cluster suicide. Associations between media portrayal and suicidal behaviours has been a subject of research for decades. Although the media is only one feature of the social environment in which suicidal behaviours can be learned [4] and the effect is probably smaller than that of other psychosocial risk factors for suicide [5], it is a significant agent in social construction of reality, especially for vulnerable persons.

Several countries and organisations—e.g., the World Health Organisation, the Samaritans, the American Foundation for Suicide Prevention—have launched and disseminated the resources to educate and empower media professionals [6,7,8,9,10]. Studies are available, which are cross-sectional and analyse whether the method of reporting about suicidal behaviours in newspapers is in line with the recommendations of international best practices [11,12,13].

There is also evidence, albeit from fewer studies, that modification of reporting on suicidal behaviour is feasible and can be effective [10,14]. Several studies have measured the style of media reporting about suicide before and after recommendations for media were launched [9,15] or before and after the trainings for editors and journalists were given [16,17]. Even if the intensity of reporting about suicide can increase after these interventions, the quality in terms of preventive accent of media coverage tend to improve [17]. As these studies do not include clear outcome measure related to suicidal behaviours, the conclusions about possible suicide preventive impact of such interventions remain ambiguous. The question how does modification of media reporting influence actual suicidal behaviours is much more interesting and challenging.

The aim of the current systematic review was to monitor and give an overview of the research performed about the roles of media in suicide prevention in order to find out possible effects media reporting on suicidal behaviours might have on actual suicidality, that is on completed suicides, attempted suicides and suicidal ideation.

The systematic review was performed following the principles of the PRISMA statement [18]. The results of the identification, screening, eligibility assessment, and inclusion of the articles is presented using PRISMA 2009 Flow Diagram with minor modifications (Figure 1).

A literature search of publications included in the electronic databases was conducted in July 2011 using MEDLINE (via PubMed), PsychINFO (via EBSCOHOST) and Cochrane Library. The search criteria was the occurrence of the combination of two keywords—suicid* and media—either in the title, abstracts or keywords of the publication. In total, 1,180 articles were identified through database searching (MEDLINE n = 588, PsychINFO n = 584, Cochrane Library n = 8). After the duplicates were removed, the number of publications included into the screening process was 981.

At the first stage of the screening all non-English articles were excluded (n = 82). Thereafter, all publications were screened based on the title and abstract in order to assess the relevance of the subject and eligibility. The rationale behind defining the eligibility criteria was that the influence of fictional suicide stories presented in the media has been found more controversial than the influence of non-fictional stories [19] and that purely descriptive research is unable to provide evidence about the real effect of media reporting on actual suicidal behaviours. Thus, the eligibility criteria for the inclusion were: (1) the article is a research article, a meta-analysis or a systematic review (letters, communications, descriptions of research trends and resources were excluded), (2) the research material is based on non-fictional media portrayal of suicidal behaviours, (3) the analysis of media reports has been linked with some suicidality-related outcome measure (e.g., suicide rate, non-fatal suicidal behaviours, suicidal ideation).

In this stage of the screening 818 clearly non-relevant articles were excluded and full-texts of 81 articles were retrieved and further assessed for eligibility. After final eligibility assessment, 56 full-text articles were included in the study. Heterogeneity of the search results limited quantitative synthesis (meta-analysis) of the articles and therefore only a qualitative synthesis was performed and key findings presented.

The main characteristics and findings of the 56 studies included in qualitative synthesis are presented in Table 1. Of all studies four were meta-analysis, four were systematic reviews and 48 were research articles.

Most of the studies (n = 19) came from North-America (the USA and one from Canada), the studies from the USA were especially over-represented before 1990. The next big region of origin of the studies was Asia (n = 16), followed by Europe (n = 12). Two studies were performed in Australia. Meta-analysis and systematic reviews had no specifically defined region or included studies from several regions.

