Intimate partner violence is a significant public health problem. Reducing physical abuse directed at women by male partners is one of 19 violence prevention objectives addressed in Healthy People 2010.1 Although progress toward this objective has been made, violence directed against women by male partners remains common with 25% of US women reporting a history of partner violence.2 The abuse of women during pregnancy is of particular concern. The majority of studies report a prevalence of 3.9% to 8.3% during the index pregnancy, which would result in the abuse of approximately 157,000 to 335,000 women who deliver live-born infants each year in the United States.3,4
Studies of the relationship between partner violence and adverse birth outcomes such as low birth weight (LBW) and preterm birth have produced conflicting results.5–12 Most studies have addressed only physical violence, and exposure data were most often collected retrospectively in the postpartum period or in one or two prenatal interviews. Moreover, exposure periods may have been imprecise (eg, partner violence in the past year) and the victim's relationship to the abuser unspecified. Although LBW and preterm birth have been most frequently studied, no outcome has consistently been associated with partner violence during pregnancy.
This study examined the impact of police-reported intimate partner violence during pregnancy on birth outcomes to determine if partner violence is associated with LBW, preterm birth, and fetal or neonatal death. The use of police data offered documentation of exposure throughout the pregnancy, and exposure data were collected independently of and before the outcomes under study.
MATERIALS AND METHODS
This was a population-based, retrospective, cohort study. Exposed subjects were women 16 to 49 years old who had reported one or more partner violence incidents to the Seattle Police Department in the years 1995 through 1998 and who subsequently had a singleton live birth or fetal death registered in the state of Washington within a time frame that indicated they were pregnant at the time of the incident. The time interval from incident to birth or fetal death was calculated from the birth certificate using last menstrual period (LMP greater than 100 and less than 320 days) or estimated gestational age (greater than 20 and less than 52 weeks) if LMP was missing; if both were missing, 300 days was used as the interval (n = 8). The comparison or unexposed group was composed of Seattle residents with a singleton birth or fetal death registered in the state of Washington without a police-reported incident in Seattle during the study period. Unexposed subjects were sorted by race/ethnicity and 5-year age groups; the files were randomized, and unexposed subjects were then frequency matched within each race/age category at a ratio of 8:1 to exposed subjects.
The computerized police files were linked to the Washington State birth and fetal death certificate files using SAS 6.12 (SAS Institute Inc., Cary, NC). The best and most conclusive matches were selected using a hierarchical set of match criteria (Appendix). Fifty-five percent of the matches for the 389 births in the final dataset (described below) were made with the first criteria composed of the mother's maiden name, first name, middle name, and date of birth. Of the approximately 13,000 incidents involving 10,088 individual women meeting the inclusion criteria, 487 incidents initially linked to births or fetal deaths. The 4.9% linkage rate corresponds to the 1996 birth rate of 48.5 per 1000 women of reproductive age in Seattle.
Of the 487 linked incidents, 402 resulted in a singleton birth or fetal death with adequate matching criteria. Three of these incidents were excluded from the analysis because they were found to not pertain to intimate partner violence when police reports were reviewed, and ten were excluded because they were a second (n = 9) or third birth (n = 1) linked to an incident report for the same woman during the study period. After exclusions, 389 abused women with a birth or fetal death remained for analysis; 3090 nonabused women with a birth or fetal death were selected for comparison.
The source of research material on each subject was data previously collected by authorized city and state agencies. No contact was made with subjects. This study was approved by the Seattle Police Department and the Washington State Department of Health Human Research Review Board.
