Background Adverse pregnancy outcomes are associated with higher cardiovascular disease risk among mothers and future health problems of offspring. Neighbourhood crime may contribute to adverse pregnancy outcomes by increasing chronic stress, yet the association has been relatively understudied.
Methods Electronic health records from 34 383 singleton births at a single hospital in Chicago (2009–2013) were geocoded and linked to 1-year rates of police-recorded crime at the neighbourhood (Chicago community area) level. Crimes included homicide, assault/battery, criminal offences and incivilities. Cross-sectional associations of total neighbourhood crime rates with hypertensive disease of pregnancy (HDP: pre-eclampsia/gestational hypertension), preterm birth (PTB), spontaneous preterm birth (sPTB) and small-for-gestational-age (SGA) birth were assessed using multilevel logistic regression with community-area random intercepts. Models controlled for maternal and infant characteristics and neighbourhood poverty. We then assessed associations between individual crime categories and all outcomes.
Results Total neighbourhood crime rates ranged from 11.6 to 303.5 incidents per 1000 persons per year (mean: 61.5, SD: 40.3). A 1-SD higher total neighbourhood crime rate was associated with higher odds of HDP (OR: 1.06, 95% CI 1.00 to 1.13), PTB (OR: 1.09, 95% CI 1.03 to 1.15), sPTB (OR: 1.09, 95% CI 1.03 to 1.16) and SGA (OR: 1.05, 95% CI 1.01 to 1.10) in fully adjusted models. Associations were generally consistent across crime categories, although only assault/battery and incivilities were associated with HDP.
Conclusions Higher neighbourhood crime rates were associated with small but significant increases in the odds of adverse pregnancy outcomes. Interventions that cultivate safer neighbourhoods may be a promising approach for improving pregnancy outcomes.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Adverse pregnancy outcomes such as preterm birth (PTB) and low birth weight put infants at risk for a variety of health conditions throughout the lifespan.1–4 In addition, adverse pregnancy outcomes have been associated with future cardiovascular disease among women.5–8 Large socioeconomic and racial/ethnic disparities in adverse pregnancy outcomes are a persistent public health problem9 10 and may result from disadvantage occurring at the individual, family or neighbourhood level.9 The neighbourhood environment has emerged as an important area for investigation that may contribute to these disparities.11 12
Prior research on pregnancy outcomes has largely focused on neighbourhood socioeconomic status, incorporating factors like neighbourhood poverty, unemployment, income and education.9 11–13 However, residents of socioeconomically disadvantaged neighbourhoods also experience greater rates of neighbourhood crime, and crime has been suggested as a more proximal exposure than economic advantage in the association of neighbourhood characteristics with adverse pregnancy outcomes.14 15 Exposure to crime has been found to increase psychological distress16 and may contribute to chronic stress.17 Chronic stress may lead to adverse pregnancy outcomes both directly through hormonal and neuroendocrine changes that may trigger PTB or restrict fetal growth18 19 and indirectly thorough health behaviours. Pregnant women who experience stress due to neighbourhood crime may use unhealthy behaviours, such as smoking and drinking alcohol, as a coping mechanism20 21 and may be less likely to participate in healthy behaviours like physical activity22 that may reduce the risk of developing adverse outcomes.23
To date, relatively few studies have examined associations of neighbourhood crime with adverse pregnancy outcomes in comparison to other neighbourhood exposures like neighbourhood socioeconomic status. Also, prior studies that have examined neighbourhood crime have not examined associations of with hypertensive disease of pregnancy (HDP).14 21 24–27 In addition, prior studies of crime and pregnancy outcomes have largely used birth certificate data, which may under-report pregnancy complications or previous maternal medical conditions.28–30 Electronic health records (EHRs) are a promising data source that may better capture these outcomes and potential confounders. As such, our objective was to examine associations of neighbourhood crime with HDP, PTB and small-for-gestational-age (SGA) birth among a racially and ethnically diverse cohort of women in Chicago using EHR data.
