Associação entre mídias sociais com autolesão não suicida e pensamentos e comportamentos suicidas: uma umbrella review

Abstract

Suicidal behavior (STB) and non-suicidal self-injury (NSSI), usually aggregated under the term self-injurious thoughts and behaviors (SITB), are related conditions that constitute a serious global public health problem, disproportionately affecting young populations, with epidemiological trends of worsening over the last decade. This period coincides temporally with the massive incorporation of social media (SM). The simultaneity of these phenomena established an epidemiological alert, driving investigations into SM exposure as a potential factor associated with NSSI and STB. Objectives To map, synthesize, and critically assess the strength and quality of evidence regarding the association between different patterns and contexts of social media use with STB and NSSI. Methods An umbrella review was conducted following the Joanna Briggs Institute guidelines, with a broad and unrestricted search across seven databases. Systematic reviews (SR) evaluating any profile of participant or user exposed to social media in which the assessed outcome corresponded to the spectrum of NSSI or STB were included. Selection and data extraction were performed by independent pairs of reviewers. The adapted AMSTAR-2 was used for methodological quality analysis and the corrected covered area (CCA) for Data were presented according to thematic synthesis and implications for future practice, research, and policy. Results Sixteen SR met eligibility criteria; 15 comprised the final synthesis, aggregating 256 independent primary studies, predominantly including young people. Methodological quality was critically low for 81.25% and low for 18.75% of the reviews, with predominance of cross-sectional designs. Absence of prior protocol (10/16), non-comprehensive search strategy (14/16), lack of specification of the excluded studies list (10/16), and absence of 11 assessment of funding sources of primary studies were the main items that decreased evidence quality. Overall overlap was low (CCA = 1.79%). For specific exposures, overlap was low for cyberbullying (0%), but high to very high for social media usage time (13.5%); sexting (50%); problematic social media use (28%), and exposure to self-harm and suicide content (11%). Most reviews (75%) included studies with young people aged 12 to 25 years. Meta-analytic evidence from cross-sectional studies demonstrated associations between some variables and different outcomes across the NSSI and STB spectrum. Problematic ) was associated with suicidal ideation (OR = 2.81, 95% CI [1.72, 4.59]; k = 4). Cybervictimization was associated with all constructs of the SITB spectrum, with a smaller effect size for suicidal ideation (OR = 2.93; 95% CI [2.43–3.54]; k = 45) and larger for non-suicidal self-injury (OR = 4.36; 95% CI [2.32–8.20]; k = 3). Sexting was associated with non-suicidal self-injury (OR = 3.07; 95% CI [2.53–3.74]; k = 2) and suicide attempt (OR = 4.24; 95% CI [3.13–5.44]; k = 1), but not with suicidal ideation (OR = 2.37; 95% CI [0.98–5.73]; k = 1). Exposure to SITB content on social media was asociated with all constructs of the SITB spectrum, with a smaller effect size for suicidal ideation (OR = 2.12; 95% CI [1.31–3.43]; k = 3) and larger for non-suicidal self-injury (OR = 3.13; 95% CI [1.41–6.96]; k = 1). Social media usage time showed no significant association with NSSI, suicidal ideation, or suicide plans. Meta-analyzed longitudinal data are available only for cybervictimization, with associations with self-injury (OR = 2.23, 95% CI [1.43–3.47]; k = 3) and suicidal behavior (OR = 1.31, 95% CI [1.11–1.55]; k = 6). Participation in online forums to discuss SITB content showed a coexistence of associated benefits and risks. Greater sense of social connection, use for emotional regulation, and perceived reduction in behavior were benefits cited by users. On the other hand, behavioral activation, presence of behavioral triggers, social comparison, competitiveness for more extreme behaviors, normalization of behavior, formation of a symptom-based identity, social learning and contagion effect, subversion of support, caregiver burden, and vicarious trauma were described as associated harms. Analysis of primary studies points to factors that possibly increase the association, particularly female sex, previous depressive symptoms, interpersonal conflicts or low offline support, younger age, sleep impairment, and increased usage time throughout adolescence. Studies were predominantly cross-sectional. Several methodological factors appear to modify the association, particularly self-report, which may attenuate the effects of usage time but underestimate the association among individuals experiencing psychological distress. Additionally, online recruitment appears to overestimate the effect in quantitative data, but not when assessing the subjective experiences of users of online forums and social networks. The association tends to increase when studies do not control for confounding variables such as baseline psychopathology, depression, or offline bullying. Conclusions The association found for specific exposure variables, to the detriment of usage time, evidences the need to move beyond the use of broad constructs in clinical practice, regulation, and research, prioritizing patterns, contexts, and content accessed. Critical analysis of the evidence points to limits for its extrapolation, but the convergence of findings justifies contextualized preventive strategies, such as protection against cyberbullying and sexting practices, against mechanisms that reduce control over use, and in favor of sensitive content moderation. The data support important considerations for new research in the field. Regarding new reviews, the use of specific exposure and outcome variables and attention to methodological quality may allow better extrapolation of results. For primary studies, particular attention is recommended regarding control for baseline comorbidities, recruitment biases, and measurement validity, as well as the adoption of longitudinal designs capable of supporting causal inference about the relationship between SM and SITB. Despite the low methodological quality, the possible risks of bias noted, and the predominance of cross-sectional studies that do not allow causal inferences, some online phenomena were consistently related to one of the outcomes with greatest lethality within psychiatry. Therefore, given the impact of this condition, it may be unfeasible to await better available evidence for regulatory decisions, at the risk of penalizing an entire generation of young people exposed to a variable never before studied in science.