Background
Preterm birth is one of the leading causes of infant morbidity and mortality. Although major strides have been made in identifying risk factors for preterm birth, the complexities between social and individual risk factors are not well understood.
Objective
This study examines the association between neighborhood youth violence and preterm birth.
Study Design
A 10-year live birth registry data set (2004 through 2013) from Richmond, VA, a mid-sized, racially diverse city, was analyzed (N = 27,519). Data were geocoded and merged with census tract and police report data. Gestational age at birth was classified as <32 weeks, 32-36 weeks, and term ≥37 weeks. Using police report data, youth violence rates were calculated for each census tract area and categorized into quartiles. Hierarchical models were examined fitting multilevel logistic regression models incorporating randomly distributed census tract–specific intercepts assuming a binary distribution and a logit link function.
Results
Nearly a fifth of all births occurred in areas with the highest quartiles of violence. After adjusting for maternal age, race/ethnicity, education, paternal presence, parity, adequacy of prenatal care, pregnancy complications, history of preterm birth, insurance, and tobacco, alcohol, and drug use, census tracts with the highest level of violence had 38% higher odds of very preterm births (adjusted odds ratio, 1.38; 95% confidence interval, 1.06–1.80), than census tracts with the lowest level of violence.
Conclusion
There is an association between high rate of youth violence and very preterm birth. Findings from this study may help inform future research to develop targeted interventions aimed at reducing community violence and very preterm birth in vulnerable populations.
Introduction
Preterm birth, defined as birth at <37 weeks’ completed gestation, is one of the leading causes of infant morbidity and mortality. In the United States, the prevalence of preterm birth has decreased to 9.6% in 2014–down from 10.4% in 2007. Despite changes in the rates of preterm birth and infant mortality over the past several decades, the magnitude of racial disparity has persisted. Preterm birth rates are notably higher among non-Hispanic black (13.2%) and American Indian/Alaskan Native (10.2%) women, and lower for Asian/Pacific Islander (8.5%), non-Hispanic white (8.9%), and Hispanic (9.0%) women. Reducing the rate of preterm birth and eliminating disparities is the focus of public health efforts in the United States.
The reasons for increased risk of preterm delivery among minority groups are not fully understood. Social determinants of health (eg, income, education, housing, partner support) including community-level factors (eg, neighborhood deprivation) likely play an important role in accounting for these disparities. Moreover, studies suggest preterm births are influenced by a complex array of maternal, family, community, and societal factors.
Violence has received international attention as a potential risk factor for adverse maternal and fetal outcomes. In the United States, adolescents and young adults are more likely to be victims and perpetrators of violence than any other age groups. In fact, homicide and intentional injuries are one of the leading causes of youth mortality and morbidity in the United States. Exposure to violence or other chronic stressors may influence the release of stress hormones during pregnancy that may subsequently lead to preterm birth. At the community level, neighborhood crime and violence may contribute to the widening racial disparities in adverse pregnancy outcomes. Neighborhood crime levels may be more modifiable than many other social determinants; nonetheless, few studies have closely examined the relationship between neighborhood crime levels and preterm birth. Masi et al examined the extent to which economic disadvantage, violent crime rate (murder, sexual assault, robbery, and aggravated assault), and group density were associated with pregnancy outcomes in live birth infants from Chicago, IL. After adjusting for individual characteristics, violent crime rate was strongly associated with low birthweight (<2500 g) and small-for-gestational-age infants (birthweight <10th percentile of birthweight for gestational age), but was not associated with preterm birth (<37 weeks of gestation). In another study, authors reported that as level of community violence increased, rates of health-compromising behaviors (eg, substance use), fetal death, and preterm birth increased. Conflicting reports in the literature may stem from inconsistent community violence or violent crime definitions, differing data sources, varying use of geographic areas, and uncontrolled confounding due to factors such as medical complications. Additional studies are needed to examine this association, addressing these limitations.
The objective for this study was to examine the association between neighborhood youth violence rates and preterm birth using linked data for residents of a medium-sized urban US city. Unlike previous studies that defined violence as murder, sexual assault, robbery, and aggravated assault, this analysis expanded the definition of violent crimes to include aggravated assault, kidnapping, homicide, sexual assault, robbery, theft (eg, pick pocketing, purse snatching), burglary, larceny, arson, destruction of property, and vandalism. It is possible that the lack of significance in the association between neighborhood violence and preterm birth reported in previous studies could be due to the more restrictive definitions used to define violent crimes. Further, this study examines the influence of violence not only on preterm birth (<37 weeks’ gestation) but also on very preterm birth (<32 weeks of gestation). The use of multiple years of birth data provided the power to examine preterm and very preterm births in a population setting. Moreover, this study extends findings from the existing literature by using a broader definition of violence, multiple years of police incident reports in a racially diverse geographic area where violence has prevailed for several decades.
