Violence As a Direct Cause of and Indirect Contributor to Maternal Death

Abstract

Background: Death during pregnancy and postpartum in the United States is an issue of urgent and growing concern. Mortality from obstetric-related causes is on the rise, and pregnancy-associated homicide remains a leading cause of death. It is unknown how the context in which women live contributes to their risk of obstetric or violent death during pregnancy and the postpartum period. This study aimed to quantify incidence of mortality from obstetric-related causes and violent death during pregnancy and up to 1-year postpartum, and to identify associations between state-level violent crime rates and incidence of pregnancy-related mortality and pregnancy-associated homicide.

Materials and Methods: We conducted a retrospective, ecologic analysis of all pregnancy-associated homicides in 17 states participating in the National Violent Death Reporting System from 2011 to 2015. Pregnancy-related mortality was identified by International Classification of Diseases-10 code for underlying cause of death in death records issued in the same states and years, data provided by the National Center for Health Statistics. We characterized decedents of both violent and nonviolent maternal death (n = 174 and 1,617, respectively). Five-year mortality ratios (deaths per 100,000 live births) were estimated for both pregnancy-related mortality and pregnancy-associated homicide in every state. Poisson regression models estimated associations between violent crime and maternal death, adjusting for area-level socioeconomic conditions.

Results: Both pregnancy-related mortality and pregnancy-associated homicide ratios were higher in states with higher rates of violent crime (relative risk [RR] = 1.05, 95% confidence interval [CI] = 1.01–1.12; RR = 1.17, 95% CI = 1.01–1.34, respectively).

Conclusion: Broad population-wide violence prevention efforts may help reduce incidence of maternal mortality from both obstetric and violent causes.

Introduction

Pregnancy-related mortality is the death of a woman while pregnant or within 1 year of the end of pregnancy from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes.¹ It is an issue of urgent concern in the United States where maternal mortality ratios rank the country 50th in the world and last among developed countries.^2,3 Moreover, incidence of these deaths appears to be on the rise, a trend evident even after adjustments for enhanced surveillance and improved data collection.⁴ An increasing number of women are entering pregnancy at older ages and/or with pre-existing conditions—hypertension,^5–7 diabetes,⁸ and chronic heart disease⁹—all of which may increase risk of an unhealthy pregnancy and likelihood of death. But beyond clinical factors, little is known about the social and contextual factors that may predispose women to severe morbidity and ultimately mortality if and when they become pregnant, and the Centers for Disease Control and Prevention (CDC) has called the identification of such factors a critical need for public health actions aimed at reducing maternal mortality in this country.¹⁰ Limited evidence suggests that contextual factors may be indirectly related to maternal mortality by elevating women's risk for development of prepregnancy chronic morbidities or for complications arising during pregnancy.^11,12 The acute and chronic stress associated with living in violent areas is a frequently hypothesized mechanism underlying the consistent association between violent crime and adverse birth outcomes.^13–17 It may similarly be associated with increasing risk of pregnancy-related mortality.

Mobilized by the recent passage of federal policy to address maternal death in the United States,¹⁸ maternal mortality review committees are convening in states across the country to evaluate clinical causes, preventability of cases, and develop recommendations to improve clinical care.¹⁹ Concurrently, a growing number of studies have identified homicide as a leading cause of death during pregnancy and postpartum,^20–30 yet it receives far less attention by clinical and public health prevention efforts. Mortality ratios for pregnancy-associated homicide—homicide of a woman during pregnancy or within 1 year of the end of pregnancy—exceed any one of the leading obstetric causes defining pregnancy-related mortality (e.g., hemorrhage, hypertensive disorders, sepsis).^29,31 These women are direct victims of violence, often perpetrated by an intimate partner,^28,29 and these deaths may have resulted from and indeed contribute to the violent context in which they occur. Studies have found that intimate partner violence (IPV) is more likely to occur in areas with high rates of community violence^32,33 as well as structural features (concentrated poverty, unemployment, residential instability, availability of resources)³⁴ where community violence is often colocated.^33,35 Women living in such contexts may therefore be more likely to experience IPV during pregnancy and, if not fatally, experience increased risks of adverse maternal health outcomes as a result.³⁶ Therefore, violent contexts may contribute to both violent and obstetric maternal deaths (pregnancy-associated homicide and pregnancy-related mortality, respectively).

Identifying features of the social environment that are potentially harmful to women of reproductive age can guide population-wide interventions to prevent maternal death from both obstetric and nonobstetric causes. The purpose of this analysis was to explore how violent contexts are associated with increased risk of both pregnancy-related mortality and pregnancy-associated homicide in states across the United States. We hypothesized that incidence of both pregnancy-associated homicide and pregnancy-related mortality would be higher in states with higher rates of violent crime.

