In Reply Our systematic review1 sought to provide a transparent and objective overview on the state of the science of suicide prediction models (SPMs). The findings demonstrate that current SPMs targeting suicide mortality cannot overcome the statistical challenge of predicting this relatively rare event. Our review focused on: (1) the positive predictive value (PPV), which is the probability that any individual flagged by the SPM will actually die by suicide, and (2) sensitivity, which is the proportion of all suicide mortalities detected. The PPV of suicide mortality was extremely low (≤0.01) across military, the Veterans Administration, and civilian health systems engendering very large cohorts of patients falsely flagged as being at risk for suicide. The model sensitivities indicated that a substantial proportion of suicides (25%-50%) will still be missed. Based on these predictive model properties, the lack of evidence on the downstream benefits and harms of SPMs,2,3 and the minimal advancements in effective treatments to prevent suicide mortality,4 it is unlikely that SPMs will move the needle on current guidelines that do not recommend population-level interventions for suicide.5
Belsher BE, Smolenski DJ, Pruitt LD. Positive Predictive Values and Potential Success of Suicide Prediction Models—Reply. JAMA Psychiatry. Published online June 26, 201976(8):870–871. doi:10.1001/jamapsychiatry.2019.1510
Browse and subscribe to JAMA Network podcasts!
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: