Intimate partner violence (IPV) can be lethal, although the total contribution of IPV to fatal violence in the US remains unknown. Researchers often study intimate partner homicide (IPH) but may overlook other IPV-related deaths. Family, children, or new dating partners can be killed in corollary homicides. A perpetrator of IPV may be killed as a result of law enforcement response to IPV (ie, legal intervention) or may die by suicide after committing homicide (ie, homicide-suicide). Additionally, IPV could contribute to suicide in the absence of other fatalities (ie, single suicide). Emerging research suggests that single suicides were more common among perpetrators of IPV than among survivors of IPV, but single suicides occurred in both groups. We sought to provide the most complete assessment available of IPV’s contribution to the burden of violent deaths in the US.

The University of North Carolina at Chapel Hill determined that this cross-sectional study was not human participant research and so was exempt from ethical review and informed consent. Our analytical process adhered to the STROBE reporting guideline.

Most US states collect data for the National Violent Death Reporting System (NVDRS). Information from death certificates and investigation reports is recorded using close-ended fields and textual summaries called death narratives. We examined intentional violent deaths in NVDRS (2015-2019) with known circumstances and available biological sex data (Figure).

In NVDRS, IPV circumstances (yes or no) are systematically recorded for homicides and legal intervention deaths but not suicides. This lack of documentation constitutes a barrier for research. To address this limitation, we used a multipronged approach to record IPV circumstances (Figure). We first verified IPH and corollary deaths using the NVDRS IPV and victim-suspect relationship variables. Next, we linked deaths by incident identifier so that IPV circumstances applied to all deaths from the same event (eg, homicide-suicides and multiple homicides). For remaining suicides, we used novel methods, including incorporating data from prior hand-reviewed NVDRS death narratives and using a validated supervised machine learning tool. This tool uses a dictionary approach with natural language processing to extract keywords and phrases from death narrative text. Vectorized keywords and phrases are then fed into a random forest model that labels IPV circumstances for each suicide (yes or no) with robust sensitivity (0.70) and specificity (0.98).

Among 178 940 intentional violent deaths (139 143 deaths among males and 39 797 deaths among females), 9.5% had IPV circumstances. The most common IPV-related fatality was single suicide (most common among males) (Table), followed by IPH (most common among females). Among deaths with IPV circumstances, 64.5% involved a firearm. We identified 55.2% of IPV-related fatalities using conventional methods and 44.8% using novel methods.

This cross-sectional study found that IPV contributed to more violent deaths in the US than previously reported; nearly 1 in 10 violent deaths had IPV circumstances. While IPV was associated with sex, findings underscore the need to engage males and females to prevent IPV while addressing shared underlying risk factors associated with IPV, homicide, and suicide. Most IPV-related fatalities involved a firearm, highlighting the importance of implementing and enforcing firearm restrictions for perpetrators of IPV. More resources may also be needed to help survivors of IPV and their families interrupt IPV before it can contribute to a fatality.

Our study has some limitations. IPV would be recorded in NVDRS only if death investigators learned about IPV and felt it was a salient factor associated with the death. In particular, coercive control or more historical IPV may be undercounted.² Thus, our findings likely underestimate the contribution of IPV to violent fatalities. NVDRS data are also not yet nationally representative, and there were limited details about prior IPV experiences (ie, surviving vs perpetrating IPV) among individuals who died. Future research should examine how different experiences or roles in an abusive relationship may be differentially associated with risk for violent death in living populations.