Materials and Methods

Two sources of data were used to identify maternal deaths due to violence (pregnancy-associated homicide) and maternal deaths due to obstetric causes (pregnancy-related mortality).

Identification of pregnancy-associated homicides

We conducted a retrospective analysis of all pregnancy-associated homicides reported to the United States National Violent Death Reporting System (NVDRS) from 2011 to 2015. NVDRS is the only state-based surveillance system that pools data on violent deaths from multiple sources into a usable, anonymous database.³⁷ These sources include state and local medical examiner, coroner, law enforcement, toxicology, and vital statistics records.³⁷ Individual state agencies coordinate the data abstraction process across sources. Both state offices and the CDC conduct quality control and reliability checks focusing in particular on abstractor-assigned variables. State participation in NVDRS has been incremental with 17 states participating by 2011. This analysis includes NVDRS Restricted Access Database data from these 17 states: Alaska, Colorado, Georgia, Kentucky, Maryland, Massachusetts, New Jersey, New Mexico, North Carolina, Ohio, Oklahoma, Oregon, Rhode Island, South Carolina, Utah, Virginia, and Wisconsin.

In NVDRS, pregnancy status of women decedents is abstracted from coroner/medical examiner records and/or the death certificate. Coding structure matches that of the 2003 Revision to the United States Standard Certificate of Death, which allows certifiers to indicate whether the woman was pregnant at the time of death, had been pregnant within the past 42 days of death, within 43 days to 1 year, or not pregnant within the past year at the time of death. For NVDRS states that had not implemented the 2003 revised death certificate, NVDRS includes options to indicate “not pregnant, not otherwise specified” or “pregnant, not otherwise specified.” For the purposes of our analysis, we grouped women marked “pregnant, not otherwise specified” with those marked pregnant at the time of death and those marked “not pregnant, not otherwise specified” with those not pregnant at the time of death. We also combined women pregnant within 42 days of death with those pregnant 43 days to 1 year of death into a single postpartum category.

To ensure complete case ascertainment, we also searched the coroner/medical examiner and the law enforcement narratives associated with incidents in which the woman victim's pregnancy status had been marked not pregnant or unknown. Search terms included words “pregnant,” “baby,” “infant,” and “birth.” Narratives that definitively indicated that the women had been pregnant or postpartum (baby age <1 year) were reclassified to the correct pregnant or postpartum status.

NVDRS includes a circumstantial variable to indicate whether the homicide was related to immediate or ongoing conflict or violence between current or former intimate partners, including all deaths where a victim is killed by his or her current or former intimate partner (defined as a current or former girlfriend or boyfriend, date, or spouse). Also included is a variable to indicate whether the death was the result of a conflict between intimate partners (jealousy/lover's triangle). These data are abstracted from review of coroner or medical examiner records and police reports. Cases where either the homicide was perpetrated by an intimate partner or was the result of a conflict between intimate partners were classified as having been IPV related. To ensure complete case ascertainment of IPV-related incidents, we also searched both the law enforcement and coroner/medical examiner narratives for descriptions involving violence between intimate partners (or resulting from a conflict between intimate partners) among incidents with an IPV circumstantial variable coded “no, unknown” in the raw data.

Identification of pregnancy-related deaths

We identified all pregnancy-related deaths among death records issued from 2011 to 2015 in the same 17 states, data available from the National Center for Health Statistics. Cases of pregnancy-related death were records with Chapter XV International Classification of Diseases (ICD)-10 codes for underlying cause of death (Pregnancy, childbirth, and the puerperium, O.00–O.99). We applied the World Health Organization Application of ICD-10 to Deaths during Pregnancy, Childbirth, and the Puerperium: ICD-Maternal Mortality³⁸ standardized coding scheme to group pregnancy-related deaths by cause.

State-level violent crime

Annual counts of violent crimes (murder and non-negligent manslaughter, rape, robbery, and aggravated assault) and total population counts are available through the Federal Bureau of Investigation's Uniform Crime Reporting Program.³⁹ We estimated the 2011–2015 state violent crime rate by averaging annual rates (crimes per 100,000 population) in each of the 17 states included in this analysis.

Mortality ratio estimation and modeling

For each state, we estimated 5-year pregnancy-associated homicide and pregnancy-related mortality ratios per 100,000 live births (sum of deaths from 2011 to 2015 divided by sum of live births from 2011 to 2015). Data on live births by state 2011–2015 were provided by the National Center for Health Statistics.

