Volume 93, Issue 8 p. 727-740
Review Article
Free Access

Deprived neighborhoods and adverse perinatal outcome: a systematic review and meta-analysis

Amber A. Vos

Corresponding Author

Amber A. Vos

Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands

Correspondence

Amber A. Vos, Erasmus University Medical Center, Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, P.O. Box 2040, 3000 CA Rotterdam, the Netherlands.

E-mail: [email protected]

Search for more papers by this author
Anke G. Posthumus

Anke G. Posthumus

Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands

Search for more papers by this author
Gouke J. Bonsel

Gouke J. Bonsel

Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands

Rotterdam Midwifery Academy, Rotterdam, The Netherlands

Department of Public Health, Erasmus University Medical Center, Rotterdam, the Netherlands

Search for more papers by this author
Eric A.P. Steegers

Eric A.P. Steegers

Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands

Search for more papers by this author
Semiha Denktaş

Semiha Denktaş

Department of Obstetrics and Gynecology, Division of Obstetrics and Prenatal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands

Search for more papers by this author
First published: 17 May 2014
Citations: 125
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

Abstract

Objectives

This study aims to summarize evidence on the relation between neighborhood deprivation and the risks for preterm birth, small-for-gestational age, and stillbirth.

Design

The design was a systematic review and meta-analysis.

Main outcome measures

The main outcome measures included studies that directly compared the risk of living in the most deprived neighborhood quintile with least deprived quintile for at least one perinatal outcome of interest (preterm delivery, small-for-gestational age and stillbirth).

Methods

Study selection was based on a search of Medline, Embase and Web of Science for articles published up to April 2012, reference list screening, and email contact with authors. Data on study characteristics, outcome measures, and quality were extracted by two independent investigators. Random-effects meta-analysis was performed to estimate unadjusted and adjusted summary odds ratios with the associated 95% confidence intervals.

Results

We identified 2863 articles, of which 24 were included in a systematic review. A meta-analysis (= 7 studies, including 2 579 032 pregnancies) assessed the risk of adverse perinatal outcomes by comparing the most deprived neighborhood quintile with the least deprived quintile. Compared with the least deprived quintile, odds ratios for adverse perinatal outcomes in the most deprived neighborhood quintile were significantly increased for preterm delivery (odds ratio 1.23, 95% confidence interval 1.18–1.28), small-for-gestational age (odds ratio 1.31, 95% confidence interval 1.28–1.34), and stillbirth (odds ratio 1.33, 95% confidence interval 1.21–1.45).

Conclusions

Living in a deprived neighborhood is associated with preterm birth, small-for-gestational age and stillbirth.

Abbreviations

  • CI
  • confidence interval
  • LBW
  • low birthweight
  • OR
  • odds ratio
  • SGA
  • small-for-gestational age
  • Key Message

    Living in a deprived neighborhood is associated with adverse perinatal outcomes. Our findings suggest that greater awareness of neighborhood deprivation should be advocated among obstetric healthcare professionals. Further studies need to identify effective interventions.

    Introduction

    The association between socio-economic status and health has been recognized for a long time 1. Socio-economic inequalities are associated with a decrease in life expectancy of up to 10 years in developed countries 2. Perinatal mortality is a key indicator for population socio-economic inequalities 3. Results from large cohort studies have shown that lifestyle risk factors (such as smoking) and social deprivation (such as household income, education level and poverty) are strongly related to adverse perinatal outcomes such as preterm birth, low birthweight (LBW) and fetal growth restriction 4-6. Over 75% of all causes of perinatal mortality are preceded by preterm delivery and fetal growth restriction, which are conditions with different prevalences in different socio-economic groups 7-9. Preterm birth and fetal growth restriction are usually a result of the concurrence of a large number of individual risk factors, known as risk accumulation 10, 11. Risk accumulation comprises the large influence of a number of smaller, seemingly less important risk factors, on the risk of adverse outcomes 10-12. Such accumulation of risk factors is more common in deprived neighborhoods, which generally show poor perinatal outcomes 11.