Our study sample came from the Northwestern Medicine Enterprise Data Warehouse, an EHR repository. We extracted records for all women delivering singleton infants at Northwestern University’s Prentice Women’s Hospital in Chicago, Illinois, from 1 January 2009 to 31 December 2013 (n=60 826 births among 46 690 women). We included records from all women whose addresses during pregnancy could be successfully geocoded and who resided within the Chicago city limits in order to enable linkage of neighbourhood crime rates.
Patient addresses were geocoded in ArcMap V.10.5 (Environmental Systems Research Institute, Redlands, California, USA). Residential addresses for a total of 57 866 births (95%) were successfully geocoded, and 40 798 were within the Chicago city limits. We subsequently excluded 514 births (1%) with missing/implausible demographic or clinical information and restricted to each woman’s first birth during the study period (5901 later births excluded), for a final total of 34 383 included births. Excluded births were more likely to be among white women and those with private insurance and were less likely to have adverse pregnancy outcomes (online supplementary table 1). As only first births during the study period were included, excluded births were also more likely to be among older and multiparous women.
Supplementary file 1
Adverse pregnancy outcomes were defined using data extracted from EHRs, including gestational age, birth weight and problem lists/diagnostic codes. Outcomes included HDP, PTB and SGA birth. HDP was defined as either gestational hypertension or preeclampsia during the current pregnancy; the field for this outcome had no missing data. PTB, defined as gestational age at birth of less than 37 weeks, was missing for 149 births (<1%). We examined both any PTB and spontaneous preterm birth (sPTB) to assess whether associations varied between sPTB and medically indicated PTB. Finally, SGA was defined as birth weight below the 10th percentile for a given gestational age31 and was missing for 284 births (<1%). A prior study comparing EHR extraction to manual abstraction found differences in abstracted birth outcomes to be small.32
Neighbourhood crime exposure
Crime data came from the City of Chicago’s Data Portal,33 which published a database of all police-recorded crimes occurring within the Chicago city limits. We obtained crime data for years 2008–2013. Crimes were categorised using Illinois Uniform Crime Reporting codes according to a previously published classification scheme.22 34 Categories included: homicide, assault and battery, criminal offences (eg, robbery, sexual assault, arson, kidnapping) and incivilities35 (non-violent crimes that may be indicators of neighbourhood physical and social disorder, eg, narcotics, prostitution, vandalism, weapons violations). We excluded crimes that occurred on an airplane/in an airport as these crimes are unlikely to affect perceptions of neighbourhood safety.
In this study, neighbourhood crime was operationalised at the community-area level. Chicago is divided into 77 community areas, which are well-defined, static neighbourhoods designated by the Social Science Research Committee at the University of Chicago and officially recognised by the city of Chicago. Prior work suggests that crime operates as an area-level rather than individual-level exposure26; as such, capturing crime exposures at a larger geographic level such as community area may be more relevant than capturing according to individual buffers (eg, crime within a 1 mile radius of residential address).
Crime exposures were linked to women by community area. For each birth, we created a count of the number of crimes (total, and for each crime category) in the mother’s community area of residence within a 12-month period preceding the date of delivery. We then calculated population-normalised crime rates. The numerator for these rates was the count of crimes in that community area during the 12-month period and the denominator was the total population of the community area based on the 2010 US Census. This rate was then multiplied by 1000 to reflect rates per 1000 persons. Crime rates were then standardised by subtracting the mean crime rate across all patients and dividing by the SD to enable comparison of associations across different crime categories. We assessed sensitivity to choice of neighbourhood definition by alternatively examining crime rates at the census tract level.
Patient covariates were extracted from EHRs and included potential confounders of the crime-adverse pregnancy outcomes relationship selected a priori. Maternal characteristics included age at delivery, race/ethnicity (Black/African American, White, Hispanic/Latino, Asian/Pacific Islander, other race, unknown), insurance status (public, private, none), multiparity (multiparous vs not), prevalent diagnosis of hypertension or diabetes, gestational diabetes during current pregnancy, history of asthma, history of mental health issue, history of a sexually transmitted disease and history of substance abuse (including alcohol and tobacco use during current pregnancy). Infant sex was also recorded. Neighbourhood poverty was defined as per cent of population of the census tract in which a woman resided that had household incomes below the poverty line based on data from the 2009–2013 American Community Survey.