We conducted Poisson regression to model the count of deaths with a live births offset (ratio denominator) to estimate the association between state-level violent crime and violent and nonviolent maternal death. Given large differences in state socioeconomic conditions—factors that may be related to both violent crime rates and risk of maternal death—models controlled for both the absolute level of resources within the state population (median household income). In addition, given documented links between income inequality and both homicide⁴⁰ and pregnancy-related mortality,⁴¹ we controlled for the degree of inequality in the distribution of resources across the population (Gini index). Both measures were obtained from the American Community Survey's 2015 five-year estimates.⁴²

We conducted sensitivity analyses to ensure the robustness of our findings in light of data quality and availability issues. Given the known difficulty in correctly classifying maternal deaths based on administrative data,^43,44 we restricted our sample to states with enhanced pregnancy surveillance (those utilizing the 2003 Standard United States Certificate of Death, which includes a checkbox for identification of the decedent as pregnant or up to 1-year postpartum at the time of death). While neither our pregnancy-associated homicide nor pregnancy-related mortality case identification relied on the checkbox information, it may be that misclassification of maternal death in our analyses was less likely among states collecting such information (in particular for identification of pregnant and postpartum women in NVDRS where the checkbox may have been used in conjunction with other case information to code the decedent by pregnancy status). Finally, we repeated the modeling after removing Utah, the only state where zero cases of pregnancy-associated homicide were identifiable throughout the study's 5-year time frame. This study of deidentified secondary data was exempt from Institutional Review Board approval.

Results

The NVDRS identified 166 pregnancy-associated homicides occurring in the 17 states participating from 2011 to 2015 (n = 113 pregnant, n = 53 postpartum). After review of coroner/medical examiner and law enforcement narratives among cases with pregnancy status variable indicating not pregnant, missing, or unknown, we identified eight additional cases (all postpartum). There were an additional six incidents where narratives referenced the victim's infant child, but contained no information on age of the infant. These victims were not included in the final total of 174 victims confirmed pregnant or up to 1-year postpartum, yielding a pregnancy-associated homicide ratio of 2.9 deaths per 100,000 live births. Ratios varied across states from a high of 5.7 in Georgia to 0.28 in Massachusetts, with no cases identified in the state of Utah (Table 1).

Table 1.

2011–2015 Five-year Estimates of Maternal Death, Violent Crime, and Median Household Income and Income Inequality Across Seventeen U.S. States

	Mean (SD)	Min	Max
Pregnancy related mortality ratio (per 100,000 live births)	24.7 (12.3)	6.9	56.9
Pregnancy-associated homicide ratio (per 100,000 live births)	2.7 (1.6)	0	5.7
IPV-related pregnancy-associated homicide ratio (per 100,000 live births)	1.7 (1.1)	0	3.5
Violent crime rate (per 100,000 inhabitants)	3.63 (1.5)	2.0	7.4
Median household income (2015 inflation-adjusted US$)	56,619 (10,653)	43,740	74,551
Income inequality (Gini index)	0.46 (0.02)	0.42	0.48

IPV, intimate partner violence; SD, standard deviation.

Of the 174 total pregnancy-associated homicides, ∼60% involved IPV (n = 104). Characteristics of IPV-related homicide victims were similar to all homicide victims (Table 2). In both cases, most victims were less than age 30, approximately two-thirds were pregnant at the time of death and >60% involved firearms.

Table 2.

Characteristics of Violent and Nonviolent Maternal Deaths, Seventeen U.S. States, 2011–2015

	All pregnancy-associated homicides (n = 174)	IPV-related pregnancy-associated homicides (n = 104)	Pregnancy-related deaths (n = 1,617)
Age
<20	20 (11.5)	11 (10.6)	58 (3.6)
20–24	61 (35.1)	33 (31.7)	215 (13.3)
25–29	40 (23.0)	25 (24.0)	295 (18.2)
30–34	30 (17.2)	19 (18.3)	329 (20.4)
35+	23 (13.2)	16 (15.4)	720 (44.5)
Race/ethnicity
White, non-Hispanic	69 (39.7)	44 (42.3)	773 (47.8)
Black, non-Hispanic	80 (46.0)	41 (39.4)	610 (37.7)
Other	25 (14.4)	19 (18.3)	234 (14.5)
Timing of death
Pregnancy	113 (64.9)	64 (61.5)	—
Postpartum	61 (35.1)	40 (38.5)	—
Weapon type
Firearm	107 (64.1)	61 (60.4)	—
Sharp instrument	24 (14.4)	15 (14.9)	—
Hanging, strangulation, suffocation	20 (12.0)	14 (13.9)	—
Other	16 (6.5)	11 (10.9)	—
IPV related
No, not available, unknown	70 (40.2)	—	—
Yes	104 (59.8)	—	—
Pregnancy-related cause of deatha
Hypertensive disorders in pregnancy, childbirth, and the puerperium	—	—	84 (5.2)
Obstetric hemorrhage	—	—	61 (3.8)
Pregnancy with abortive outcome	—	—	34 (2.1)
Pregnancy-related infection	—	—	22 (1.4)
Contributory conditions	—	—	343 (21.2)
Other obstetric complications	—	—	228 (14.1)
Nonobstetric complications	—	—	511 (31.6)
Other and unknown	—	—	334 (20.7)

