Abstract
Intimate partner homicide (IPH) is a leading cause of maternal mortality in the United States. However, very little information exists as to the circumstantial factors associated with IPH during pregnancy. We conducted a descriptive study of the demographic characteristics, psychosocial service engagement, and crises experiences (i.e., life and relationship stressors) among pregnant and nonpregnant victims to understand what differences, if any, exist in their risk profile for IPH. Data from the Centers for Disease Control and Prevention’s National Violent Death Reporting System (NVDRS) were used for this study. The NVDRS is a national opt-in tracking system of all violent deaths in the United States. Pregnant victims (N = 293) were significantly more likely to be 5 years younger than nonpregnant victims, African American, and never married. Pregnant victims were more likely to be seen in the emergency room following the fatal incident. Nonpregnant victims (N = 2,089) were significantly more likely to have suspected alcohol use at the time of their death. In strictly proportional terms, we also observed higher rates of mental health problems, a history of mental health treatment, and a reported history of intimate partner violence (IPV), crisis, or family problems among nonpregnant victims. A wider range of IPH-related risk factors (e.g. substance abuse) need to be included IPV assessments. Future studies should seek to develop effective interventions to prevent IPH, particularly among reproductive aged women.
Introduction
Intimate partner homicide (IPH) represents a significant public health threat that disproportionately affects women in the United States (Campbell et al., 2007; Stöckl et al., 2013). Research has shown that women are more likely to be killed by an intimate partner than any other type of perpetrator (Stöckl et al., 2013) and that IPH is one of the leading causes of premature death for women under the age of 44 years (Campbell et al., 2007). In fact, when compared with IPH rates for male victims (3%–6%; Stöckl et al., 2013), 30% to 60% of homicides involving female victims are perpetrated by an intimate partner (Campbell et al., 2007; Catalano et al., 2009; Cooper & Smith, 2012; Garcia et al., 2007; Petrosky et al., 2017). Thus, even though men are more likely to be murdered than women, the proportion of women killed by a partner is nearly 6 times greater than that of men, and women represent nearly 70% of all victims killed by an intimate partner (Catalano et al., 2009; Cooper & Smith, 2012; Stöckl et al., 2013). More alarmingly, a recent study by Fox and Fridel (2017), which reviewed national homicide data from the Federal Bureau of Investigation, found that rates of IPH of women have risen since 2014.
Studies on risk factors associated with IPH are limited but have shown that victims often have a history of intimate partner violence (IPV), including severe physical abuse, sexual abuse, stalking and isolation, more frequent abuse, and threats of homicide (Aldridge & Browne, 2003; Campbell et al., 2003, 2007; Messing et al., 2017; Moracco et al., 1998; Nicolaidis et al., 2003; Saunders & Browne, 2000). Research also shows that the risk of IPH or attempted IPH is heightened when there is a change in the relationship status between the victim and her perpetrator (e.g., an attempt to leave; Garcia et al., 2007). In addition, a study by Wadman and Muelleman (1999) found that approximately 40% of victims of IPH visit an emergency department in the 2 years prior to the fatal incident for IPV-related injuries (Wadman & Muelleman, 1999). Likewise, Sharps et al. (2001) found that around 44% of victims saw a health care provider for behavioral services within a year prior to their death (Parsons & Harper, 1999; Sharps et al., 2001). Furthermore, qualitative research among survivors of attempted IPH has found that many victims do not perceive the perpetrator as a threat to their life (Nicolaidis et al., 2003) and thus may not recognize their partners’ behaviors as potentially lethal. Identifying the full range of risk factors for IPH is paramount then to improving screening in clinical settings and reducing IPV-related mortality.