    The term “neighborhood deprivation” does not have a standard definition. Rajaratnam et al. 13 examined which neighborhood characteristics are routinely addressed in perinatal health studies. That study identified 12 broad categories of factors used to characterize neighborhoods, such as education, employment, occupation and income 13. Deprivation is often characterized by indexes with cut-off points to categorize deprivation levels. The most commonly used indexes are the Carstairs–Morris score, Index of Multiple Deprivation, Townsend deprivation index, and the Jarman score. The Carstairs–Morris score measures domestic overcrowding, male employment, car ownership and social class distribution 14. The Index of Multiple Deprivation combines income, employment status, health and disability, education, housing problems and crime 15. The Townsend Deprivation Index converts zip codes into deprivation scores by taking into account local unemployment, car ownership, overcrowding and housing 16. The Jarman score is a continuous measure which combines unemployment, overcrowding, single parents, under-fives, elderly living alone, ethnicity, low social class and residential mobility 17. However, often other self-composed indices are used.

    To date it is unknown to what extent the effect of deprivation goes beyond the effect of poor individual level of socio-economic status of citizens in deprived neighborhoods 18, 19. An additional effect of neighborhood has been demonstrated in relation to diseases in adulthood 20, and was suggested to occur with regard to adverse perinatal outcomes 3, 21, 22. For instance, a recent meta-analysis demonstrated such an additional role in the occurrence of LBW, which combines both growth restriction and preterm cases 23. The newest systematic review takes the heterogeneous evidence into account by including studies with neutral, possible and positive associations of deprivation with perinatal outcomes 24. Here we present a systematic review and meta-analysis complementary to the existing LBW analysis for other perinatal outcomes, i.e. preterm birth and fetal growth restriction separately, and stillbirth. It seems timely to show the broader evidence of the association of neighborhood deprivation with perinatal outcomes other than LBW.

    Material and methods

    Sources

    We performed an electronic search on 1 May 2012 in Medline, Embase and Web of Science from inception to May 2012 for meta-analysis, randomized controlled trials, cohort studies, longitudinal studies and case-control studies. A search strategy was developed and adapted for each database. It included search terms regarding adverse perinatal outcome (such as “stillbirth” “fetal death*”, “fetal mortalit*”, “adverse pregnancy outcome*”, “small for gestational age*”, “low birth weight*”, “dysmatur*”, “intrauterine growth restrict*”, “preterm deliver*”, “preterm birth*”, “prematur*”), and search terms regarding deprivation (including “neighborhood*”, “neighbourhood*”, “urban*”, “city”, “town*”, “disadvantag*”, “deprived”, “pover*”, “indigen*”, “disadvantaged communit*”, “residential segregation”). The search terms regarding adverse perinatal outcome were restricted to “prematurity”, “small-for-gestational age” or “intra-uterine growth retardation” or “fetal growth retardation”, “stillbirth” and “perinatal mortality.” Because the search term “congenital anomaly” resulted in too much heterogeneity in the results (for both underlying cause as well as type of congenital anomaly), this search term was not used in the present study. Reference lists from main articles and relevant reviews were hand-searched for additional eligible studies. The search was restricted to studies in humans. No language restrictions were applied. Ethical approval was not required in the Netherlands.

    Study selection

    For inclusion the studies had to meet the following criteria: They had to (i) be a randomized controlled trial, cohort (including longitudinal), cross-sectional or case-control study; (ii) report how deprived neighborhood or neighborhood index was defined; (iii) report data of perinatal outcomes on the whole neighborhood population; (iv) report any of the main outcome measures preterm birth, LBW, small-for-gestational age (SGA), stillbirth, and/or perinatal mortality; (v) report prevalences, odds ratios (OR) or relative risks; (vi) be conducted in a developed country, defined as all countries listed as such by the World Bank 25.

    Two reviewers (A.V., A.P.) independently examined titles, abstracts and full-text articles for eligibility. They independently extracted all relevant data into a pre-formatted spreadsheet. In case of discrepancies or uncertainties regarding the data extraction, the two reviewers achieved consensus between themselves or by approaching a third party (the senior investigator, SD). In the case of missing tabular data in studies deemed eligible for meta-analysis, we contacted authors of the respective study 26, 27. The procedures we followed were in accordance with the PRISMA statement 28.

