Predicting suicide and suicide attempts in adults in acute hospitals: A systematic review of diagnostic accuracy evaluating risk scales

Abstract

Background: Prediction of suicide and suicide attempts is an aspirational goal for clinicians and policy makers. This requires scrutiny of various diagnostic accuracy measurements. Objective: To evaluate the diagnostic accuracy of suicide risk scales in inpatient and emergency departments, and combined settings within hospitals, and consider the best scale in predicting suicide and suicide attempts in adults. Design: Systematic review of diagnostic accuracy. Settings: Inpatient and emergency rooms of acute hospitals. Methods: The searched databases included PubMed, Embase, ProQuest, MEDLINE, Web of Science, CINAHL, Cochrane, PsycINFO and Scopus. Grey literature databases (ProQuest Dissertations and Theses, Google Scholar and OpenSIGLE) and reference lists of all included studies were also searched. Eligibility criteria included prospective cohort studies of suicide or attempted suicide in adults. A narrative synthesis was done for true positives, true negatives, false positives and false negatives, and they were calculated for sensitivity, specificity, likelihood ratios, predictive values and diagnostic odds ratios. Results: The included 33 studies differed in the methodological characteristics and scales used. The sensitivity of scales, specificity, positive predictive value, and diagnostic odds ratios ranged from 17 to 100%, 11–93%, 0.43–40.9%, and 0.77–18.5, respectively. Conclusion: The choice of scale for predicting suicide attempts in adults in hospitals depends on many factors. None of the scales evaluated produced accurate diagnostic measurements. The Manchester Self-Harm Rule and Suicide Crisis Syndrome should be used in the emergency department and inpatient setting, respectively, while the Suicide Intent Scale should be used in combined settings.