Materials and Methods
Data for the current study came from 3 sources–live birth registry, police crime reports, and census data. A 10-year live birth registry data set (2004 through 2013) from the Virginia Department of Health Vital Statistics was used to obtain individual-level characteristics and birth outcomes for women who resided in Richmond, VA. The data included all singleton births for women residing within the city limits. Birth files were geocoded (creating coordinates based on addresses) using geographic information systems software (ArcGIS, Version 10, Environmental Systems Research Institute, Inc., Redlands, CA). Community-level data were extracted from the 2010 US Census. Crime data from the Richmond Police Department provided information on all incidents from 2004 through 2013. Data were restricted to Class-A reportable offenses involving youth aged 10-24 years. While we did not include violent offenses by all age groups, studies show that youth violence often reflect the overall violence rate in the community. In fact, in the United States, young adults and adolescents are more likely to perpetrate and be affected by violence than any other age groups.
The incidence rate of youth violence was calculated for each census tract and included in the data set. The final merged data set included information on maternal demographic characteristics, health behaviors, reproductive characteristics, and birth outcome, as well as census tract–level community characteristics and violence rates (N = 27,519). The study received approval from the Virginia Commonwealth University Institutional Review Board and Virginia Department of Health Institutional Review Board.
The outcome variable, preterm birth, was determined using the clinical estimate of gestation as recorded on the birth certificate. Gestational age at birth was classified as <32 weeks, 32-36 weeks, and term ≥37 weeks. The exposure variable, youth violence rate, was calculated using class-A reportable offenses among individuals aged 10-24 years residing in Richmond, VA. Class-A offences included aggravated assault, kidnapping, homicide, sexual assault, robbery, theft (eg, pick pocketing, purse snatching), burglary, larceny, arson, destruction of property, and vandalism. Average violence rates were calculated for each census tract area per 1000 youth population in the specified time period. Youth violence rates (events per 1000) were then categorized into quartiles. Quartile 1 (≤96.4845) reflected the lowest violence rates, followed by quartile 2 (96.4846-106.3162), quartile 3 (106.63-141.8227), and quartile 4 (>141.8227).
Covariates examined included individual- and census tract–level factors. Demographic factors included maternal age, race/ethnicity, education, and if the father was named on the birth certificate as a proxy for paternal presence. Maternal health risk behaviors included use of tobacco, alcohol, or illicit drugs during pregnancy, and were dichotomized for each risk behavior. Adequacy of prenatal care, method of payment or insurance, and reproductive health variables were also examined. Community-level factors captured from the 2010 US Census data included percentage of non-Hispanic black population, percentage of female-headed households, and percentage of individuals living <100% of the 2010 Federal Poverty Level. To examine the effect of changes in the youth violence and preterm birth rates over the years, time was assessed as a potential confounding factor. Although youth violence and preterm births were significant problems in the city, a slight decline in the rates was observed. However, examination of the association between youth violence and preterm births for each year revealed that the effect estimates remained consistent over the years. Time was not found to be a statistically significant confounding factor and was not included in the analysis.
Data were analyzed using statistical software (SAS, Version 9.4, SAS Institute, Cary, NC) for descriptive and multilevel analyses. Descriptive analyses were performed on population characteristics to provide means and percentages. Continuous variables were examined using t test and P values. Categorical variables were assessed using χ 2 test, odds ratios (OR), and 95% confidence intervals (CI). Race/ethnicity was examined as a potential effect modifier; it was not statistically significant ( P > .05), but was retained in the models as a potential confounder. Several hierarchical models were fit to assess the association between youth violence and preterm birth using PROC GLIMMIX in SAS; multilevel logistic regression models incorporated randomly distributed census tract–specific intercepts assuming a binary distribution and a logit link function. OR with 95% CI were calculated for all models. Model I examined the association between youth violence and preterm birth. Model II included youth violence and individual-level characteristics. Model III included all the variables from model II and community-level covariates. The intraclass correlation coefficient was assessed to determine the variability explained by the individual- and census-level factors. Laplace estimation was used to assess the best fit model. A significance level of α = 0.05 was used for all analyses.