    Two reviewers (A.V., A.P.) assessed the quality of each included study independently by using the Newcastle–Ottawa scale 29. Since no randomized controlled trials were retrieved from the search, we used a quality assessment scale suitable for observational studies. The Newcastle–Ottawa scale was developed to assess the quality of non-randomized studies with regard to design, content and ease of use directed to the task of incorporating the quality assessments in the interpretation of meta-analytic results. The scale was scientifically evaluated and is regarded as suitable to use for quality assessment of observational studies 30. We defined study quality as “high” if the study was assigned the maximum of nine stars on this scale, “medium” in the case of seven or eight stars, and “low” in the case of seven or fewer stars. Discrepancy in quality assessment was resolved by the two reviewers.

    As LBW could not be distinguished between LBW related to preterm birth and LBW babies at term, we decided only to analyze results from studies when growth restriction was explicitly defined as SGA. Studies that met the inclusion criteria, but in which the reported determinant or outcome measure was not eligible for meta-analysis (such as not using quintiles as cut-off point or LBW), remained eligible for the systematic review but were not incorporated in the meta-analysis. Statistical analysis was performed with BIOSTAT COMPREHENSIVE META-ANALYSIS version 2 (Biostat, Englewood, NJ, USA). Higgins I2 (with a significance level at < 0.05) and tau2 were calculated to assess statistical heterogeneity across studies. We used random-effects meta-analysis to estimate unadjusted and adjusted summary ORs with the associated 95% confidence intervals (CI). Adjusted ORs were obtained from fully adjusted models as presented in the original articles. If the risk estimate of interest was not explicitly stated, ORs were calculated with comprehensive meta-analysis. Because different indices were used to determine deprivation, a subgroup moderator analysis was performed to explore the effect of these different neighborhood indices on the outcome of interest.

    Assessment of deprivation

    Apparently different cut-off points were used across studies in the categorizing of neighborhood deprivation. We opted for categorization into quintiles (with the lowest quintile representing the least deprived neighborhoods) for three reasons. The most important reason was that the majority of studies applied a division into quintiles, assuming the extreme categories to be large enough to be relevant and small enough to demonstrate contrasts if present. Another reason was that we considered quintiles valuable for the cross-study comparability. We assume that the relative socio-economic position within a country is much more important as a determinant than are absolute measures. Thirdly, the key indicators of interest are summary risk estimates, which themselves are relative indicators as well, because baseline risk is set to one or zero. We asked 21 authors of papers without quintile divisions to re-categorize their determinants into quintiles, and subsequently re-analyze their study data. One author was willing and able to do so 27. In the meta-analysis, we evaluated the contrast between least and most deprived neighborhood quintiles in preterm birth (birth before 37th week of gestation, SGA (birthweight below the 10th percentile for gestational age), and stillbirth (≥20 weeks).

    Results

    The initial search identified 2863 articles, which were potentially relevant based on title/abstract screening. All studies were in English except for one study, which was in Dutch. After exclusion of studies that did not meet our inclusion criteria, 108 articles remained, of which the full texts were evaluated. We identified another four studies from reference list tracking, of which two met inclusion criteria. In total, 24 studies met the inclusion criteria for the systematic review and seven were included in the meta-analysis (Figure 1).

    Details are in the caption following the image
    Flow diagram of included studies in the systematic review and meta-analysis.

    Table 1 11, 21, 22, 26, 27, 31-49 summarizes the characteristics of the studies included in the systematic review (n = 17) and meta-analysis (n = 7). Many articles presented results for multiple outcomes. The included studies were conducted in either the UK (= 10), Canada (= 5), the Netherlands (n = 4), United States (n = 2), Spain (n = 1), Sweden (n = 1) and Australia (n = 1), with data collected from 1985 up to 2008. Four of the included studies included a multilevel analysis 21, 27, 31, 41, of which one study was included in the meta-analysis 27. The remaining 20 studies assessed neighborhood-level exposure.