Pregnancy or the unique contextual factors associated with risk of IPH during pregnancy is one area that has not been fully explored. However, there is support in the literature to suggest that pregnant women may be at an increased risk of IPH (Martin et al., 2007). Research on IPV has shown that women of reproductive age are the most at risk of abuse, and pregnancy increases that risk (Campbell, 1998; Catalano, 2012). Furthermore, there is evidence to suggest that abuse intensifies during pregnancy, meaning pregnant women are more likely to experience more severe forms of abuse, including IPH (Shadigian & Bauer, 2005). To that end, homicide is now considered one of the leading causes of maternal mortality or pregnancy-related deaths (Campbell et al., 2003; Horon & Cheng, 2001; Shadigian & Bauer, 2005; Taillieu & Brownridge, 2010). In addition, attempted homicide and other forms of abuse during pregnancy have adverse consequences that can reduce women’s engagement in prenatal care (Cha & Masho, 2014; Chambliss, 2008; Subramanian et al., 2012) and increase their susceptibility to behavioral health issues such as substance use (Anderson et al., 2002; Caetano et al., 2005; El-Bassel et al., 2005) and depression (Connelly et al., 2013). Such issues can affect maternal health (El Kady et al., 2005; Shah & Shah, 2010; Yost et al., 2005) and, in turn, contribute to greater mortality risks for both mothers and their children (Ackerson & Subramanian, 2009; Alhusen et al., 2015; El Kady et al., 2005; Shadigian & Bauer, 2005). Despite understanding the risks associated with pregnancy and the impact that IPV has on maternal mortality, very little information exists as to the contextual, or circumstantial, factors associated with IPH during pregnancy.
We conducted a descriptive study of the demographic characteristics, psychosocial service engagement and crises experiences (i.e., life and relationship stressors) among pregnant and nonpregnant victims using the Centers for Disease Control and Prevention’s (CDC, 2019) National Violent Death Reporting System (NVDRS) to understand what differences, if any, exist in pregnant and nonpregnant victims’ risk profile for IPH. By comparing pregnant and nonpregnant victims’ risk profile for IPH, we hope to inform and improve strategies for addressing and preventing IPH for pregnant women and reduce the risk of maternal mortality associated with IPV. Furthermore, we hope to shed light on the precipitating factors associated with IPH for women of reproductive age.
Method
We analyzed data from the NVDRS from the years of 2003–2015. Created in 2002, the NVDRS is a national opt-in tracking system of all violent deaths in the United States and its territories (CDC, 2018). In 2003, the first year of NVDRS data collection, seven states (Alaska, Maryland, Massachusetts, New Jersey, Oregon, South Carolina, and Virginia) contributed data. Six additional states joined the reporting system in 2004 (Colorado, Georgia, North Carolina, Oklahoma, Rhode Island, and Wisconsin), four additional states joined in 2005 (California, Kentucky, New Mexico, and Utah), and two joined the NVDRS in 2010 (Ohio and Michigan). In 2015, an additional 14 states started collecting data for the NVDRS (Arizona, Connecticut, Hawaii, Iowa, Indiana, Illinois, Kansas, Maine, Minnesota, New Hampshire, New York, Pennsylvania, Vermont, and Washington) for a total of 33 states by 2015. Due to the inconsistency of reporting among states and the limited number of states included in this analysis, our data are, at best, representative of the states collecting data in each year and likely underrepresent the true incidence of the study’s measures.
As of 2018, all 50 U.S. states and territories became official participants in the NVDRS (CDC, 2018), which includes more than 600 variables for included homicides, suicides, legal intervention deaths, terrorism-related deaths, unintentional firearm deaths, unintentional/accidental poisoning deaths, and deaths of undetermined intent (CDC, 2015). Variables are extracted from death certificates, coroner/medical examiner reports, toxicology reports, and law enforcement records and merged with qualitative reports from a data abstractor (CDC, 2015). Data address physical and mental health status, relationship issues, crises and life stressors, psychosocial service engagement, and detailed demographic factors regarding deceased victims and perpetrators (CDC, 2018).