Materials and Methods
Data for the current study came from 3 sources–live birth registry, police crime reports, and census data. A 10-year live birth registry data set (2004 through 2013) from the Virginia Department of Health Vital Statistics was used to obtain individual-level characteristics and birth outcomes for women who resided in Richmond, VA. The data included all singleton births for women residing within the city limits. Birth files were geocoded (creating coordinates based on addresses) using geographic information systems software (ArcGIS, Version 10, Environmental Systems Research Institute, Inc., Redlands, CA). Community-level data were extracted from the 2010 US Census. Crime data from the Richmond Police Department provided information on all incidents from 2004 through 2013. Data were restricted to Class-A reportable offenses involving youth aged 10-24 years. While we did not include violent offenses by all age groups, studies show that youth violence often reflect the overall violence rate in the community. In fact, in the United States, young adults and adolescents are more likely to perpetrate and be affected by violence than any other age groups.
The incidence rate of youth violence was calculated for each census tract and included in the data set. The final merged data set included information on maternal demographic characteristics, health behaviors, reproductive characteristics, and birth outcome, as well as census tract–level community characteristics and violence rates (N = 27,519). The study received approval from the Virginia Commonwealth University Institutional Review Board and Virginia Department of Health Institutional Review Board.
The outcome variable, preterm birth, was determined using the clinical estimate of gestation as recorded on the birth certificate. Gestational age at birth was classified as <32 weeks, 32-36 weeks, and term ≥37 weeks. The exposure variable, youth violence rate, was calculated using class-A reportable offenses among individuals aged 10-24 years residing in Richmond, VA. Class-A offences included aggravated assault, kidnapping, homicide, sexual assault, robbery, theft (eg, pick pocketing, purse snatching), burglary, larceny, arson, destruction of property, and vandalism. Average violence rates were calculated for each census tract area per 1000 youth population in the specified time period. Youth violence rates (events per 1000) were then categorized into quartiles. Quartile 1 (≤96.4845) reflected the lowest violence rates, followed by quartile 2 (96.4846-106.3162), quartile 3 (106.63-141.8227), and quartile 4 (>141.8227).
Covariates examined included individual- and census tract–level factors. Demographic factors included maternal age, race/ethnicity, education, and if the father was named on the birth certificate as a proxy for paternal presence. Maternal health risk behaviors included use of tobacco, alcohol, or illicit drugs during pregnancy, and were dichotomized for each risk behavior. Adequacy of prenatal care, method of payment or insurance, and reproductive health variables were also examined. Community-level factors captured from the 2010 US Census data included percentage of non-Hispanic black population, percentage of female-headed households, and percentage of individuals living <100% of the 2010 Federal Poverty Level. To examine the effect of changes in the youth violence and preterm birth rates over the years, time was assessed as a potential confounding factor. Although youth violence and preterm births were significant problems in the city, a slight decline in the rates was observed. However, examination of the association between youth violence and preterm births for each year revealed that the effect estimates remained consistent over the years. Time was not found to be a statistically significant confounding factor and was not included in the analysis.
Data were analyzed using statistical software (SAS, Version 9.4, SAS Institute, Cary, NC) for descriptive and multilevel analyses. Descriptive analyses were performed on population characteristics to provide means and percentages. Continuous variables were examined using t test and P values. Categorical variables were assessed using χ 2 test, odds ratios (OR), and 95% confidence intervals (CI). Race/ethnicity was examined as a potential effect modifier; it was not statistically significant ( P > .05), but was retained in the models as a potential confounder. Several hierarchical models were fit to assess the association between youth violence and preterm birth using PROC GLIMMIX in SAS; multilevel logistic regression models incorporated randomly distributed census tract–specific intercepts assuming a binary distribution and a logit link function. OR with 95% CI were calculated for all models. Model I examined the association between youth violence and preterm birth. Model II included youth violence and individual-level characteristics. Model III included all the variables from model II and community-level covariates. The intraclass correlation coefficient was assessed to determine the variability explained by the individual- and census-level factors. Laplace estimation was used to assess the best fit model. A significance level of α = 0.05 was used for all analyses.