    Table 1. Design, characteristics and quality assessment of included studies ordered by publication year (= 24).
    Study Study design and participants Index for deprivation Primary outcome Quality
    Spencer et al. 38 Retrospective cohort study, 1991–1993, West Midlands, UK, = 194 081 Townsend deprivation index – deciles; Register General's social class index – quintiles LBW, VLBW Medium
    Smeeton et al. 44 Case control study, 1996–1998, London, UK, = 2735 Jarman score – continuous Stillbirth, early neonatal death Medium
    Luo et al. 49 Birth cohort study, 1985–2000, British Colombia, Canada, = 697 477 Neighborhood income – quintiles PTB, SGA, stillbirth High
    Manning et al. 35 Retrospective review of neonatal unit admission records, 1990–2002, UK, = 47 614 Townsend deprivation index – quartiles Admission at neonatal unit Low
    Dibben et al. 32 Retrospective birth cohort study, 1996–2000, UK, = 306 067 Index of multiple deprivation – quintiles LBW, VLBW High
    Delpisheh et al. 42 Retrospective analysis, 1993, 1998 and 2001, UK, = 4637 Townsend deprivation index – (−6 to 12) Smoking, birthweight Low
    Janghorbani et al. 34 Prospective case-record study, Plymouth, UK, 1996–1997, = 3834 Townsend deprivation index – tertiles PTB High
    Collingwood Bakeo and Clarke 40 Retrospective cohort study, 1991–2000, England and Wales, UK, = 116 261 Carstairs – quintiles LBW High
    Luo et al. 47 Birth cohort study, 1991–2000, Quebec, Canada, = 825 349 Neighborhood income – quintiles PTB, SGA, stillbirth High
    Cubbin et al. 41 Case record study, 1997–1998, Washington and Florida, USA, = 8359 Townsend deprivation index – tertiles LBW High
    Urquia et al. 48 Retrospective cohort study, 1996–2001, Toronto, Canada, = 143 030 Neighborhood income – quintiles PTB High
    De Graaf et al. 22 Retrospective birth cohort study, 2002–2006, the Netherlands, = 877 816 Dutch deprivation score – binary PM Medium
    Gray et al. 45 Retrospective cohort study, 2000–2003, UK, = 149 690 Carstairs–Morris scores – quintiles PTB Medium
    Beard et al. 31 Retrospective cohort study, 1994–2004, Australia, = 877 951 Index of relative socio-economic disadvantage – quartile SGA < p3 High
    Agyemang et al. 21 Prospective cohort study, 2003–2004, Amsterdam, the Netherlands, = 7883 Neighborhood income – quartile PTB, SGA Medium
    Smith et al. 37 Prospective cohort study, 1998–2007, UK, = 7402 Composite neighborhood index – quintiles Very preterm birth (<33 weeks) Medium
    Janevic et al. 43 Case record study, 1998–2002, New York, USA, = 517 994 Messer neighborhood deprivation index c quartile Term LBW, PTB High
    Liu et al. 46 Retrospective birth cohort study, 2004–2006, Canada, = 334 231 Neighborhood income Prevalences and risk estimates of the included studies ordered by publication year (= 24) quintiles PTB, SGA, stillbirth High
    Timmermans et al. 11 Retrospective birth cohort study, 2002–2006, Rotterdam, the Netherlands, = 8668 Dutch deprivation score – binary PTB, SGA Medium
    Poeran et al. 36 Retrospective birth cohort study, 2000–2006, Rotterdam, the Netherlands, = 50 000 Dutch deprivation score – top 5 highest deprived neighborhoods vs. top 5 lowest PTB, SGA, PM Medium
    Sundquist et al. 39 Prospective cohort study, 1992–2004, Sweden, = 720 357 Composite neighborhood index – tertiles SGA < p2, 5 High
    Garcia Subirats et al. 33 Retrospective cohort study, 2000–2005, Barcelona, Spain, = 61 676 Contextual socioeconomic variables (such as unemployment) – quintiles PTB, LBW, SGA < p3 High
    Taylor-Robinson et al. 26 Retrospective cohort study 2002–2008, UK, = 31 785 Index of multiple deprivation – quintiles PTB Medium
    Urquia et al. 27 Birth cohort study, Ontario, 2000–2007, Canada, = 397 470 Neighborhood income – quintiles PTB High
    • LBW, low birthweight (<2500 g); PTB, preterm birth <37 weeks unless otherwise specified; SGA, small-for-gestational age (birthweight <10th centile) unless otherwise specified; VLBW, very low birthweight (<1500 g).