The full NVDRS data set for years 2003–2015 includes 234,612 records. Female victims of IPH constitute 13,310 of these records. We chose to restrict our analyses to only those female victims of IPH who were of reproductive age, 11–45 years, which numbered 7,799 records. The NVDRS pregnancy status variable is coded into eight categories: not pregnant, pregnant at death, not pregnant but pregnant ≤42 days ago, not pregnant but pregnant 43 days to 1 year ago, not pregnant time unknown, pregnant unspecified, not applicable, or unknown. Of the 7,799 females, 2,382 had known data about pregnancy status. Thus, our analysis was carried out on this group of 2,382 females. It should be noted that an additional 5,417 female victims of IPH aged 11–45 years also were part of the NVDRS data set; however, as their pregnancy status was listed as unknown (N = 5,103), not applicable 1 (N = 129), or unspecified (N = 29), we did not include them in the primary analysis.
We classified as pregnant those coded as pregnant at death (combined N = 293). We classified as not pregnant those coded as not pregnant, not pregnant but pregnant ≤42 days ago, not pregnant but pregnant 43 days to 1 year ago, or not pregnant time unknown (combined N = 2,089). Two of the pregnancy status categories represent the postpartum period: not pregnant but pregnant ≤42 days ago (N = 21) and not pregnant but pregnant 43 days to 1 year ago (N = 83). Because of the small size of this postpartum subgroup, we did not treat them as a separate group in the analyses, but we did consider how results were affected by alternately including them as part of the pregnant subgroup rather than the not pregnant subgroup. This is discussed further in the results and limitations section.
Service engagement data were not available for all incidents included in the analyses; thus, when available, trends in service engagement were analyzed. To this end, we culled variables in the victim and circumstances sections of the NVDRS data set. SAS software version 9.3 (Cary, NC) was used to manage and analyze the data. Age by pregnancy status is presented as mean ± standard deviation and range of values; age by pregnancy status was compared using the t-test. Frequency distributions of categorical variables were compared by pregnancy status using the chi-square test. This study was approved by the University of Pittsburgh Committee for Oversight of Research and Clinical Training Involving Decedents (CORID) to ensure compliance with university ethical procedures and standards.
Results
Sample Characteristics
Table 1 shows the sample characteristics (age, race, marital status, and education level) by pregnancy status. The two-tailed t-test (t = 9.21, df = 2,380, p < .0001) indicated that pregnant victims (M = 25.3, SD = 6.4) were on average 5 years younger than nonpregnant victims (M = 30.3, SD = 9.1, Hedges’s g = 0.57). Data for race/ethnicity, marital status, and education level were only available for a portion of the sample. Nonpregnant victims (n = 2,088) were primarily White (49%), followed by Black/African American (34%) and Hispanic (11%). Among those with available marital status (n = 2,069), most had never been married (52%). For those with available educational information (n = 1,535), a little more than one third (39%) of nonpregnant victims had a high school/GED diploma, whereas around one quarter (24%) had between a 9th and 12th grade education. Pregnant victims (n = 293) were primarily Black/African American (44%), followed by White (39%) and Hispanic (12%). Among those with available marital status (n = 287), a majority had never been married (69%). Finally, among those with available educational information (n = 174), a majority had a high school/GED diploma (40%) or between a 9th and 12th grade education (30%). We found significant differences in the distribution of race and marital status between nonpregnant and pregnant victims. No statistically significant differences between the groups were found with respect to the distribution of education levels.
Demographic Variables by Pregnancy Status of Victims (Postpartum Grouped With Not Pregnant).
Note. GED = General Educational Development.
Significant values are in bolded text.
Circumstantial Variables
Table 2 shows the circumstantial variables by pregnancy status. Around one third (32%) of nonpregnant victims were killed during an argument (n = 672), one quarter of those with known data had been seen in the emergency department, 2 and one fifth of those with known data had suspected alcohol use. In addition, one fifth (21%) were killed as the result of another serious crime. 3 Overall rates of reporting for other variables were low, and differences between groups were not found to be statistically significant. In strictly proportional terms, we observed higher rates of mental health problems, current and past treatment for mental health, reported interpersonal violence and crisis in the past month, and problems with a family member among nonpregnant victims when compared to pregnant victims.
