Birth outcomes in relation to intimate partner violence

Introduction

Intimate partner violence (IPV) is a public health issue as well as a serious social problem. It is estimated that 5.3 million IPV victimizations occur each year.¹ More than1 in 4 women experience IPV during their lifetime.² Abused women are at higher risk for physical and mental health problems, including injury, chronic pain, gynecological and gastrointestinal problems, substance abuse, depression, anxiety, and posttraumatic stress disorder (PTSD).^3–7 The CDC estimate that IPV costs society $5.8 billion annually for physical and mental health care, and lost productivity.⁸

Pregnant women are particularly vulnerable to the harmful effects of IPV, because the violence may affect both maternal and neonatal health. The prevalence of IPV during pregnancy is 0.9 to 26%, depending on variant IPV definitions and study designs.^9–11 Violence during pregnancy may be more common than preeclampsia, gestational diabetes, and placenta previa.^{10, 12} It is well documented that IPV around the time of pregnancy is associated with physical and mental health problems and negative health behaviors. Studies have found an increased risk of maternal injury and death, inadequate prenatal care, smoking and alcohol use in women who experienced IPV.^10–13–16

Research on the impact of IPV on neonatal outcomes has yielded mixed results and conflicting findings may be due to variant definitions, different outcome measures, and study designs.¹⁰ In a review of 23 studies, 10 studies reported no significant differences or mixed results in birth outcomes between abused and non-abused women. The remaining 13 studies found significant differences in birth outcomes: preterm delivery, low birth weight, fetal death, miscarriage and neonatal intensive care.¹⁰ Among studies that found significant differences, the risk was 2 to 4 times greater for delivering a low birth weight infant.^{17, 18} Abused compared to non-abused women were 1.6 to 2.7 times as likely to have preterm delivery.^{19, 20} Another systematic review selected 8 studies for meta-analysis that assessed the association between IPV and low birth weight.²¹ The bivariate association was found in 6 of those studies. However, among the 6 studies, one study did not control for other confounders and 2 studies found no differences after controlling for other factors.

This study aimed to clarify the role of IPV and its association with adverse birth outcomes. We analyzed data from chart reviews of pregnant women, who were screened for IPV at the first prenatal visit and the postpartum visit. Birth outcomes of the abused pregnant women were compared with those of the non-abused pregnant women. We hypothesized that abused pregnant women would have poorer birth outcomes compared to non-abused pregnant women. Outcome measures were preterm delivery, neonatal intensive care, and low birth weight.

Materials and Methods

Study sample

The participants for this study were patients of an urban university affiliated prenatal clinic and its on-site hospital. The prenatal clinic has 12 obstetrics and gynecology faculty and residents who provide medical care to approximately 2,000 pregnant women per year. Inclusion criteria were pregnant women who were seen at the clinic, and who gave birth at the on-site hospital between January 1, 2003 and December 31, 2009. Women without documented IPV screening results, or women who did not give birth at the on-site hospital were excluded.

Procedures

All pregnant women were screened for IPV at the first prenatal visit and the postpartum visit by the providers, and the screening results recorded in the computerized medical charts. We generated a computerized list of women who were seen at the prenatal clinic between January 1, 2003 and December 31, 2009 and have available IPV screening results. This computerized list included information on the names, unique medical record number, IPV screening results, age of the women, and date of visits. If a woman gave birth to more than 1 child during the study time frame, only the most recent pregnancy was included. The unique medical record numbers were used for random selection employing a computer-generated random sequence. Random selection was stratified by victim status. We abstracted from the first prenatal visit to postpartum visits.

Prior to the start of data collection, part-time chart abstractors received online training provided by the university in computerized medical record systems and issues of confidentiality. The research team developed a chart abstraction form. Training materials included explicit criteria for all variables abstracted. One investigator (MV) conducted 10 chart reviews to pilot test the chart abstraction form and training material. The data abstractors received intensive training by an investigator. At the beginning of data collection, interrater reliability for chart abstractors was assessed for the major outcome variables. Each abstractor reviewed a sample of the same 25 charts and a Kappa statistic was calculated. The data abstraction form was modified if the Kappa statistic was less than 0.7; and the data abstractor was retrained to insure the accuracy and reliability of chart abstraction.^{22, 23} Intrarater reliability was used to accomplish ongoing monitoring of data quality. An investigator (PC) reviewed any discrepancies and corrected the data.