    Deprivation indicators varied across studies. One study used the Carstairs–Morris score, five studies used the Index of Multiple Deprivation, another five used the Townsend Deprivation Index, one study used the Jarman score, and five used neighborhood income as a proxy for deprivation at the neighborhood level.

    Table 2 shows the prevalences and risk estimates of the included studies for preterm birth, SGA and stillbirth. Twenty-one of the 24 included studies showed positive associations between adverse perinatal outcomes and neighborhood deprivation. The prevalence of preterm birth ranged from 3.8 to 6.7% in the least deprived quintile and 5.6 to 11.9% for the most deprived quintile. For SGA the prevalence was respectively 4.8–10.4% and 6.2–14.5%. Stillbirth rates ranged from 3.2 to 6.3 per 1000 births in the least deprived quintiles and from 4.6 to 7.0 per 1000 births in the most deprived quintile. All studies included in the meta-analysis used a wide variety of variables to adjust for potential confounders. One study did not report crude ORs, but we calculated these with comprehensive meta-analysis 46.

    Table 2. Prevalences and risk estimates of the included studies ordered by publication year (= 24).
    Study Prevalence of the outcome of interest Risk estimates (95% CI) for most deprived compared to least deprived neighborhood Covariates in fully adjusted model
    Spencer et al. 38

    LBW TDI MD: n = 3578 (9.8%), LD: n = 888 (5.1%) RGSC: MD: n = 128 (11.3%), LD: n = 63 (5.5%),

    VLBW TDI MD: n = 329 (0.9%), LD: n = 103 (0.6%); RGSC: NA

    LWB TDI: RR 1.99 (1.85–2.18)

    RGSC: RR 2.04 (1.53–2.73)

    VLBW TDI: RR 2.11 (1.73–2.57); RGSC: NA

    NA
    Smeeton et al. 44

    Stillbirth overall: n = 351;

    Early neonatal death overall: n = 198

    Stillbirth not significant (results not reported) Neonatal death OR 0.947 (0.849–0.997, = 0.038) NA
    Luo et al. 49a

    PTB MD: n = 10 163 (7.4%), LD: n = 5855 (6.3%);

    SGA LD: n = 4461 (4.8%), MD: n = 8515 (6.2%);

    Stillbirth LD: n = 585 (0.63%), MD: n = 961 (0.70%)

    PTB OR 1.16 (1.09–1.23), aOR 1.26 (1.17–1.35)

    SGA OR 1.41 (1.33–1.49), aOR 1.50 (1.40–1.60)

    Stillbirth OR 1.17 (0.95–1.43), aOR 1.30 (1.04–1.63)

    Infant sex, parity, plurality, ethnicity, maternal age, marital status, abortion history, mode of delivery, maternal illness, community size, and distance to the nearest hospital with obstetricians
    Manning et al. 35b PTB MD: n = 334 (8.2%), LD: n = 156 (3.8%) NA NA
    Dibben et al. 32

    LBW overall: 6.0%, MD: n = 25 005 (8.2%), LD: n = 12946 (4.2%);

    VLBW overall: 0.9%, MD: n = 3672 (1.2%), LD: n = 2020 (0.7%)

    LWB OR: NA, aOR 1.03 (0.99–1.07)

    VLBW OR: NA, aOR 1.14 (1.12–1.16)

    Age, social class of household, registration status, estimated household income, age–household income interaction, area income deprivation (AID), age–AID interaction
    Delpisheh et al. 42

    PTB MD: 14%, LD: 9%

    Term LBW MD: 2%, LD: 0%;

    LBW MD: 8% LD: 2%

    PTB NA

    Term LBW OR: NA, aOR 2.9 (0.9–9.6)

    LBW NA

    Maternal smoking, household smoking, parents in paid employment, father's employment, maternal employment, unemployed parents; Townsend score
    Janghorbani et al. 34b PTB overall: n = 202 (5.3%), MD: n = 92 (6.1%), LD: n = 49 (4.7%) PTB RR 1.31 (95% CI 0.94–1.84) Townsend score, age, gender
    Collingwood Bakeo and Clarke 40

    LBW overall: n = 3390 (5.8%)