The vast majority of the studies support the idea that media coverage of suicidal behaviours and actual suicidality are associated. Only four studies found no significant associations [20,26,27,28] and five studies expressed hesitations about clear associations or reported incoherent results [29,30,36,50,73]. All studies reporting no associations were conducted before 1990. Unfortunately it is impossible to evaluate whether these one-way results, i.e., either positive or negative significant associations between media reporting and suicidality, reflect the real situation. There is a risk of reporting bias in the sense that researchers are eager to report meaningful positive results, but could keep silent if the results are not beneficial.

Mass media imitation theory presumes that if modelling works in one way (copycat suicides), it can work also on the other way (positive model) [74]. For indicating negative, provoking effect of media portrayal, the expression ‘Werther effect’ was introduced by Phillips [75] already decades ago and for the opposite, preventive effect the expression ‘Papageno effect’ was proposed recently by Niederkrotenhaler and colleagues [72]. The ‘Werther effect refers to Goethe’s novel The sorrows of Young Werther (1774), where a young man takes his life for love by shooting himself. The ‘Papageno effect’ refers to Mozart’s opera The Magic Flute (1791), where a young man in love becomes suicidal, but copes well thanks to his friends’ intervention. Based on the results of the current systematic review more research is available about provocative outcomes of irresponsible media reports inducing an increase in suicidal behaviours after publicized suicide stories than for protective effects. Only six research articles evaluated the protective effect of media coverage, caused either by newspaper blackout, by reducing the quantity of reporting or by changing the quality of media reporting, and observed subsequent decrease in suicidality [10,21,35,38,42,72].

It has been argued that certain individuals are more vulnerable to incorporate the idea and act of suicide into their concepts of self and one possible source for learning is media [76]. Several studies included in the current review proposed the suicide contagion effect of media reporting to be present only for specific vulnerable groups, as the degree of media influence is contingent on audience receptivity [36]. Subsequent similar suicides indicate strong modelling effect based on age [21,30,31,58,66,67,68] and gender [21,53,58,61,67,68]. These notions can be explained by differential identification theory, which takes age and gender as key dimensions of social life upon people build a sense of identity [77]. The fact that media-induced contagious effect has been highlighted more among young people and especially among elderly leads to the conclusion that suicide rates of the middle-aged people are less related to imitative suicide [74,77]. This is probably due to relatively low level of suicidogenic life conditions for middle-aged people. Unlike elderly, the middle-aged people have comparatively good health and are financially well off. Compared to the young, they are more settled into society’s institutional framework, including marriage, the family, work and politics [77].

One important issue which can not be neglected when assessing media influence is reporting on specific suicide methods. Studies included in the current systematic review revealed how media reports ‘advertised’ several dramatic and highly lethal suicide methods: burning [44], charcoal burning [44,66,68,69], shooting [49], jumping [51,52,58], railway suicide [71], and subway suicide [35,38,42]. Studies about newsworthiness of suicide have found that media reports are not representative to official suicide data and tend to exaggerate certain types of suicides, like suicides by celebrities and suicides involving unusual circumstances or methods [78]. The methods of drowning, jumping, shooting and rare methods are more likely to be reported than hanging [79], which is actually the most frequent suicide method in many countries.

Most of the studies have focused on short-term effect (e.g., from 1–2 days to 3–4 weeks) of media reporting, which is highly relevant period for provoking fatal and non-fatal suicidal acts. However, suicidal ideation (suicidal thoughts in the community) have found to be influenced by a publicized celebrity suicide for a longer, approximately 1-year period [62]. The long-term effect becomes probably even more important in the contemporary world, where majority of the newspapers are electronic and easily accessible in the Internet for an undetermined period of time. The Internet is global medium, which makes it difficult to determine specific area and exact dose of media reporting. These facts challenge research methods for evaluating associations between media reporting and subsequent suicidal behaviours.

