Direct versus indirect psychosocial and behavioural interventions to prevent suicide and suicide attempts: a systematic review and meta-analysis

Summary Background Psychosocial and behavioural interventions that address suicidal thoughts and behaviour during treatment (direct interventions) might be more effective in preventing suicide and suicide attempts than indirect interventions that address symptoms associated with suicidal behaviour only (eg, hopelessness, depression, anxiety, quality of life). To test this hypothesis, we did a systematic review and meta-analysis of psychosocial and behavioural interventions aimed at preventing suicide and suicide attempts. Methods For this systematic review and meta-analysis, we searched MEDLINE and PsycINFO from inception to Dec 25, 2015, for randomised controlled trials that reported suicides or suicide attempts as an outcome, irrespective of participants' diagnoses or the publication language. We excluded studies with pharmacological or device-based interventions, those that targeted communities or clinicians, primary prevention trials, and trials that reported events of non-suicidal self-injury as suicide attempts. Trials that had no suicides or suicide attempts in both groups were also excluded. Data were extracted by one investigator and independently verified by a second investigator. We used random-effects models of the odds ratio (OR) based on a pooled measure of suicides and the number of individuals who attempted suicide, immediately post-treatment and at longer-term follow-up. Findings Of 2024 unique abstracts screened, 53 articles met eligibility criteria and reported on 44 studies; 31 studies provided post-treatment data with 6658 intervention group participants and 6711 control group participants at baseline, and 29 studies provided follow-up data. The post-treatment difference between direct interventions and indirect interventions did not reach statistical significance at the 0·05 level (OR 0·62 [95% CI 0·45–0·87] vs 0·93 [0·77–1·12], p=0·06) and represented a large effect size (Cohen's d =0·77). At longer-term follow-up, the difference was not significant (OR 0·65 [0·46–0·91] vs 0·82 [0·70–0·96], p=0·25) but still represented a medium effect size (Cohen's d =0·47). These effect sizes emphasise the clinical importance of direct interventions. Post-hoc subgroup and sensitivity analyses showed that our results are robust and unlikely to be notably affected by between-study heterogeneity or publication bias. Interpretation Psychosocial and behavioural interventions that directly address suicidal thoughts and behaviour are effective immediately post-treatment and long term, whereas treatments indirectly addressing these components are only effective long term. Moreover, although the differences shown between direct and indirect strategies were non-significant, the difference in favour of direct interventions represented a large post-treatment improvement and medium improvement at longer-term follow-up. On the basis of these findings, clinicians working with patients at risk of suicide should address suicidal thoughts and behaviours with the patient directly. Although direct interventions are effective, they are not sufficient, and additional efforts are needed to further reduce death by suicide and suicide attempts. Continued patient contact might be necessary to retain long-term effectiveness. Funding National Institute of Nursing Research.