For the purpose of this study, any intimate partner violence exposure was defined as having one or more police-reported incidents in Seattle, Washington, during pregnancy, perpetrated by a man with whom the victim had an intimate relationship (ie, a spouse, boyfriend, or dating relationship) at the time of or before the incident, as determined by the subject's self-report. Additionally, incidents were defined as physical or nonphysical (ie, psychological or emotional) violence based on the offense assigned to the case by the police. An incident classified as physical assault or reckless endangerment was categorized as physical violence. Reckless endangerment, as defined by the local municipal codes, is an action that creates a substantial risk of death or serious physical injury; only one case was coded as such. The remaining offenses were categorized as nonphysical violence (criminal trespass, custodial interference, disturbance, harassment, menacing, property damage, stalking, threats, theft/burglary, violation of court order, suspicious circumstances, and warrant). An incident was selected only if it was determined by the police to be founded (incident assigned to an officer, charge made by an officer, incident deemed to be “family violence,” or other relevant action taken). In cases with more than one incident reported during the study period, the incident involving physical violence was selected for analysis; if the incidents were of the same category, the most recent incident was selected.
Birth outcomes were selected from the first pregnancy to occur during the study period. Outcomes were determined from the birth certificate data and included LBW (less than 2500 g), very LBW (VLBW less than 1500 g), preterm birth (20–36 weeks' gestation), very preterm birth (20–31 weeks' gestation), fetal death (20 weeks' gestation or more), and neonatal death (before hospital discharge or completion of birth certificates in the case of nonhospital births). Birth weight data were nearly complete with 1% missing for exposed and unexposed subjects, and no data were missing for fetal and neonatal death. Gestational age was calculated from the LMP. Gestational age data were missing in 24% and 28% of exposed and unexposed subjects, respectively; out-of-range data (less than 20 weeks' or greater than 44 weeks' gestation) occurred uniformly in 2% of subjects.
All analyses were conducted with Stata 6.0 (Stata Corp., College Station, TX). Pearson χ2 tests were calculated to compare demographic, behavioral, and obstetric history characteristics of exposed and unexposed subjects; a P value of < .05 was considered statistically significant.
Unconditional logistic regression was used to compute adjusted odds ratios (aORs) with 95% confidence intervals (CIs) in the analysis of each birth outcome and any partner violence, physical violence, and nonphysical violence. Potential confounding factors, obtained from the birth and fetal death certificate files, were sequentially added to the model if they altered the OR of the exposure by 10% or more.13 Confounding by race/ethnicity and age was accounted for by frequency matching of exposed and unexposed subjects in the study design.
For each outcome and exposure modeled, the following demographic variables were considered as potential confounders: marital status of the infant's parents (married, not married), mother's education level (less than 12, 12, more than 12 years), source of payment for delivery (subsidized/no insurance, commercial insurance), and public health benefits received (Special Supplemental Nutrition Program for Women, Infants and Children, First Steps [an expanded Medicaid program for pregnant women], Aid to Families With Dependent Children, or local health department prenatal care services). After adjusting for these factors, the following behavioral and obstetric variables were then considered as potential confounders: smoking during pregnancy (yes, no), alcohol use during pregnancy (yes, no), adequacy of prenatal care using the Kotelchuck Summary Adequacy of Prenatal Care Utilization Index (adequate, inadequate; inadequate = initiation of prenatal care begun in month 5 or later or less than 50% of expected visits received),14 parity (0, 1, 2 or more), previous spontaneous abortion (fetal death less than 20 weeks; yes, no), previous induced abortion (yes, no), and previous fetal death (20 weeks or more).
A series of dummy variables were created for each potentially confounding factor. A separate category was created for missing values for each of the above factors in the multivariable analysis, with the category of missing assigned the highest risk level. This was important given that a substantial proportion of data in the birth certificate files was missing for several variables and because missing data occur more commonly among women at higher risk.15 The highest percentage of missing data occurred with demographic variables, with data missing at comparable levels in the exposed and unexposed groups for subjects' educational level (approximately 30%), source of payment (20%), and marital status (1%). Missing data for receipt of public health benefits were not discernable as the coding mechanism only allowed for a positive response. For behavioral and obstetric history variables, the proportions of missing data were equally distributed between the two groups; for the most part, these proportions were more modest (4–7%), with the exception of drinking during pregnancy (approximately 14%) and adequacy of prenatal care (28%).
Of the 389 abused women, 342 (87.9%) reported one incident during the index gestational period, 37 (9.5%) two incidents, and ten (2.6%) three to five incidents. Of the 389 incidents selected for analysis, the majority (72%) were categorized as physical violence. Approximately one third of women with reported partner violence during pregnancy were of non-Hispanic, white race/ethnicity and one third were non-Hispanic, black (Table 1). Over 80% were under 30 years old, and 20% were teenagers.