The majority of studies about associations between media and suicidality follow the tradition of quantitative evaluation of media reporting, which fails to take into account the capacity of audience to make meaning out of message. Quantitative method could be complemented by examining qualitatively the multiple meanings that audiences give to media messages [80]. Niederkrotenthaler and colleagues [72] have applied both content analysis and quantitative evaluation and reached to novel findings about possibilities to provide suicide-protective effect by advertising positive coping in adverse circumstances. Even more, the readers can be active agents in constructing reality in media. For example, online newspapers provide possibilities for interaction between the suicide story and the reader, but also between the readers themselves, which in turn can construct the reality of exposed persons. Some hints about social construction and interpretation of suicide stories presented in online media can be obtained by exploring the content of spontaneous readers’ comments on media portrayal of a suicide story [13,81].

The vast majority of the studies about possible effects media reporting on suicidal behaviours might have on actual suicidality (fatal and non-fatal suicidal acts or suicidal ideation) support the idea that these two phenomena are associated. However, there is a risk of reporting bias in the sense that only positive results could be considered as worth to publish and zero results may remain unknown. This is especially true for the studies performed during the last two decades and investigating merely a single suicide story. Only four studies [20,26,27,28] included in the current systematic review found no significant associations and all of them were conducted before 1990.

More research is available about how irresponsible media reports can provoke suicidal behaviours (the ‘Werther effect’) and less about protective effect media can have by newspaper blackout or by changing the quality and content of media reporting (the ‘Papageno effect’). Contemporary tendencies in public health encourage researchers to ascertain protective factors as opposed to risk factors, although most of the time these are just two sides of the same coin.

Strong modelling effect of media coverage on suicide is based on age and gender. Individuals with demographic background similar to the person who committed highly publicized suicidal act (in most of the cases celebrities) are more vulnerable and receptive for identification. As it is true for suicide prevention as a whole, the research on media effects should also be target group specific, because universal approaches are less promising.

Several studies have revealed how media reports tend to ‘advertise’ dramatic and highly lethal suicide methods (burning, charcoal burning, shooting, jumping, railway and subway suicides), which are rare in real life. Media reports are not representative to official suicide data and tend to exaggerate sensational suicides.

The studies about media reporting on suicide come from a limited number of countries, although from different regions of the world. Researcher should be encouraged to assess the situation in different countries, although the fact about globalisation of news via Internet should kept in mind. Future studies have to encounter the challenges the global medium Internet will offer in terms of research methods. Circulation of news in the Internet is not limited either spatially or in time. However, online media can provide valuable innovative qualitative research material like readers’ spontaneous comments on media portrayals, which enables to analyse different meanings audience gives to messages and also interactions between different agents in this process.

One possible limitation of the study is that if alternative search regimens were utilized, additional studies might have been found. Especially the studies from earlier periods (e.g., from 1960’s to 1980’s) are difficult to find using standard search engines, an example of such omission could be the study performed by Stack [82]. The bibliographies of articles and books need to be searched, which is very time consuming. After all, it is difficult to get copies of many of these papers that are simply not accessible online. However, the authors believe that the conclusions may be largely the same, if the search procedure was amended, so this may not be as serious limitation as one may think.

One of the inclusion criteria for the current systematic review was that the study should be based on non-fictional media portrayal of suicidal behaviours, as the influence of fictional suicide stories presented in the media has been found more controversial [19]. However, the research on the description of portrayals of suicide not just in the news, but also in the movies might be of great interest. To the extent that portrayals of suicide in the cinema are inaccurate, they can contribute to public misunderstandings of the nature of suicide and hurt the development of effective suicide prevention programs. Given that people spend more time watching movies than any other leisure time pursuit, it is important to develop and apply suicide guidelines for film. Suicides in film are generally insistent with media guidelines and, for example, give graphic details on suicide methods, depict the actual, often bloody, suicide act in progress, and de-emphasize mental illness risk factors in suicides [83].