For birth outcomes, data were abstracted from the hospital electronic medical record (EMR) system. After the women gave birth at the on-site hospital, the newborn’s unique medical record number was recorded in the mother’s chart. The newborn also had her or his own chart, containing name, birth date, unique medical record number, mother’s name, and mother’s birth date, and other information. We used mother’s name and mother’s birth date to identify the charts of the newborns of the selected women. We abstracted birth outcomes of the infants from birth to hospital discharge.

Instruments and measures

IPV during pregnancy

Our main independent variable was IPV measured by a 4-item IPV screening tool - HITS (Hurt, Insulted, Threatened with harm, and Screamed at).²⁴ HITS has been developed for use in primary care settings, and tested with diverse populations.^24–27 HITS measures IPV in a current relationship and is comprised of the following four items: (1) “How often does your partner physically hurt you?” (2) “How often does your partner insult you or talk down to you?” (3) “How often does your partner threaten you with harm?” and (4) “How often does your partner scream or curse at you?” Answers to each question are based on a 5-point scale from never to frequently (1 to 5). Answers are summed to form an interval scale of the total HITS score from 4 to 20. Using a cutoff score of 10.5, HITS has accurately classified 91% of non-victims and 96% of female victims.²⁴ HITS has demonstrated good reliability and concurrent validity with the Conflict Tactics Scale (CTS), an established gold standard for measuring partner violence. Cronbach’s alpha was 0.80 for HITS and the correlation is 0.85 between HITS and CTS.

Strategies were used in the IPV screening protocols on the study site to minimize underreporting of IPV, including building relationships with the respondents, ensuring privacy and confidentiality, and providing the respondent with multiple opportunities for disclosure. All physicians and medical staff received training on screening techniques and the use of HITS. As most women regularly see the providers for routine exams, a relationship between providers and patients has been established. Women are seen alone in a private room where providers screen for IPV. Providers enter the information into the EMR and the computer automatically calculates the HITS score. The computer then generates a 'pop-up' warning when the patient scores above a particular threshold indicating additional investigation should occur. Physicians then refer victims to an on-site social worker for intervention. The victims were referred to IPV service advocates by the social worker, if needed. This same screening procedure was used throughout the entire time period of chart review.

Maternal and neonatal health

The main outcome variables for maternal and neonatal health were preterm delivery, low birth weight, and neonatal intensive care. Preterm delivery was defined as gestational age less than 38 weeks.^28–30 Neonatal intensive care was measured by prevalence of receiving intensive care. Based on published studies, low birth weight was defined as those who are born at <2,500g.^{16, 21, 28, 29, 31–33} Infants weighing >=2,500g at birth were considered normal weight. As the three outcome measures may be clinically related, we also created a combined birth outcome measure to indicate any of the three poor birth outcomes versus none.

Confounding variables for maternal and neonatal health

Confounding variables for maternal and neonatal health: race/ethnicity, age (<18, 18–34, >35),²⁹ education level, employment status, marital status, insurance type, living arrangements, planned pregnancy, current use of alcohol, weight at the first prenatal visit, and a comorbidity index (0,1,2,3+). The Charlson comorbidity index (CCI) is a weighted index of 19 categories of diseases that have been found to be related to mortality.^{34, 35} Charlson comorbid conditions (and their corresponding weightings) include myocardial infarction (1), congestive heart failure (1), peripheral vascular disease (1), cerebrovascular disease (1), dementia (1), chronic pulmonary disease (1), connective tissue disease (1), ulcer disease (1), mild liver disease (1), diabetes without complications (1), diabetes with complications (2), hemiplegia (2), renal disease (2), any tumor without metastases (2), leukemia (2), moderate or severe liver disease (3), metastatic solid tumor (3), and acquired immunodeficiency syndrome (6). Increasing scores on the CCI reflect an increasing burden of comorbid conditions.

Statistical analysis

We used descriptive statistics and bivariate analyses to characterize the sample based on victim status. Adverse birth outcomes by victim status were compared using Fisher’s exact test for dichotomized variables and chi-square test for categorical variables. We then assessed the association between each of the other predictor variables with each of the 3 separate outcomes: preterm delivery, low birth weight, and neonatal intensive care. We set statistical significance at the 0.05 level. SPSS, Version 21 was used to conduct the analysis.