Distributions of birth outcomes, neighbourhood crime rates and covariates were examined in the study sample using means and SD for continuous variables and proportions for categorical variables. We compared the distribution of covariates and outcomes among women by tertile of total neighbourhood crime rate. We then used multilevel logistic regression with community area random intercepts to calculate adjusted ORs of the three adverse pregnancy outcomes associated with a 1-SD increase in neighbourhood crime rates. First, we ran an unconditional model to estimate the interclass correlation coefficient or the proportion of the variance in each adverse pregnancy outcome attributable to differences across community areas. Then, we ran the model with the neighbourhood crime term and individual-level (level 1) and neighbourhood-level (level 2) covariates. Models controlled for maternal age at delivery, race/ethnicity, insurance status, multiparity, prevalent hypertension, prevalent diabetes, history of asthma, mental health conditions, sexually transmitted disease or substance abuse, gestational diabetes in current pregnancy, infant sex and neighbourhood poverty. We examined associations of total neighbourhood crime rate and then examined associations of rates of the four crime categories (homicide, assault/battery, criminal offences, incivilities) to see if associations differed by crime type.
We conducted several sensitivity analyses. As some studies have found ambient temperature or season to be associated with adverse pregnancy outcomes,36 we conducted a sensitivity analysis adjusting all models for season of delivery. In addition, we tested sensitivity to fixed-cohort bias, a type of selection bias occurring in retrospective studies of pregnancy outcomes in which shorter pregnancies are missed early in the study and longer pregnancies are missed at the end of the study.37 As such, we excluded participants with an estimated date of conception at least 20 weeks before the study start (1 January 2009) or within 43 weeks of the study end (31 December 31 2013) to make the distribution of gestational age consistent across calendar time (n=3254 births excluded).
During the study period, there were 34 383 singleton births. The number of births per community area ranged from 7 to 3461 (median 163). Table 1 displays the distribution of past-year neighbourhood crime rates in the community areas patients resided in. Past-year total crime rates ranged from 11.6 to 303.5 per 1000 persons (mean: 61.5, SD: 40.3, table 1). The mean rates per 1000 persons for specific crime categories were: homicide: 0.1 (SD: 0.1); assault/battery: 23.7 (SD: 16.5); criminal offences: 14.3 (SD: 7.4) and incivilities: 23.4 (SD: 17.9). The distributions of neighbourhood crime rates by adverse pregnancy outcome status are presented in online supplementary table 2.
Available demographic and clinical characteristics of the study population are shown in table 2. The mean age at delivery was 31.1 years. Half (49.6%) of women were White, 18.0% were Hispanic/Latina, 11.0% were Black/African American, 6.8% were Asian/Pacific Islander, 4.7% were classified as other race and race/ethnicity was unknown for 9.9%. Most (76.6%) had private insurance and 35.7% were multiparous. The average neighbourhood poverty level was 17.3%. The prevalence of adverse pregnancy outcomes was 5.1% for HDP, 8.1% for PTB (5.5% for sPTB) and 10.5% for SGA.
We compared demographic and clinical characteristics among patients in each tertile of total neighbourhood crime rate (table 2). Women in highest crime neighbourhoods were slightly younger on average, were more likely to be Black/African American, have public insurance, be multiparous, have a recorded history of asthma, a sexually transmitted disease or substance abuse, have prevalent hypertension and live in census tracts with a higher per cent of population below the poverty line. Women in the highest crime neighbourhoods were less likely to have a recorded history of a mental health condition. Women in the middle tertile were more likely to have gestational diabetes in the current pregnancy and more likely to be Hispanic/Latina than those in the other tertiles. The proportion with adverse pregnancy outcomes increased with increasing tertile of neighbourhood crime (HDP from 4.3% to 5.8%, PTB from 7.4% to 9.6% and SGA from 9.4% to 11.5%, table 2). Table 3 displays bivariate associations of demographic and clinical characteristics with each adverse pregnancy outcome.
