Intimate Partner Violence as a Precipitating Factor For Suicide: Strengthening Data Systems and Identifying Opportunities For Prevention

Background: The link between intimate partner violence (IPV) and suicide outcomes remains largely unexplored given existing limitations with quantitative data fields in national data sources. To address this gap, the aims of this dissertation were to 1) assess the feasibility of developing an automated tool for detecting textual mention of IPV circumstances for suicide deaths in existing mortality data, 2a) document the prevalence of IPV-related suicides in the US, and 2b) examine correlates for IPV-related suicide to inform prevention efforts. Methods: Using n=10,000 hand-reviewed cases from the National Violent Death Reporting System (NVDRS), we developed and evaluated a supervised machine learning (SML) model. Information was extracted from death narrative text in NVDRS using natural language processing. After finalizing the SML tool, we applied it to help examine the prevalence of IPV-related suicide in a dataset of n=130,550 deaths from NVDRS (2015-2019). Finally, we examined the correlates for IPV-related suicide (compared to other suicide events) using logistic regression analyses stratified by gender and incident type. Results: We developed a tool that could detect IPV-related suicide with robust precision (0.72), sensitivity (0.70), specificity (0.98), and Coen’s Kappa values (0.69). Using this tool and other methods, we determined that 7.1% or more of suicides in the US are precipitated by IPV. We observed higher odds of IPV circumstances when the person who died by suicide was involved in the justice system and when a firearm was used as the suicide method. Compared to referent group single suicides, there were lower odds of IPV circumstances when traditional suicide risk factors such as physical or mental health problems were recorded. Conclusion: IPV has been overlooked in suicide research. Our SML tool can now be used to inform future research and surveillance efforts focused on IPV-related suicide. We found that IPV circumstances may represent a distinct pathway to suicide compared to other commonly understood mental health problems or traditional suicide “warning signs.” This has important implications for prevention, including collocating or integrating IPV and suicide intervention efforts.