Women reporting any partner violence during pregnancy were significantly more likely than women who had not reported partner violence to have a lower educational level, be unmarried, and use subsidized or no insurance. Although women reporting violence were more likely to access public health programs, such as the Women, Infants, and Children's and First Steps programs, they were no more likely than women without reported violence to access these programs when restricting the analysis to those eligible, as assessed by insurance type (data not shown). Women reporting violence were also more likely to smoke or drink alcohol during pregnancy, receive inadequate prenatal care, to be of higher parity, and to have experienced a previous spontaneous abortion, induced abortion, or fetal death.
All of the outcomes measured, with the exception of fetal death, were significantly associated with any reported partner violence after adjusting for demographic, behavioral, and obstetric history risk factors as shown in Table 2. Women reporting any violence were more than 1.5 times as likely to have a LBW infant or preterm birth compared with women without reported violence (aOR 1.70; 95% CI 1.20, 2.40; and aOR 1.61; 95% CI 1.14, 2.28, respectively). Exposure to any reported partner violence increased the risk of VLBW by more than two-fold, and the risk for very preterm birth was increased more than three-fold.
Any partner violence was also associated with a threefold increased risk of neonatal death. Seven (58%) of 12 neonatal deaths with known, reliable gestational age (20–44 weeks) were preterm births. Three (75%) of the four deaths in the exposed group and four (50%) of the eight deaths in the unexposed group were preterm births; all but one were very preterm births. The remaining births occurred at term (39–41 weeks). The crude OR for fetal death was 0.44 (95% CI 0.06, 3.30) for any partner violence. The small number of fetal deaths in this sample (one in the exposed group and 18 in the unexposed group) precluded further analysis.
In separate analyses of physical and nonphysical partner violence, patterns of increased risk remained for both groups. The estimates in the physical violence subcohort were comparable to those of the entire cohort (Table 2); however, most estimates in the nonphysical subcohort were not statistically significant after adjusting for other factors (data not shown).
This study is one of the few population-based studies conducted to date and the first to examine the relationship between intimate partner violence and birth outcomes using police data. Measurement has been a consistent problem in previous studies of both prevalence of partner violence during pregnancy and outcomes related to partner violence. These studies rely on self-report, which may be constrained by the subject's fear of her partner, shame or loss of self-esteem, entrapment and lack of empowerment, and rationalization or coping mechanisms.16
Our findings of increased risk for adverse birth outcomes associated with partner violence are noteworthy in light of the inconsistent results found in previously published studies.5–12 In a case–control study by Campbell et al, in which women were matched on hospital, ethnicity, and age group, physical and nonphysical partner abuse during pregnancy were assessed postpartum before hospital discharge.11 Physical and nonphysical abuse were significantly associated with LBW (OR 3.29; 95% CI 1.18, 9.18; and OR 3.78; 95% CI 1.31, 10.90, respectively) in full-term, but not preterm infants. The association became nonsignificant for both exposure groups, however, when adjusted for other demographic as well as behavioral and obstetric risk factors.
Similarly, Bullock and McFarlane interviewed women delivering in private and public hospital settings 24 hours postpartum.6 Women delivering in a private hospital who were physically abused by their current male partners, either during or before pregnancy, were found to have an increased risk of LBW (OR 4.80; 95% CI 1.79, 12.90); abused women in public hospital settings were not at increased risk. When the data for public and private hospitals were combined, the risk of LBW remained significant after adjusting for race, behavioral, and obstetric risk factors, and hospital.