    MD: n = 297 (7.2%), LD: n = 1388 (4.2%)

    LBW OR: NA, aOR 1.78 (1.54–2.05) Economic activity, number of people in the household, number of rooms in the household, household access to a car, housing tenure, region of usual residence Carstairs deprivation quintile, ethnicity, limiting long term illness status
    Luo et al. 47a

    PTB MD: n = 14917 (8.2%), LD: n = 9939 (6.7%);

    SGA MD: n = 22376 (12.3%), LD: n = 13500 (9.1%);

    Stillbirth MD: n = 836 (0.46%), LD: n = 474 (0.32%)

    PTB OR 1.23 (1.2–1.26), aOR 1.14 (1.10–1.17)

    SGA OR 1.40 (1.37–1.43), aOR 1.18 (1.15–1.21)

    Stillbirth OR 1.44 (1.29– 1.62), aOR 1.30 (1.13–1.48)

    Infant sex, parity, plurality, maternal age, education, ethnicity, marital status, and neighborhood income quintile
    Cubbin et al. 41

    LBW Washington overall: n = 171(4.4%)

    MD: n = 198(5.1%), LD: n = 151 (3.9%);

    LWB Florida overall: n = 290 (6.5%)

    MD: n = 353 (7.9%), LD n = 219 (4.9%)

    LBW Washington OR 1.21 (0.61–2.40), aOR 1.21 (0.61–2.40)

    LBW Florida OR 1.34 (1.17–1.54), aOR 0.99 (0.85–1.17)

    Neighborhood–level deprivation, income, education, paternal education, race/ethnicity, marital status, age, parity
    Urquia et al. 48a

    PTB overall: n = 7580 (5.3%)

    MD: n = 1605 (5.6%), LD: n = 1340 (4.7%)

    PTB OR 1.19 (1.11–1.28); aOR 1.25 (1.15–1.37) Infant sex, maternal age group, neighborhood income quintile, and recent immigrant status
    De Graaf et al. 22b PM MD: 13.5‰, LD: 9.3‰ NA NA
    Gray et al. 45a

    PTB overall: n = 8394 (5.6%)

    MD: NA, LD: NA

    PTB OR 1.15 (1.13–1.18); aOR 1.07(1.04–1.10) Deprivation, age, height, parity, sex, smoking, and obstetric intervention
    Beard et al. 31

    SGA(<p3) overall: n = 26 592 (3.4%)

    MD: n = 9879 (4.5%), LD: n = 4171(2.4%)

    SGA OR 1.88 (1.22–1.34), aOR 1.45(1.37–1.53) Quartile of disadvantage, baby's gender, maternal smoking, mother's age, aboriginality, pre-existing diabetes, pre-existing hypertension, gestational hypertension, gestational diabetes, previous pregnancy, year of birth, onset antenatal care, ethnicity, season of birth
    Agyemang et al. 21b

    PTB MD: n = 138 (6.8%), LD: n = 96 (5.4%);

    SGA MD: n = 339 (16.6%), LD: n = 150 (8.4%)

    PTB OR NA, aOR 1.03 (0.76–1.40)

    SGA OR NA, aOR1.62 (1.25–2.08)

    Age, parity, educational level, ethnicity, smoking and obesity
    Smith et al. 37

    VPTB overall: n = 103 (1.4%)

    MD: n = 35 (1.8%), LD: n = 10 (1.0%)

    VPTB RR 1.91 (1.7–2.06)
    Janevic et al. 43 NA

    LBW OR 1.99 (1.80–2.19), aOR 1.19 (1.11–1.27)

    PTB 32–36 weeks: OR 13.7 (1.30–1.44), aOR 1.06 (1.01–1.11)

    PTB <32 weeks: OR 1.55 (1.45–1.65), aOR 1.24 (1.13–1.36)

    Age, education level, parity, ethnicity, nativity, and smoking
    Liu et al. 46a

    PTB MD: n = 5026 (7.5%), LD: n = 4192 (6.3%);

    SGA MD: n = 7719 (11.6%), LD: n = 5046 (7.6%);

    Stillbirth MD: n = 434 (0.65%), LD: n = 300 (0.45%)

