|
Abstract
|
Suicide is a leading cause of death in the USA, and between 2000 and 2018, rates increased by 36% to 14·5 deaths per 100 000 people. Although this trend is pervasive, risk is not distributed evenly across the country or the population. 80% of deaths by suicide occur in men, and rates are highest among Native Americans and White Americans, with 29·4 suicide deaths per 100 000 occurring among White men aged 65 years and older. Conversely, suicide rates have decreased by 5% since 2018 among White individuals, but have continued to increase in other groups, including among young Black men. Beyond demographics, individual risk factors for suicide include violent victimisation, major depressive disorder, chronic pain, substance abuse, and financial stress. Situational and community-level factors such as unemployment rates and accessibility of lethal means, particularly firearms, can also increase the risk of suicide.
Individual-level models incorporating risk factors can classify people as being at high or low risk of death by suicide, but the predictive accuracy of these models is close to zero, making targeted prevention efforts challenging. If individual-level models alone cannot provide the information necessary to prevent death by suicide, population-level models might be crucial to identifying areas of concern and allocating resources appropriately. In The Lancet Public Health, Sasikiran Kandula and colleagues use geospatial modelling to estimate county-level suicide risk, generating accurate predictions that have the potential to inform future intervention.6
By taking into account spatial and temporal autocorrelation, this approach allows the identification of high-risk counties and identifies areas of focus for intervention. Suicide risk was estimated to increase with each SD increase in firearm ownership (2·8% [95% credible interval (CrI) 1·8 to 3·9]), prevalence of major depressive episode (1·0% [0·4 to 1·5]), and unemployment rate (2·8% [1·9 to 3·8]). Conversely, risk was estimated to decrease by 4·3% (-5·1 to -3·2) for each SD increase in median household income and by 4·3% (-5·8 to -2·5) for each SD increase in population density.
This useful analysis yields several important follow-up questions. First, a small set of covariates representing known risk factors for suicide were included in the models. Additional important contributors are well documented, and including these could improve model performance and usefulness. In particular, no county-level metric of education or access to mental health care was included in their approach. Additionally, state-level rates of major depressive disorder and firearm ownership were used to estimate county-level rates. Although this decision might reflect low data availability, both of these factors vary substantially within states, further hindering model performance and interpretation. Data availability is a pervasive issue and might represent one of the key obstacles in the optimisation and dissemination of population risk prediction models for suicide. Second, a large proportion of US counties are rural, with relatively small populations. In these counties, small changes in absolute numbers of deaths translate to large changes in death rates at the population level, making modelling particularly challenging, as we have previously described...
Language: en |