In two prospective studies of predominately low-income women receiving prenatal care in urban clinics, women experiencing physical or sexual abuse by any perpetrator were found to be at significantly greater risk of delivering a LBW infant.7,9 Upon further analysis of the study by Parker et al, abuse became nonsignificant when demographic, obstetric, and substance use variables were included in the model.8 In a third, prospective, clinic-based study, Covington et al found preterm birth in teens, but not adults, to be associated with severe physical prenatal violence after taking race, behavioral, and obstetric history factors into account.12
One of the strengths of this study was the use of police-report data as a method to identify subjects. The police data provided documentation of partner violence and the ability to assess exposure before the outcomes. On the other hand, the confidentiality of police reports prevents researchers from contacting subjects for additional data gathering or subsequent interventions. In addition, generalizability of the findings using police data is limited to police-reported partner violence, particularly because these incidents tend to be more severe than unreported incidents.17,18 Our study encountered a small number of nonphysical compared with physical violence events. Although few studies have differentiated between physical and nonphysical partner violence, at least two self-report surveys revealed a higher prevalence of nonphysical than physical violence in the current relationships of women.19,20 It is also possible that some women with police-reported partner violence during the study period moved out of state and were not captured in the data linkage process, although women who delivered out of state and who registered their births in Washington would have been included in the sample (only three subjects, all unexposed, had done so). It is unlikely that a large number of abused women did leave the state altogether, given the lack of social support and financial resources of many abused women, which may limit their mobility.21,22 Our study also may have misclassified a portion of women in the cohort as unexposed because only 25% to 50% of partner violence is estimated to be reported to police.2,18 Nevertheless, the use of police-reported incidents would likely result in higher risk estimates for adverse birth outcomes than would be found with all partner violence because more severe violence may have a greater impact than that of nonpolice-reported violence.
Although police report data may not represent the totality of the intimate partner violence experience for each woman, police-reported incidents are likely indicators of ongoing violence because partner violence tends to be chronic and often escalates over time. Nonphysical, as well as physical, violence may also induce a physiologic response as a result of this ongoing stress. Although the current study identified too few nonphysical events to detect an association with adverse birth outcomes, the association of partner violence and psychosocial stress has been documented in several studies,23–25 and stress in pregnancy is thought to be associated with adverse birth outcomes through effects mediated by physiologic responses.26–28 In addition to the key findings related to adverse birth outcomes, important demographic associations with intimate partner violence were revealed in this study. Women reporting any partner violence during pregnancy were significantly more likely than women without reported partner violence to be of lower socioeconomic status and to have a poor obstetric history. Nevertheless, the relationship of partner violence and adverse birth outcomes persisted after controlling for these factors in the multivariable analyses, suggesting that these factors alone do not explain this relationship. Stress in pregnancy, as previously noted, may have direct physiologic effects on birth outcomes above and beyond those of socioeconomic factors.
A limitation of this study, however, was the use of birth certificate data to assess confounding by demographic factors and maternal reproductive histories. Birth certificate data have well-known limitations in terms of data availability and accuracy.15,29–32 The proportion of missing data in our study was substantial for some of the demographic and obstetric history variables. This may have diminished our ability to control for the confounding effect these important factors could have had on our estimates. This is particularly important in light of the fact that at-risk subgroups are overrepresented in records with missing data.15,33 We did account for missing data in multivariable analyses by creating a missing data category for each variable and assigning this category to the highest risk group. In doing so, we may have reduced this bias to some extent. Previous studies of birth certificate data have suggested that data accuracy is also of concern.31,32,34,35 The findings in those studies, however, varied in terms of which variables are most valid when compared with medical records and whether prenatal care is over- or underestimated.
In light of the limitations of using public data sources, particularly with regard to identifying potentially confounding factors, the findings in this study should be regarded with caution. It is important to consider, however, that controlling for behavioral and obstetric risk factors in particular may be “over-adjusting” because many of these factors may be a step in the causal pathway. Although few studies have been conducted to determine cause and effect, factors such as substance abuse, unintended pregnancy (resulting in higher parity), and access to health care may occur as a result of partner violence; controlling for these factors might then be inappropriate.