    PTB OR 1.21 (1.16–1.26), aOR 1.17 (1.12–1.23)

    SGA OR 1.60 (1,54–1.66), aOR 1.51 (1.46–1.57)

    Stillbirth OR 1.45 (1.25–1.68), aOR 1.39 (1.19–1.62)

    Maternal age, parity, smoking during pregnancy, maternal health problems, initiation prenatal care in 1st trimester
    Timmermans et al. 11b

    PTB: MD: n = 163 (5.9%), LD: n = 219 (4.8%);

    SGA: MD: n = 402 (14.5%), LD: n = 476 (10.4%)

    PTB aOR 1.22 (1.00–1.48)

    SGA aOR 1.41 (1.2 –1.59)

    Indicators for adverse perinatal outcome was related to all individual risk factors and the deprivation indicator
    Poeran et al. 36b

    PTB overall: n = 3865 (7.7%), MD: n = 204 (11.9%), LD: n = 84 (4.7%)

    SGA Overall: n = 4704 (9.4%), MD: n = 378 (11,8%), LD: n = 92 (4.7%)

    PM Overall: n = 600 (1.2%), MD: n = 53 (2.3%), LD: n = 2 (0.2%)

    NA NA
    Sundquist et al. 39

    SGA (<p2.5) overall n = 20487 (2.8%)

    MD: n = 4696 (3.5%), LD: n = 3942 (2.5%)

    SGA OR 1.38 (1.32–1.44), aOR 1.28 (1.22–1.34) Age, marital status, family income, educational level, urban/rural status, employment, mobility
    Garcia Subirats et al. 33

    PTB overall: 3395 (5.6%), MD: n = 382 (7.1%), LD: n = 197 (3.5%);

    LBW overall: n = 3931 (5.6%), MD: n = 372 (6.9%), LD: n = 708 (4.5%);

    SGA(<p3) overall: n = 1741 (2.3%), MD: n = 197 (3.5%), LD: n = 358 (2.2%)

    PTB aOR 1.51 (1.27–1.79)

    LBW aOR 1.56 (1.37–1.78)

    SGA aOR 1.66 (1.29–1.12)

    Maternal age, country of origin, parity, sex ofnewborn
    Taylor-Robinson et al. 26a PTB overall: n = 1612 (5.1%), MD: n = 1146 (5.6%), LD: n = 27 (4.1%) PTB OR 1.55 (1.36–1.76), aOR 1.32 (1.12–1.55) Maternal age, parity, smoking status, BMI, ethnicity
    Urquia et al. 27a PTB overall: n = 24623 (6.2%), MD: n = 5594 (7.0%), LD: n = 4559 (5.7%) PTB OR 1.25 (1.20–1.30), aOR 1.26 (1.2–1.31) Maternal age, parity and immigrant status
    • aOR, adjusted odds ratio, 95% CI, 95% confidence interval; LBW, low birthweight (<2500 g); LD, least deprived; MD, most deprived; NA, not available; OR, odds ratio; PM, perinatal mortality PTB, preterm birth <37 weeks unless otherwise specified; SGA, small-for-gestational age (birthweight <10th centile) unless otherwise specified; VLBW, very low birth weight (<1500 g); VPTB, very preterm birth (<33 weeks).
    • a Studies included in the meta-analyses.
    • b Exclusion based on index subdivision (not in quintiles).

    Meta-analysis

    The two most common reasons for exclusion in our meta-analysis were lack of results at neighborhood level (e.g. results were provided for whole villages, counties or states, = 38) or outcome measures were defined that were not eligible for meta-analysis (including LBW, data not reported in quintiles, = 17). The meta-analysis eventually included cohort studies on adverse perinatal outcomes associated with neighborhood deprivation. If outcomes for several years were reported, the most recent results were used for the meta-analysis. Assessment of study quality showed that five studies were of high quality 27, 46-49 and two of medium quality 26, 45.