Circumstance/Engagement Variables by Pregnancy Status of Victims (Postpartum Grouped With Not Pregnant).
Note. Significant values are in bolded text.
Findings for pregnant victims show that around one third (32%) of those with known data were seen in the emergency department and (29%) were killed during an argument (n = 84), whereas one quarter (25%) were killed as the result of another serious crime (n = 74). Again, reporting for all other variables was low.
We found that pregnant victims were more likely to have been seen in the emergency department than nonpregnant victims (32% vs. 24%, respectively; p = .05). We found that nonpregnant victims when compared to pregnant victims were more likely to have suspected alcohol use (21% vs. 11%, respectively, p = .0002).
We observed very similar relationships in terms of frequency distributions and tests of significance between the nonpregnant and pregnant victims even when we grouped the postpartum females with the pregnant victims.
Discussion
Using the NVDRS, we conducted a descriptive study of the demographic characteristics, psychosocial service engagement, and crises experiences (i.e., life and relationship stressors) among pregnant and nonpregnant victims. We found several significant differences between pregnant and nonpregnant victims, namely that of age, race, marital status, emergency room visit, and use of alcohol at time of death. Although we did not find significant differences between pregnant and nonpregnant IPH victims in other observed NVDRS measures, the frequency distributions nonetheless have implications for improving screening for IPH during pregnancy, in particular, and for reducing IPV-related morbidity and mortality for women of reproductive age in general.
We found that pregnant IPH victims were significantly younger than nonpregnant victims and that a higher proportion of pregnant victims were African American and unmarried. These findings are in keeping with previous literature, which has shown that young, unmarried women and minority women are disproportionately represented among IPH victims (Bailey, 2010; Chang et al., 2005; Petrosky et al., 2017; Sugg, 2015). Thus, our study supports previous research on the elevated risk of IPH among young, minority women and suggests that prevention and intervention methods should be appropriately tailored to meet the needs of this particularly vulnerable population (Lee et al., 2002).
We also found that pregnant victims were significantly more likely to been seen in the emergency room following the fatal event. Although it is unclear why this may be the case (e.g., victims were already deceased upon their arrival at the emergency room, complications regarding the pregnancy), it nonetheless is in keeping with studies that have shown that IPV has a significant relationship with antenatal engagement with clinical care (Davidov et al., 2017; Helton et al., 1987; Hillard, 1985; Lipsky et al., 2004). For example, pregnant victims are more likely than nonpregnant victims to be hospitalized for injuries or IPV-related concerns (Cokkinides et al., 1999; Lipsky et al., 2004; Mendez-Figueroa et al., 2013) as well as often present in emergency rooms or other clinical settings with pregnancy-related complications (Greenberg et al., 1997; Leone et al., 2010; Salazar-Pousada et al., 2012). Despite the relatively higher contact with clinical professionals, IPV assessment in clinical settings, including emergency rooms, among pregnant and nonpregnant victims remains low (Bailey, 2010; Glass et al., 2001; Kramer et al., 2004), and providers report a number of barriers to addressing IPV in clinical contexts, including a general unwillingness, personal discomfort, lack of knowledge of how to help if IPV is disclosed, and belief that patients will not disclose and time, among other factors (Bailey, 2010). Furthermore, even when providers do ask about IPV, many screening tools separate out the time periods before and at the start of pregnancy or use a single “catchall” question that does not provide for comprehensive assessment of IPV experiences. Thus, providers often only ask about physical abuses, which may fail to provide a more comprehensive and longitudinal view of abuse (Bailey, 2010; Charles & Perreira, 2007; Taillieu & Brownridge, 2010). Our study, therefore, suggests the need for continued efforts to improve assessment for IPV, particularly during pregnancy, and to adapt counseling and educational approaches to address lethality (Messing et al., 2017).