^aSee WHO ICD-MM for ICD-10 codes included in each maternal death grouping category.³³

WHO, World Health Organization; ICD-MM, International Classification of Diseases-Maternal Mortality.

There were 1,617 pregnancy-related deaths from 2011 to 2015 in the 17 states included in this analysis with mortality ratios averaging 24.7 deaths per 100,000 live births and ranging from 6.9 in Massachusetts to 56.9 in Georgia (Table 1). These women tended to be older than women victims of violent death with hypertensive disorders and hemorrhage as leading causes (Table 2). Timing of obstetric death was not discernable from birth record data alone.

Violent crime averaged 3.6 events per 100,000 state residents from 2011 to 2015, ranging from 2.0 in Virginia to 7.4 in Alaska (Table 1). States with higher violent crime rates had higher incidence of pregnancy-associated homicide, specifically IPV-related pregnancy-associated homicide (Table 3). Moreover, states with higher violent crime rates had higher incidence of obstetric death with the pregnancy-related mortality ratio increasing by 5% for every 1-point increase in the violent crime rate, independent of socioeconomic conditions within the state (adjusted relative risk [RR] = 1.05, 95% confidence interval [CI] 1.01–1.12).

Table 3.

Associations Between Violent Crime and Pregnancy-Associated Homicide and Pregnancy-Related Mortality in Seventeen U.S. States, 2011–2015

	Pregnancy-associated homicide		IPV-related pregnancy-associated homicide		Pregnancy-related mortality
	RR	95% CI	RR	95% CI	RR	95% CI
Violent crime (per 1-unit increase)	1.17	(1.01–1.34)	1.26	(1.07–1.50)	1.05	(1.01, 1.12)
Median household income (per US $10,000 increase)	0.77	(0.6–0.99)	0.89	(0.73–1.06)	0.95	(0.91, 0.99)
Income inequality (per 1-U increase)	1.06	(0.94–1.19)	1.05	(0.91–1.22)	1.15	(1.11, 1.20)

CI, confidence interval; RR, relative risk.

Results from both sensitivity analyses were consistent with our primary findings (data not shown). Violent crime remained associated with both violent and nonviolent maternal deaths in the sample limited to states with enhanced pregnancy surveillance (GA, KY, MD, NJ, OH, OK, OR, RI, SC, UT) and in the sample that excluded Utah.

Discussion

In this analysis of maternal deaths occurring across 17 U.S. states from 2011 to 2015, we found a significant positive association between the level of violent crime in the state and both pregnancy-related and pregnancy-associated homicide incidence. Our purpose was to demonstrate that efforts beginning far enough upstream—those that aim to alter the context in which women are born, live, work, and potentially become pregnant and give birth—may reduce maternal death from both obstetric and nonobstetric causes simultaneously.

Our results suggest that violence prevention efforts may not only directly prevent homicide during pregnancy and postpartum (a leading single cause of maternal death), but may also indirectly reduce maternal mortality from obstetric causes through the elimination of a psychosocial environmental stressor.

Previous research on social environment and reproductive health hypothesizes that persistent exposure to violence induces a chronic stress response,^15,45,46 which may in turn lead to dysfunction of maternal cardio metabolic processes⁴⁷ contributing to the pathogenesis of severe maternal morbidities. High crime areas may also influence maternal health through associations with harmful health behaviors¹⁷—influencing decisions to smoke or use alcohol—or by limiting access to sexual and reproductive health care services and other health promoting resources (safe spaces for physical activity, healthy foods, social support, and others)⁴⁸ before, during, and after pregnancy.