To test the hypothesis that abused women are more likely to have adverse birth outcomes, we built a series of models for each of the 3 separate dichotomized outcomes. We analyzed each outcome in separate models. Variables significantly associated with each outcome at the p<0.05 level were retained in the model. Colinearity was checked by correlation analysis to avoid misestimating the contribution of confounders. Adjusted odds ratios and 95% confidence intervals were calculated. As in the bivariate analysis, statistical significance levels were set at 0.05.

Results and Discussion

Participants

Figure 1 presents the flow chart of the study. During the chart review period from 2008 to 2010, the health informatics department generated a list of 6423 pregnant women seen between 2003 and 2009. Approximately 25% of these charts were randomly selected for review (N=1542). Chart reviews indicated that 85 women did not meet the inclusion criteria and an additional 19 women had incomplete screening records. Therefore, 1,438 pregnant women were included in the final analysis. Of these, 7.5% were victims (N=109) and 92.5% were non-victims (1,433) based on HITS score.

Figure 1.

Flow chart of study design

Table 1 presents the demographic characteristics of participants. The victims and non-victims were similar on all but 2 variables. Victims compared to non-victims were more likely to be current users of alcohol (7.6% vs. 2.7%, p=.012). Victims were more likely to have comorbidities than non-victims (57.8% vs. 43.9%, p=.0035).

Table 1.

Sample characteristics of participants by intimate partner violence screening protocol

Variable	Victims	Non- victims	Total	P value
	N=109	N=1329	N=1,438
Race/ethnicity (%)
White	0.9	1.4	1.4	0.645
African American	54.1	50.7	50.9
Hispanic	40.4	40.3	40.3
Other	4.6	7.6	7.4
Age (mean, years)	27.1	25.9	26.0	0.070
Educational level (%)				0.703
Some high school or less	47.3	44.8	45.0
High school completed	33.8	37.7	37.5
Some college	9.5	10.9	10.8
College completed	9.5	6.5	6.7
Employed (including part-time) (%)	25.2	28.6	28.4	0.265
Marital status (%)				0.688
Single/widowed/divorced	77.1	77.5	77.5
Married	19.3	20.1	20.1
Separated	3.7	2.4	2.5
Medicare/Medicaid (%)				0.309
Commercial fee for service	0.0	0.5	0.5
Commercial HMO	1.9	4.6	4.4
Medicaid or charity care	67.6	67.6	67.6
Medicaid HMO	25.7	25.1	25.1
Medicare	4.8	2.2	2.4
Living arrangement				0.972
Participant’s house/apartment	77.5	77.1	77.1
Someone else’s house/apartment	21.6	21.5	21.5
An institution	1.0	1.3	1.3
Other	0.0	0.2	0.1
Planned pregnancy (%)	19.6	25.1	24.7	0.227
Current use of alcohol (%)	7.6	2.7	3.1	0.012
Weight at first prenatal visit (%)	163.5	160.8	161.0	0.534
Comorbidity (%)
0	42.2	56.1	55.0	0.035
1	39.4	29.6	30.3
2	11.9	10.5	10.6
>=3	6.4	3.8	4.0

Prevalence of poor birth outcomes among victims and non-victims

Table 2 shows significant differences between victims and non-victims in the prevalence of preterm delivery, low birth weight, and neonatal intensive care. Compared to non-abused women, abused women were more likely to have preterm deliveries (18.3% vs. 10.3%; p=.016). Compared to infants of non-victims, infants of victims were more likely to have low birth weight (21.5% vs. 11.0%; p=.003) and to receive neonatal intensive care (23.4% vs. 7.8%; p=.000).

Table 2.

Unadjusted prevalence of preterm deliveries, low birth weight, and neonatal intensive care by IPV victim status

Variable	Victims	Non-	Total	P value
	N=109	N=1329	N=1,438
Preterm deliveries (%)	18.3	10.3	10.9	.016
Low birth weight (%)	21.5	11.0	11.8	.003
Neonatal intensive care (%)	23.4	7.8	9.0	.000
Any of the poor birth outcomes (%)	34.3	17.2	18.5	.000

Premature babies were more likely to be low birth weight (p<.001), and low birth weight and premature babies were more likely to need intensive care (p<.001 and p<.001, respectively) (not shown). Victims were twice as likely to have any of the poor birth outcomes (34.3% vs. 17.2%, p<.001).