The limitations of using birth certificate data to determine birth outcomes, particularly gestational age, have also been documented.36–38 The LMP, used in this and most other studies to calculate gestational age from birth certificates, has been found to overestimate gestational age. Mothers of high-risk infants, disproportionately represented in low socioeconomic populations, are more likely to receive late or inadequate prenatal care and have the poorest recall of LMP.37,38 Thus, if the data for these births are preferentially excluded because of missing, incomplete, or inaccurate gestational age, the net effect in this study would have been to decrease the proportion of preterm births, producing a bias toward the null. The alternative to LMP is the estimated gestational age variable on the birth certificate. As Alexander and Allen noted in an analysis of US birth data, the use of these data has not been fully explored.38 Furthermore, DiPietro and Allen concluded in their review paper that this estimate is inadequate and that its reliability and validity have not been well established.37
The key findings in this study suggest that when pregnant women are identified at the time of a reported incident, they should be provided health information and referrals to social, health, and crisis intervention services. Referrals to high-risk pregnancy programs, particularly preterm birth and LBW prevention programs, should be aggressively facilitated once violence-affected women are identified because partner violence may be a strong marker for high-risk pregnancy. Conversely, it is important that all avenues of partner violence detection are used, including the incorporation of universal and systematic screening throughout pregnancy and the postpartum period by all health care providers.
Finally, police data provide a critical perspective of partner violence. Further research with enforcement agencies and the judicial system should be conducted, including larger samples of women exposed to partner violence, to discern the effects of different types of violence against women on birth outcomes.
1. US Department of Health and Human Services. Healthy People 2010. Conference Edition. Washington: DHHS, January 2000.
2. Tjaden P, Thoennes N. Extent, nature, and consequences of intimate partner violence, findings from the National Violence Against Women Survey. NCJ 181867. Washington: National Institute of Justice, July 2000.
3. Gazmararian JA, Lazorick S, Spitz AM, Ballard TJ, Saltzman LE, Marks JS. Prevalence of violence against pregnant women. JAMA 1996;275:1915–20.
4. MacDorman MF, Minino AM, Strobino DM, Guyer B. Annual summary of vital statistics—2001. Pediatrics 2002; 110:1037–52.
5. Petersen R, Gazmararian JA, Spitz AM, Rowley DL, Goodwin MM, Saltzman LE, et al. Violence and adverse pregnancy outcomes: A review of the literature and directions for future research. Am J Prev Med 1997;13:366–73.
6. Bullock LF, McFarlane J. The birth-weight/battering connection. Am J Nurs 1989;89:1153–5.
7. Parker B, McFarlane J, Soeken K. Abuse during pregnancy: Effects on maternal complications and birth weight in adult and teenage women. Obstet Gynecol 1994;84:323–8.
8. McFarlane J, Parker B, Soeken K. Abuse during pregnancy: Associations with maternal health and infant birth weight. Nurs Res 1996;45:37–42.
9. Curry MA, Perrin N, Wall E. Effects of abuse on maternal complications and birth weight in adult and adolescent women. Obstet Gynecol 1998;92:530–4.
10. Cokkinides V, Coker A, Sanderson M, Addy C, Bethea L. Physical violence during pregnancy: Maternal complications and birth outcomes. Obstet Gynecol 1999;93:661–6.
11. Campbell J, Torres S, Ryan J, King C, Campbell DW, Stallings RY, et al. Physical and nonphysical partner abuse and other risk factors for low birth weight among full term and preterm babies: A multiethnic case-control study. Am J Epidemiol 1999;150:714–26.
12. Covington DL, Justason BJ, Wright LN. Severity, manifestations, and consequences of violence among pregnant adolescents. J Adolesc Health 2001;28:55–61.
13. Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993;138:923–36.
14. Kotelchuck M. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index. Am J Public Health 1994;84:1414–20.
15. Gould JB. Vital records for quality improvement. Pediatrics 1999;1031 Suppl E:278–90.
16. Smith PH, Tessaro I, Earp JA. Women's experiences with battering: A conceptualization from qualitative research. Womens Health Issues 1995;5:173–82.
17. McFarlane J, Soeken K, Reel S, Parker B, Silva C. Resource use by abused women following an intervention program: Associated severity of abuse and reports of abuse ending. Public Health Nurs 1997;14:244–50.
18. US Department of Justice. Violence against women: Estimates from the redesigned survey. Washington: US Department of Justice, August 1995.
