Article Characteristics.

A large number of research studies have previously assessed adherence to suicide-reporting guidelines in media from print, television, and radio outlets (14, 16, 21, 24, 25). Furthermore, previous research has demonstrated that media content which poorly adheres to the suicide-reporting guidelines causes objective health harms, such as increases in deaths from suicide clusters and increased suicide rates (14, 16). In light of these compelling health consequences, the evolving nature in which individuals consume media, and persistent limited uptake of safe-reporting practices, our study sought to assess adherence among articles shared on social media, a leading way in which individuals now view and interact with news information. Consistent with prior studies, we found that publicizing the decedent’s name, describing the method and location of the death in detail, and featuring the word suicide prominently in the headline were present in a large proportion of articles (16, 21). Similarly, inclusion of protective information was low. For example, only 16% of articles included information about a suicide prevention telephone hotline despite the general availability of such resources in the nations we studied and the relative ease with which such information can be included. The reasons for noninclusion of protective elements in media reporting about suicide have not been fully elucidated in prior research; however, there are limited incentives for publishers to include additional content, no matter how brief it may be. Indeed, as over 60% of the articles we studied included none of the protective elements we measured, increasing inclusion of protective content is an area of safe reporting where considerable gains can be made.

Notable differences in article scores by article type were found for each category. Articles about homicide-suicides had the lowest mean score while general articles about suicide had the highest score. Particularly low scores among articles about homicide-suicides may be a reflection of the fact that such multiple-fatality events are common targets of sensational reporting. Furthermore, even where publishers are aware of best practices for safe reporting about suicide, publishers may not view homicide-suicide events as fitting within the paradigm of suicide reporting and hence fail to include standard protective elements. Although studies on the effect of suicide contagion from media reporting on homicide-suicides has been mixed (13, 14, 16), there may be increased risks for suicide among females (13), suggesting caution is still warranted for this class of articles.

General articles about suicide (which do not focus on a specific suicide) include a broad range of articles, such as positive stories of individuals living with mental health challenges or news on suicide prevention programs, had the highest scores. Prior studies have found that media covering individuals with suicidal ideation and advancing the notion that positive coping skills are possible have been empirically associated with lower population-level suicide rates (14).

Importantly, there were notable differences across countries. Articles from Australia had the highest scores of all nations studied, which is consistent with concerted efforts by suicide prevention organizations in Australia to increase uptake of suicide-reporting guidelines (34). Articles from the United States had the lowest mean score of all high-income, English-speaking nations studied and were statistically indistinguishable from articles from India. Notably, there has been a growing amount of research from economically developing nations, particularly India, which have documented low adherence to suicide-reporting guidelines; our results are consistent with these studies (22, 35).

Reader Engagement.

In contrast to commonly held beliefs, our empirical findings from one of the largest social networking platforms indicate that, at least for mental health content, articles with lower amounts of sensationalism and higher quality health reporting are more often reshared and have the potential to reach a greater number of viewers. Even a single 1-point increase in positive reporting on a 27-point scale confers a 19% increase in the odds of an article being reshared. Given that the overall degree of adherence to safe-reporting practices is still relatively low across the entire corpus of suicide-related news, it is likely that robust efforts to improve safe-reporting practices would likely shift safe-reporting scores by more than a single point increase, resulting in even greater odds of article resharing. Indeed, the top fifth of articles by score received ∼470% as many reshares as articles with scores in the bottom fifth. Creating even a small shift in population-level exposure to harmful content, such as is possible given the extremely large user base of leading social media platforms, would be anticipated to lead to a population-level impact on suicide reduction. Indeed, leading public health strategies that create seemingly small shifts in the distribution of exposure to risk factors, result in sizable population-level reductions in death (36).

Lastly, with regards to the engagement analysis, it should be noted that higher scoring articles also had increased odds of receiving a love reaction and lower numbers of “anger” reactions. Although previous research indicates that the love reaction is used in a largely monosemantic positive way (37), the nature of anger reactions toward articles is less clear as readers may simply be expressing anger at the death of an individual rather than at the content of an article. Importantly, however, previous research has identified anger as a strong motivating factor for sharing of information by individuals (38), yet our findings diverge from this as articles with the highest safe-reporting scores and resharing had the lowest odds of anger reactions. This suggests that the expression of anger in relation to suicide-related news and resultant social actions are a nuanced area that merits further research.

Publisher Characteristics.