Results
In 2004 through 2013, the overall youth violence rate in Richmond, VA, was 110.9 per 1000 youths. Approximately 17.3% of the women resided in areas with the highest quartiles of violence (quartile 4) and 34.7% lived in areas with the lowest rates of violence (quartile 1). There was a statistically significant association between youth violence rate and preterm births. Areas in the first, second, third, and fourth quartiles of violence had 9.5%, 10.1%, 11.7%, and 13.0% premature births, respectively ( P < .001) ( Table 1 ). Women residing in communities with the highest youth violence rates had greater percentages of preterm births (10.1%, 32-36 weeks’ gestation; 2.9%, <32 weeks’ gestation), followed by those from areas in the third quartile of violence (9.3%, 32-36 weeks’ gestation; 2.4%, <32 weeks’ gestation).
| Characteristics | Total, % | Quartile 1, N = 9548 | Quartile 2, N = 4937 | Quartile 3, N = 8278 | Quartile 4, N = 4756 | P value |
|---|---|---|---|---|---|---|
| Individual-level characteristics | ||||||
| Age, y | <.001 | |||||
| <19 | 7.4 | 4.7 | 7.3 | 8.9 | 10.3 | |
| 19–24 | 33.5 | 24.5 | 39.6 | 35.9 | 41.0 | |
| 25–34 | 46.5 | 53.8 | 44.0 | 43.1 | 40.5 | |
| ≥35 | 12.6 | 17.0 | 9.0 | 12.1 | 8.2 | |
| Race/ethnicity | <.001 | |||||
| Non-Hispanic white | 30.6 | 50.2 | 15.23 | 25.6 | 16.1 | |
| Non-Hispanic black | 56.4 | 35.2 | 50.3 | 70.6 | 80.8 | |
| Non-Hispanic other | 2.0 | 2.5 | 2.4 | 1.4 | 1.7 | |
| Hispanic | 10.9 | 12.2 | 32.1 | 2.4 | 1.4 | |
| Education | <.001 | |||||
| Less than high school | 24.7 | 17.8 | 33.8 | 23.6 | 31.3 | |
| High school graduate | 27.7 | 19.8 | 29.2 | 32.4 | 34.0 | |
| More than high school | 47.6 | 62.5 | 37.0 | 44.1 | 34.7 | |
| Paternal presence (no) | 58.7 | 42.2 | 65.6 | 64.7 | 74.1 | <.001 |
| Tobacco user | 7.6 | 5.3 | 5.3 | 9.3 | 12.0 | <.001 |
| Alcohol user | 0.8 | 0.8 | 0.5 | 0.8 | 0.9 | .07 |
| Illicit drug user | 1.9 | 1.1 | 0.9 | 2.7 | 3.2 | <.001 |
| Kotelchuck index | <.001 | |||||
| Inadequate/intermediate | 30.8 | 26.2 | 32.1 | 31.0 | 38.1 | |
| Adequate | 44.8 | 47.5 | 45.7 | 44.0 | 39.8 | |
| Adequate plus | 24.5 | 26.34 | 22.2 | 25.0 | 22.1 | |
| Labor/delivery complications (yes) | 19.1 | 16.5 | 15.9 | 21.3 | 23.5 | <.001 |
| Parity | <.001 | |||||
| No prior live births | 44.7 | 48.1 | 43.4 | 44.0 | 40.2 | |
| 1 Prior live birth | 28.5 | 30.0 | 28.1 | 27.9 | 27.0 | |
| 2 Live births | 14.9 | 13.4 | 16.7 | 14.7 | 16.5 | |
| ≥3 Live births | 11.9 | 8.5 | 11.8 | 13.4 | 16.2 | |
| Prior preterm birth (yes) | 0.7 | 0.8 | 0.5 | 0.6 | 0.7 | .17 |
| Insurance | <.001 | |||||
| Medicaid | 43.4 | 27.5 | 37.2 | 53.9 | 63.6 | |
| Private insurance | 43.8 | 59.9 | 33.5 | 39.5 | 29.7 | |
| Self-pay | 12.8 | 12.6 | 29.3 | 6.6 | 6.7 | |
| Preterm birth | <.001 | |||||
| Term (≥37 wk) | 89.1 | 90.5 | 89.9 | 88.4 | 87.0 | |
| 32–36 wk | 8.7 | 7.9 | 8.0 | 9.3 | 10.1 | |
| <32 wk | 2.2 | 1.6 | 2.1 | 2.4 | 2.9 | |
| Community-level characteristics, mean (SD) | ||||||
| Black, % | 56.5 (22.08) | 36.4 (25.76) | 58.6 (0.00) | 69.1 (3.14) | 72.5 (11.30) | <.0011 |
| Female-headed household, % | 41.9 (13.67) | 29.5 (14.53) | 42.4 (0.00) | 46.9 (0.97) | 57.6 (7.45) | <.0011 |
| Poverty, % | 25.3 (9.45) | 19.7 (10.08) | 20.5 (0.00) | 27.4 (5.53) | 38.1 (1.07) | <.0011 |
| Preterm rate | 10.9 (1.82) | 9.5 (2.08) | 10.1 (0.00) | 11.7 (0.17) | 13.0 (0.66) | <.0011 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
