Using a spatiotemporal model to estimate the impact of suicide prevention in small areas

Evaluating the impact of suicide prevention programs is more challenging in smaller counties, where reliable estimates of suicide rates are difficult to obtain. This includes rural and tribal communities disproportionately burdened by suicide. We describe a novel application of a spatiotemporal model developed for disease mapping to assess the impact of suicide prevention in small areas. As an example, we use small counties exposed to the Garrett Lee Smith (GLS) program. Specifically, the impact of suicide prevention programming on suicide-related hospital use among youth between 2008 and 2018 was explored with this novel method in a sample of rural counties across 10 states. While, on average, suicide-related hospital use was close to what would be expected in the absence of the program, there was considerable variation across counties. For example, among a group of counties in South Dakota, there was a substantial decrease in suicide-related hospital use among youth for several years after the start of exposure. Subsequent exploration suggested that implementing a high proportion of longer trainings was associated with membership in this favorably impacted group. The illustrated approach can be used to estimate the impact of interventions targeting other outcomes that are relevant but relatively infrequent.