IPV may lie along the pathway to both pregnancy-associated homicide and pregnancy-related mortality in places where high violent crime increases IPV incidence in turn resulting in either fatal assault or the development of obstetric complications leading to maternal death.³⁶ Moreover, it may be that some cases of pregnancy-related mortality may not have occurred in the absence of IPV and therefore may justifiably be deemed homicides at their most fundamental underlying cause. Unfortunately death records lack detailed situational information to explore the contribution of IPV to the development of fatal maternal morbidities and the intersection of pregnancy-related mortality and pregnancy-associated homicide.

Violence as a harmful contextual feature may be more easily modifiable than others (such as entrenched socioeconomic depravation). A growing body of literature highlights public health approaches to violence prevention that are increasingly focused at the community- and societal levels with the potential to achieve broad violence prevention impact across the population.³³ Such efforts could be integrated into maternal mortality review committee recommendations for the prevention of future maternal deaths. Considering the broader contexts in which maternal deaths occur enables review committees to identify area- and systems-level contributing factors,¹⁹ including violence. This multilevel approach may yield recommendations for place-based prevention (via partnerships with community-based organizations located in high violence areas) as well as recommendations for policy- and systems-level interventions (such laws that restrict firearm possession by domestic violence offenders, or incentivizing behavioral health screening with enhanced reimbursement by insurers) that may ultimately result not only in fewer maternal deaths but community violence prevention more broadly.

This study utilized the most comprehensive source of population-level data on pregnancy-associated homicide (NVDRS), an outcome notoriously under-reported in administrative records. By combining multiple data sources (death records, coroner and medical examiner information, toxicology data, and law enforcement reports), the NVDRS represents a significant improvement in ascertainment of pregnancy status over reliance on death records alone.⁴⁹ We sought to further the goal of complete case ascertainment by searching incident records to identify narratives describing a victim's recent (<1 year prior) pregnancy or child (<1 year old). Our classification of the eight pregnancy-associated homicides identified in the search may have been incorrect where police report narratives were incorrect. However, there is still likely to be under ascertainment of cases in particular among women with early pregnancies or postpartum women who do not have custody of their children or had experienced an infant loss.⁵⁰ The analytic implications of under ascertainment suggest that our estimates of association between violent contexts and pregnancy-associated homicide incidence are conservative and may be higher. We also acknowledge the possibility of misclassification among pregnancy-related mortality cases where decedents were incorrectly assigned a Chapter XV ICD-10 code for underlying cause of death.⁵¹ Thus, our estimates of pregnancy-related mortality should be interpreted with caution.

Our analysis has additional limitations to consider. First, this analysis is cross-sectional, and prohibits causal interpretation of the association between violent crime and maternal death. Second, given the relative rarity of maternal deaths, our analysis combined 5 years of data at the state level to achieve an adequately large number of cases for analysis. As such we are unable to explore annual trends in mortality longitudinally, or mortality at a finer geographic resolution. It may be that the association between violence and maternal mortality is even greater at city- or metropolitan area levels where rates of violent crime and victimization are consistently higher than nonmetropolitan or rural areas.⁵² Third, with data from only 17 states included in this analysis, we are unable to explore racially stratified incidence of pregnancy-related mortality and pregnancy-associated homicide in association with violent crime as an even smaller number of states would have a sufficient number of deaths in separate racial categories, thereby substantially limiting analytic power. We acknowledge that such an analysis is critical to addressing the large and persistent inequity in maternal mortality experienced by non-Hispanic black women relative to whites, and future studies utilizing a larger sample of states or years should explore how social context differentially impacts women by race and contributes to the maintenance of racial health inequities.

Conclusions

Our findings highlight the ongoing need for clinical and public health efforts to recognize community violence as a significant contributor to maternal mortality in this country. They also highlight the magnitude of violence as a leading cause of maternal death and engage in meaningful responses to identify, review, and take steps to prevent these deaths with the same rigor and vigilance paid to obstetric causes. Illinois provides a promising example of addressing this issue where the establishment of a maternal mortality review committee dedicated to violent deaths works in parallel to the long-standing committee to review clinical care pathways of obstetric deaths.⁵³ Adoption of similar approaches in other states may aid in the development and wide-scale adoption of violence prevention efforts that protect the health of pregnant and postpartum women and their families.

Disclaimer

Its contents are solely the responsibility of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or the Department of Health and Human Services or the National Institutes of Health or the Louisiana Department of Health. None of the funders had involvement in the study design, analysis, or interpretation of data, writing of the report, or the decision to submit the article for publication.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported in part by the American Public Health Association/Centers for Disease Control and Prevention National Violent Death Reporting System New Investigator Award and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development grant number R01HD092653.