We also found that alcohol use appears to have played a role in the homicides of nonpregnant victims. The relationship between substance use in general and IPV victimization has been well-established among both nonpregnant and pregnant women (Burns et al., 2011; Haller & Miles, 2003; Tuten et al., 2004; Velez et al., 2006). Thus, our study supports previous research on IPV and its relationship to substance abuse issues in victims. However, although a small body of research has examined the intersection of IPH and substance use in nonpregnant victims (Mathews et al., 2009; Sharps et al., 2003), to our knowledge, virtually no research has examined the intersection of substance use and pregnancy among female victims of IPH. Future studies should seek to better understand the contextual factors associated with substance use and IPH among both pregnant and nonpregnant victims and, more specifically, to what degree substance use or other behavioral health issues places women of reproductive age at risk of IPH.
Although not statistically significant, we also found that nearly one fourth of pregnant victims’ deaths and one fifth of nonpregnant victims’ deaths were precipitated by other serious crimes (e.g., drug trafficking, robbery, burglary, motor vehicle theft). Although the significance of this finding is unclear, as it is unclear if the precipitating crime was committed by the perpetrator, victim, or someone else, this finding nonetheless suggests that IPH among reproductively aged victims may be associated with higher degrees of criminal activity among both the perpetrators and the victims. There is some support for this notion in the literature, as research has found that male perpetrators of IPV are often involved in and arrested for other crimes at rates relatively higher than general populations of men (Ramirez, 2005; Straus & Ramirez, 2004). Furthermore, research has shown the presence of criminogenic influences among incarcerated pregnant women, including being the victim of physical and/or sexual abuse (Gilfus, 2002; Greene et al., 2000; Salisbury & Van Voorhis, 2009). Thus, future studies should also strive to understand the relationship of both perpetrator and victim involvement in other criminal activities and risk of IPH among women of reproductive age.
Finally, we also found that certain factors including past and present mental health issues, a reported history of IPV or other crisis (unspecified) a month prior to the homicide, recent family problems, and/or an argument with the perpetrator at the time of death were proportionately higher among nonpregnant victims. However, it is entirely possible that these issues were underreported among pregnant victims. Nonetheless, our study supports some of the extant findings on known precipitating events related to IPH (i.e., previous history of abuse, prior help-seeking for behavioral health issues; Campbell et al., 2003; Jones et al., 2014) and further suggests that other precipitating factors may need to be investigated further and included in screening for IPH risk. In particular, future research should focus on identifying the nature and context of other crises IPV victims experience, including disturbances un-related to the intimate partnership, that may not be apparently related to IPV but that may signal the potential for a lethal event. In addition, current lethality assessments may need to be updated to include questions related to victims’ experiences of crises with their families, and lives, more generally.
Limitations
This study has several limitations worth mentioning. First, the limited nature of the NVDRS data and missing information in the data set make deriving definitive conclusions about the risk of IPH among pregnant and nonpregnant victims difficult. Likewise, underreporting of IPV and a failure to confirm when homicides are IPV related in the data may also mean that many instances of IPH are not captured in the NVDRS. Furthermore, the NVDRS is an “opt-in” system, meaning that only those states that have both the desire to report in and the capacity to do so participate, thus limiting the generalizability and robustness of the NVDRS data. It is entirely likely that our analysis, therefore, may have yielded different results given higher quality (i.e., more complete, accurate capture of all IPH) and/or more comprehensive (i.e., reporting from all 50 states) data. Finally, the small number of postpartum victims included in the sample limited our ability to consider them a unique category; thus, we were unable to identify any unique contextual factors associated with the postpartum period.
Conclusion
Despite these limitations, our study has implications for thinking about how to improve early identification of risk of IPH. Consideration should be given to inclusion of a wider range of IPH-related risk factors in IPV assessments, particularly substance use. Furthermore, directly asking about behavioral health or other contextual factors indicating potential crises (e.g., involvement in criminal justice, familial crises) could potentially assist in identifying IPV victims’ risk of IPH and providing an opportunity for intervention. Future studies should seek to develop effective interventions to prevent IPH, particularly among reproductive aged women.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