Multivariate analysis of preterm deliveries, low birth weight, and neonatal intensive care

Multivariate predictors of preterm deliveries, low birth weight, and neonatal intensive care are presented in Table 3. As current use of alcohol and comorbidities were significantly associated with IPV in the bivariate analysis, they were included as confounders in the multivariate analysis. After controlling for current use of alcohol and comorbidity, IPV was associated with preterm deliveries, low birth weight, and neonatal intensive care (p=.049, .007, and .000, respectively). Victims of IPV had 1.72 times the odds of preterm deliveries, 2.03 times the odds of low birth weight, and 3.33 times the odds of neonatal intensive care as non-victims. In addition, victims of IPV had 2.19 times the odds of any of the 3 poor birth outcomes as non-victims (p=.001).

Table 3.

Multivariate analysis of preterm deliveries, low birth weight, and neonatal intensive care

	Preterm deliveries (N=1394)		Low birth weight (N=1361)		Neonatal intensive care (N=1341)		Any poor birth outcomes (N=1318)
Predictor Variables	Odds Ratio (95% CI)	P value	Odds Ratio (95% CI)	P value	Odds Ratio (95% CI)	P value	Odds Ratio (95% CI)	P value
IPV
Non-victims	1.00		1.00		1.00		1.00
Victims	1.72	.049	2.03	0.007	3.33	.000	2.19	.001
	(1.00–2.95)		(1.22–3.39)		(1.99–5.58)		(1.40–3.43)

Discussion

This study showed a significant association between IPV and poor birth outcomes. In this population, approximately one fifth of victims had preterm deliveries and delivered low birth weight babies. Almost one fourth of the infants of victims received neonatal intensive care. As expected, the three outcome measures are all clinically related. Consistent with some previous studies, victims of IPV were 2–3 times more likely to have preterm deliveries, low birth weight, and neonatal intensive care.^{10, 17–20} Unlike the majority of prior studies, our study did show that the association between IPV and poor birth outcomes exists even after controlling for potential confounders.²¹ These conflicting findings may be due to variant definitions, different outcome measures, and study designs of previous studies.¹⁰

In a recent update, the United States Preventive Services Task Force (USPSTF) recommended that clinicians screen women of childbearing age for IPV and provide or refer women who screen positive for intervention services.^{36, 37} Previous studies have found that IPV is queried in less than 15% of encounters and interventions are inadequate in up to 90% of cases.³⁸ In this study, although screening and brief intervention were conducted with patients, poor birth outcomes were still found in victims of IPV. Perhaps, more education and promotion strategies are needed to overcome provider and patient barriers. Provider barriers include lack of office protocols, limited time, lack of education, lack of resources, and competing demands.^{4, 39–41} Access to care and lack of understanding of providers are barriers perceived by patients.^42–45 There is some evidence that patient education and treatment intervention leads to increased safety planning, reduced anxiety, and depression among victims of IPV.^46–48 Further research is needed to develop office-based intervention for both providers and IPV victims.

Our study has several limitations. First, patients may choose not to disclose their abuse status. Therefore, this study may underestimate the difference between the victims and non-victims in terms of outcomes. Second, the study was conducted in an urban and academic setting so the study results may not be generalized to other clinical settings. Third, this was a chart review study, which is limited by lack of information on certain variables. For instance, income was not documented in the chart. Nonetheless, employment and insurance could be considered proxies for income. Due to the large number of missing data, we did not include some confounders in the analytic sample (e.g., history of child abuse). No imputation of data was done in the analysis. However, there were no significant differences between the analytic sample and the sample with missing data.

Implications

This is one of the few randomized retrospective cohort studies to report the association of IPV to preterm deliveries, low birth weight, and neonatal intensive care, controlling for potential confounders. Our study has considerable strengths. This study would not have been possible were it not for the universal screening that was done using an EMR system. The cohort study design allows us to look at abused women over time throughout the pregnancy, and compare them with their counterparts. This study is based on a large sample of pregnant women. Thus the study has sufficient power to detect statistically significant differences between abused and non-abused women. Unlike the majority of prior studies, this study used a screening tool that demonstrates good validity and reliability. HITS measures verbal, emotional, and physical abuse, thus providing information on various aspects of IPV.

Healthy People 2020 has specified the goals and objectives that address improved maternal and child health, including reduction in low birth weight.⁴⁹ Besides poor birth outcomes, victims of IPV and their children may suffer long-term negative health effects. It is crucial to establish a standardized comprehensive intervention that promotes birth outcomes for victims of IPV.^{4, 45}