In the unconditional models, intraclass correlation coefficients for the adverse pregnancy outcomes ranged from 3% to 5%. In multivariable models, a 1-SD increase in total neighbourhood crime rate was associated with a 6% increase in the odds of HDP (OR: 1.06, 95% CI 1.00 to 1.13), a 9% increase in the odds of PTB (OR: 1.09, 95% CI 1.03 to 1.15) and of sPTB (OR: 1.09, 95% CI 1.03 to 1.16) and a 5% increase in the odds of SGA (OR: 1.05, 95% CI 1.01 to 1.10) (table 4). When we examined associations for specific types of crimes, we found that patterns were generally consistent across crime categories (table 4), although incivilities had the largest association for three out of four outcomes (HDP, PTB and sPTB). For HDP, the magnitude of association was small and CIs crossed the null for homicide and criminal offences. Results were similar in sensitivity analyses when an alternative neighbourhood definition (census tract) was used, although slightly attenuated for HDP and SGA (online supplementary table 3). Results were also similar on adjustment for season of delivery (online supplementary table 4) and were slightly stronger for PTB and HDP when we assessed sensitivity to fixed-cohort bias by excluding participants with dates of conception at least 20 weeks prior to the study start and within 43 weeks of the study end (online supplementary table 5).
Higher neighbourhood crime rates were associated with significantly higher odds of HDP, PTB (any or spontaneous) and SGA among a large cohort of women in Chicago after adjustment for available maternal and infant demographics, clinical characteristics and neighbourhood poverty. When we examined associations of specific categories of crimes (homicide, assault/battery, criminal offences and incivilities), patterns were generally consistent across crime categories, although only assault/battery and incivilities were associated with HDP.
Our findings are consistent with several prior studies that found higher crime rates to be associated with higher risk of PTB21 26 27 and low birth weight or SGA.14 15 21 24–26 While there was substantial heterogeneity in how crime was parameterised in prior studies, the effect sizes in our study were generally consistent with prior findings. For example, Morenoff et al found higher violent crime rates in Chicago to be associated with a 5% increase in odds of low birth weight,15 and Masi et al found similar associations for SGA (OR 1.1–1.2 across race/ethnic groups on adjustment for neighbourhood economic disadvantage, also in Chicago).14 Larger effect sizes were seen in several studies in North Carolina26 27 that compared highest to lowest quartiles of violent crime rate. For example, women in neighbourhoods in the two highest quartiles of violent crime had 1.5 times the odds of PTB (95% CI 0.9 to 2.6) for white women and 1.4 times the odds of PTB (95% CI 1.0, 2.1) for black women compared with those living in lower crime neighbourhoods.27 Prior studies were not able to adjust for maternal conditions such as prevalent hypertension, diabetes or asthma that might be risk factors for adverse pregnancy outcomes due to lack of availability of this information in birth records.
The mechanisms through which neighbourhood crime influences development of adverse pregnancy outcomes are not explicitly known, but may operate through multiple pathways. Crime may influence pregnancy outcomes through a direct biological pathway in which crime increases maternal physiological distress/chronic stress,16 17 leading to hormonal and neuroendocrine changes such as higher levels of corticotropin-releasing hormone and cortisol that may trigger sPTB or fetal growth restriction.18 19 In addition, neighbourhood crime may make women less likely to engage in physical activity22 or walking in their neighbourhoods.34 As obesity is an important risk factor for HDP,38 crime may influence this outcome and, subsequently, downstream birth outcomes (PTB and SGA), through health behaviours. Crime may also lead women to adopt unhealthy behaviours such as smoking and drinking alcohol during pregnancy as a coping mechanism.20 21
The positive association between neighbourhood crime rates and HDP is a novel finding. While researchers have previously found that neighbourhood physical degradation/disorder like litter, graffiti and poor housing conditions was associated with HDP39 40 and hypertension was previously included in an aggregate measure of pregnancy complications,41 to our knowledge, the association of neighbourhood crime with HDP alone has not been examined. HDP increases the risk of SGA and PTB, but is also an important outcome on its own given its links to future cardiovascular disease among affected women.5 6 8
Our finding that only incivilities and assault/battery were significantly associated with HDP may be a result of the relatively higher rates of these crime categories compared with homicide and criminal offences. However, it is also possible that incivilities, which includes crimes such as vandalism and prostitution that are related to physical and social disorder in the neighbourhood,35 is a particularly salient neighbourhood feature for adverse pregnancy outcomes, as the strongest association was observed for this crime category for three out of four outcomes. Proliferation of incivilities causes residents of a neighbourhood to perceive greater problems and to lose confidence in their neighbourhood and in police ability to prevent or control crime.35 It is possible that incivilities may be particularly relevant for determining how individuals perceive the safety of their neighbourhood, as physical and social disorder may be more readily apparent on a daily basis than less frequent crimes like homicide. Prior research has shown physical disorder to be associated with HDP.39 40 In addition, incivilities-related crime rates have been associated with lower odds of walking for transportation.22 As physical activity may reduce the risk of developing HDP,23 incivilities may be particularly important for this outcome relative to more serious crimes like homicide.