Investigation into the characteristics of publishers more likely to have higher safe-reporting scores revealed that national publishers had higher mean scores than local ones. The nature of outreach and training of news organizations on best practices for suicide reporting in the United States has not been quantified. With recent renewed national awareness of the importance of local media and calls to strengthen local news organizations (39), our findings suggest an increased role for capacity building for suicide reporting among local news organizations. Our analysis also detected a trend toward higher scores from publishers that are primarily print media organizations compared to outlets that are primarily television or radio; this is consistent with some studies (13), yet other studies have found lower adherence among print media (40). It is important to acknowledge that the definitional boundaries of print, television, and radio now poorly define media organizations in online environments. For example, it is not uncommon for a television news station to have a robust online news site with large volumes of digital print content. In regression analyses, there were no statistically significant differences among legacy and digital-native media outlets. This finding may be a result of the fact that these distinctions are also increasingly artificial as nearly all legacy publishers are also building robust digital publishing platforms given changes in media consumption preferences.

Limitations.

There are important limitations of our study to note. First, our study assessed media content about suicide over a focused time period. Ongoing evaluation of safe-reporting practices should be performed. It should also be noted that the articles studied in this analysis were those that appeared most frequently on Facebook. We focused on such articles as content that is more prevalent is most likely to impact population health and our sample of articles remained remarkably diverse (representing 307 unique publishers across a wide variety of nations and characteristics); however, assessing the nature of news media that is relatively rarer is an important direction for future work. Second, as with any study involving manual annotation, there is the possibility of misclassification of suicide-reporting elements. However, any misclassification is unlikely to be differential across articles and with the large number of elements evaluated per article the effect of occasional misclassification on overall score distribution is small. Also, given that we sampled articles based on the presence of the word suicide in the article text or user annotation, it is possible that we excluded some articles about suicide that did not explicitly use the term; the number of such articles is likely to be low and this is a limitation of other published research as well. Third, while our study is the only assessment of reporting about suicide on Facebook and has significant global coverage, more can be learned by expanding the international reach of our evaluation (41, 42). Lastly, our evaluation treated all elements of safe reporting as equally important in the creation of our scores, which is consistent with prior published studies; however, prior research also suggests that some elements may be more contagion promoting than others. Future work may consider incorporating weighting schemes to allow more refined study of the most harmful elements.

Implications.

The public health implications of our findings are considerable and can help advance better reporting on suicide and therefore a potential reduction in suicide deaths from contagion. Our results have relevance for journalists/publishers, public health organizations, social media companies, academic researchers, and lay readers of health content. First, for publishers, while poor adherence to suicide-reporting guidelines is recognized by health professionals as harmful to the health of individuals, our results provide empirical evidence that such reporting may also be harmful to publishers as well. The observation that news articles with lower adherence to suicide-reporting guidelines are less likely to be shared by audiences provides strong rationale for publishers to adopt greater adherence to the voluntary guidelines.

Second, our findings help to accelerate the efforts of public health organizations. Given the rapid and sustained increases in suicide rates nationally, robust multisector partnerships to address suicide are needed. Suicide-reporting guidelines, which were primarily developed in the era of print media, would benefit from continued adaptations relevant to the increasingly digital nature of communications. Public health professionals and relevant nongovernmental partners can also serve as key entities to help advance the creation of positive content about individuals overcoming mental health challenges that can lower suicide rates (14). Moreover, our study found that news articles infrequently had information on national and local resources, warning signs of suicide, and statistical information and quotes from official sources, all of which are items which health organizations can help strengthen.

Third, social media platforms can play an important role in disseminating information about the importance and nature of safe reporting to the many publishers active on their platforms in a way that may have even greater reach than previous efforts by health organizations, particularly as the number of publishers increase and diversify with the advance of digital publishing. Furthermore, beyond the stated goals of social media organizations to protect against health misinformation, social media platforms can also play an important role in advancing content which protects against suicidal behaviors, such as the positive stories noted above; greater visibility of such content could have a significant benefit to public health.

Fourth, this research points to key areas of collaboration for academic researchers, particularly in measuring the impact and effectiveness of multisector approaches to increase uptake of the guidelines. Further research is also needed to expand the literature on engagement with articles beyond our initial presentation herein and to more fully understand who is most vulnerable to harmful content. While most of the research literature has focused on suicide fatalities as a consequence of harmful reporting, less is known about nonfatal self-harm including nonsuicidal self-harm. Additionally, with regards to public health strategies focused on increasing positive content, while previous research on the Facebook platform has indicated that increased viewing of positive content is associated with increased positive sentiment in subsequent posts made by viewers (43), no study has assessed whether a similar relationship holds true for positive suicide-related content and additional research on the nature and characteristics of positive content online that is most protective and beneficial to mental health is needed.

Lastly, our findings suggest a final important area for expanded engagement—with lay readers of health content. Awareness of the impact of sensationalized content about suicide on health may be even lower among the general population than among publishers. Friends and family in the social networks of individuals at risk for suicide can play an important role in protecting the health of loved ones through a greater awareness of the impact of sharing content about suicide. Furthermore, with the advance of social technologies, all individuals are increasingly empowered to be content creators and thus expanded efforts to educate all individuals on suicide prevention strategies are increasingly important. Broad collaborations to improve safe reporting about suicide stand to benefit both population health as well as news media goals of reach and engagement.