Injuries/Violence
Predicting US suicide death rates over time using measures of structural inequity Jonathan Platt* Jonathan Platt Amanda Sursely Avinash Mudireddy
Introduction: Suicide is the leading cause of violent death in the US, and death rates have been increasing especially among historically marginalized groups. Despite congressional calls, research focusing on social inequities and suicide is limited. This project aimed to identify key social inequities to accurately predict suicide death rates, utilizing an intersectional framework to account for multidimensional social structures, and machine learning methods to model that complexity with statistical efficiency.
Methods: US Vital Statistics data were used to calculate county-level suicide rates from 2005-2020 (n=5603). We assembled a dataset of 36 indicators of structural inequities across racial, sex/gender, and class domains. For prediction, we specified a time-series regression model with a Convolutional Neural Network (CNN) layer, Gated Recurrent Unit (GRU) layers, and dense layers. The integrated CNN-GRU structure utilizes a hybrid loss function and applies regularization techniques to enhance the model’s ability to capture local patterns and temporal ordering.
Results: Model fit and error rates suggest good prediction accuracy (Mean Absolute Error=3.95, Root Mean Squared Error=5.54, R-squared=0.77; see figure). The final model contained 10 inequity indicators; structural classism indicators included income inequality and receipt of public assistance, while structural sexism indicators included sex disparities in education attainment, employment, poverty, political representation, and state laws restricting domestic violence perpetrators from possessing firearms. No indicators of structural racism were retained.
Discussion: An effective population health approach to suicide prevention must account for both historical and emerging inequities to identify high risk groups and prevent disparities. The indicators of structural classism and sexism we have identified highlight key targets for equity-promoting interventions to reduce preventable suicide deaths.