Our use of EHR data is a strength of this study, as birth records may not reliably capture HDP as well as several covariates for which we controlled. In addition, although our study population included women who sought care at a single academic hospital, we included a large, racially/ethnically diverse group of women from all 77 Chicago community areas, which may increase the generalisability of our findings. Our study was also subject to several limitations. First, we relied on police-recorded crime, which likely does not capture all crime incidents. Crimes may be differentially reported by neighbourhood due to differential police responses, which could lead to measurement error. In addition, we focused on objectively measured crime only and were unable to determine whether women perceived their neighbourhoods to be safe or unsafe. In addition, while EHR data may provide improved capture of pregnancy complications and medical history relative to birth records, it still may not have perfect capture of all data as these systems were designed for clinical practice rather than research. We lacked information on health behaviours like diet and physical activity that are likely mediators of the crime-pregnancy outcome relationship, as these behaviours were not routinely captured in the EHR. In addition, prepregnancy body mass index was not captured consistently in the EHR, and gestational weight gain was missing for approximately 49% of our study population. Also, implausible/impossible values entered into the EHR for gestational age and birth weight resulted in missing data for PTB and SGA. In addition, we controlled for reported history of substance abuse, but this variable may have been misreported due to social desirability or recall bias and did not distinguish between use during the current pregnancy and prior use. In addition, we lacked information on residential mobility and date of diagnosis of HDP. Finally, community areas may not accurately capture each woman’s definition of her ‘neighbourhood.’ However, results were generally consistent in a sensitivity analyses using census tract as an alternative neighbourhood definition.
Higher neighbourhood crime rates were associated with small but significant increases in the odds of four adverse pregnancy outcomes: HDP, PTB, sPTB and SGA. Interventions that promote safer neighbourhoods and crime prevention may be a promising approach for reducing rates of adverse pregnancy outcomes among women living in urban areas.
What is already known on this subject
Neighbourhood exposures such as low socioeconomic status have been linked to adverse pregnancy outcomes. Neighbourhood crime rates may contribute to these adverse outcomes by increasing mothers’ levels of chronic stress, yet relatively few studies have examined these associations, particularly for the outcome of hypertensive disease of pregnancy.
What this study adds
Using electronic health record data, we found neighbourhood crime rates to be associated with higher odds of hypertensive disease of pregnancy, preterm birth (overall and spontaneous) and small-for-gestational-age birth.
The authors thank Kelsey Ryland from Northwestern University for assistance with geocoding.
Contributors SLM designed the study, analysed the data, interpreted results and drafted the article. LRP contributed to data acquisition, interpretation of results and critically revised the article. WAG contributed to interpretation of results and critically revised the article. KNK contributed to study design, data acquisition, interpretation of results and critically revised the article. All authors approved the final version for publication.
Funding This research was supported by an NHLBI training grant in cardiovascular epidemiology and prevention (award number T32HL069771).
Competing interests None declared.
Patient consent Detail has been removed from this case description/these case descriptions to ensure anonymity. The editors and reviewers have seen the detailed information available and are satisfied that the information backs up the case.
Ethics approval Northwestern University IRB.
Provenance and peer review Not commissioned; externally peer reviewed.