19. Coker AL, Smith PH, McKeown RE, King MJ. Frequency and correlates of intimate partner violence by type: Physical, sexual, and psychological battering. Am J Public Health 2000;90:553–9.
20. Koenig LJ, Whitaker DJ, Royce RA, Wilson TE, Callahan MR, Fernandez MI. Violence during pregnancy among women with or at risk for HIV infection. Am J Public Health 2002;92:367–70.
21. Lutenbacher M, Cohen A, Mitzel J. Do we really help? Perspectives of abused women. Public Health Nurs 2003; 20:56–64.
22. Thompson MP, Kaslow JN, Kingree JB, Rashid A, Puett R, Jacobs D, et al. Partner violence, social support, and distress among innercity African American women. Am J Community Psychol 2000;28:127–43.
23. Amaro H, Fried LE, Cabral H, Zuckerman B. Violence during pregnancy and substance use. Am J Public Health 1990;80:575–9.
24. Campbell JC, Poland ML, Waller JB, Ager J. Correlates of battering during pregnancy. Res Nurs Health 1992;15:219–26.
25. Curry MA. The interrelationships between abuse, substance use, and psychosocial stress during pregnancy. J Obstet Gynecol Neonatal Nurs 1998;27:692–9.
26. Copper RL, Goldenberg RL, Das A, Elder N, Swain M, Norman G, et al. The preterm prediction study: Maternal stress is associated with spontaneous preterm birth at less than thirty-five weeks' gestation. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol 1996;175:1286–92.
27. Wadhwa PD, Porto M, Garite TJ, Chicz-DeMet A, Sandman CA. Maternal corticotropin-releasing hormone levels in the early third trimester predict length of gestation in human pregnancy. Am J Obstet Gynecol 1998;179:1079–85.
28. Hobel CJ, Dunkel-Schetter C, Roesch SC, Castro LC, Arora CP. Maternal plasma corticotropin-releasing hormone associated with stress at 20 weeks' gestation in pregnancies ending in preterm delivery. Am J Obstet Gynecol 1999;180:S257–63.
29. Adams MM, Berg CJ, McDermott JM, Gaudino JA, Casto DL, Wilson HG, et al. Evaluation of reproductive histories constructed by linking vital records. Paediatr Perinat Epidemiol 1997;11:78–92.
30. Clark K, Fu CM, Burnett C. Accuracy of birth certificate data regarding the amount, timing, and adequacy of prenatal care using prenatal clinic medical records as referents. Am J Epidemiol 1997;145:68–71.
31. Dobie SA, Baldwin LM, Rosenblatt RA, Fordyce MA, Andrilla CH, Hart LG. How well do birth certificates describe the pregnancies they report? The Washington State experience with low-risk pregnancies. Matern Child Health J 1998;2:145–54.
32. McDermott J, Drews C, Green D, Berg C. Evaluation of prenatal care information on birth certificates. Paediatr Perinat Epidemiol 1997;11:105–21.
33. Hellerstedt W, Macary S, Oswald J, Jones-Webb R. The reliability of maternal characteristics' data on birth certificates compared with self-report. Paediatr Perinat Epidemiol 2001;15:A15.
34. Buescher PA, Taylor KP, Davis MH, Bowling JM. The quality of the new birth certificate data: A validation study in North Carolina. Am J Public Health 1993;83:1163–5.
35. Piper JM, Mitchel EF Jr, Snowden M, Hall C, Adams M, Taylor P. Validation of 1989 Tennessee birth certificates using maternal and newborn hospital records. Am J Epidemiol 1993;137:758–68.
36. Mustafa G, David RJ. Comparative accuracy of clinical estimate versus menstrual gestational age in computerized birth certificates. Public Health Rep 2001;116:15–21.
37. DiPietro JA, Allen MC. Estimation of gestational age: Implications for developmental research. Child Dev 1991; 62:1184–99.
38. Alexander GR, Allen MC. Conceptualization, measurement, and use of gestational age. I. Clinical and public health practice. J Perinatol 1996;16:53–9.