Predicting suicidal behaviours using clinical instruments: systematic review and meta-analysis of positive predictive values for risk scales.

Background Prediction of suicidal behaviour is an aspirational goal for clinicians and policy makers; with patients classified as 'high risk' to be preferentially allocated treatment. Clinical usefulness requires an adequate positive predictive value (PPV). Aims To identify studies of predictive instruments and to calculate PPV estimates for suicidal behaviours. Method A systematic review identified studies of predictive instruments. A series of meta-analyses produced pooled estimates of PPV for suicidal behaviours. Results For all scales combined, the pooled PPVs were: suicide 5.5% (95% CI 3.9-7.9%), self-harm 26.3% (95% CI 21.8-31.3%) and self-harm plus suicide 35.9% (95% CI 25.8-47.4%). Subanalyses on self-harm found pooled PPVs of 16.1% (95% CI 11.3-22.3%) for high-quality studies, 32.5% (95% CI 26.1-39.6%) for hospital-treated self-harm and 26.8% (95% CI 19.5-35.6%) for psychiatric in-patients. Conclusions No 'high-risk' classification was clinically useful. Prevalence imposes a ceiling on PPV. Treatment should reduce exposure to modifiable risk factors and offer effective interventions for selected subpopulations and unselected clinical populations.