    Figure 2 summarizes the comparison of the least and most deprived neighborhoods for preterm birth, SGA and stillbirth. None of the included studies reported data on perinatal mortality, but all reported stillbirth by >20 weeks of gestation. The random effects model comprising all seven studies suggests a positive association for preterm birth crude OR 1.28 (95% CI 1.20–1.37), adjusted OR 1.23 (95% CI 1.18–1.28). Heterogeneity was noted among the individual study effects [I2 (crude) 94% (< 0.001), I2 (adjusted) 77% (< 0.001)]. Countries included in the meta-analysis have comparable preterm birth rates 50. For the outcomes SGA and stillbirth, only studies that indicated deprivation using the index neighborhood income remained after the selection process. Similar positive associations between least and most deprived neighborhood quintiles were found. The OR from crude results for SGA was 1.47 (95% CI 1.34–1.60), I2 95%, < 0.001, adjusted OR 1.31 (95% CI 1.28–1.34), I2 99%, < 0.001). For stillbirth, the crude OR was 1.38 (95% CI 1.23–1.54), and adjusted OR was 1.33 (95% CI 1.21–1.45) without marked heterogeneity [I2 (crude) 41%, = 0.185, I2 (adjusted) 0%, = 0.793]. The studies in our meta-analysis showed a consistent association between living in a deprived neighborhood and adverse perinatal outcomes.

    Details are in the caption following the image
    Forest plot of pooled random effects adjusted odds ratio (aOR) and 95% confidence intervals (CIs) of adverse perinatal outcome comparing the most deprived neighborhood quintile with the least deprived neighborhood quintile. White squares indicate the aOR in each study, with square sizes inversely proportional to the standard error of the OR. Horizontal lines represent 95% CIs. p-Values represent p for heterogeneity. (a) Pooled effects for preterm birth. (b) Pooled effects for small for gestational age. (c) Pooled effects for stillbirth.

    Subgroup analysis

    In a subgroup analysis comprising five studies that used neighborhood income as measure of deprivation, we also found a positive association between the least and most deprived neighborhood quintiles: crude OR 1.22 (95% CI 1.19–1.25), I2 24%, = 0.261; adjusted OR 1.21 (95% CI 1.15–1.27), I2 78%, = 0.001. Since six studies were excluded based on how they categorized their neighborhood index (cut-off point other than quintiles), we performed a univariate meta-regression analysis with neighborhood cut-off point as moderator to assess the empirical relation between neighborhood cut-off point and the log of the observed OR. Crude ORs from four of these six studies were available for this analysis in preterm birth 11, 21, 34, 35, and crude ORs from two studies were available for SGA 11, 21. This figure indicates that we found no empirical relation for the cut-off point in this analysis (Figure 3).

    Details are in the caption following the image
    Scatter plot representing the meta-regression analysis to test the association between the cut-off point for neighborhood index and the log adjusted odds ratio (aOR) of preterm birth and small-for-gestational age (SGA). The area of each circle is inversely proportional to the variance of the log relative risk estimate. (a) Preterm birth. (b) SGA.

    A sensitivity analysis was performed for all outcomes to evaluate the stability of the results. We performed a subgroup moderator analysis to compare the mean OR for five studies using income as a measure of deprivation and two other studies using another index in preterm birth: ORincome 1.25 (95% CI 1.15–1.37) and ORother 1.38 (95% CI 1.25–1.70). These studies were also the studies rated as being of high. This difference was not significant (Q = 0.70, = 0.404). Although no asymmetry was seen on the funnel plot (data not shown), the largest analysis included only seven studies. This number of studies is too small to perform an adequate assessment of publication bias 51. We tried to minimize risks of publication bias with our search strategy, first by including Web of Science as one of the search engines, because of its provision of conference abstracts, and secondly, by a review of reference lists of all 108 studies which were eligible based on title or abstract, and of one meta-analysis and several reviews on this topic.

    Discussion

    This systematic review and meta-analysis indicates that neighborhood deprivation is associated with preterm birth, SGA and stillbirth. Compared with the least deprived neighborhood quintile, ORs of adverse perinatal outcomes in the most deprived quintile were significantly increased for preterm birth, SGA and stillbirth. This is the first meta-analysis in which preterm birth and SGA were analyzed separately. Whereas the previous analysis on LBW showed an excess prevalence due to deprivation effect of 11% 23, our meta-analysis, in which we analyzed SGA and preterm birth separately, provided a prevalence of 31 and 23%, respectively. Our findings suggest that these two disease entities share only a part of the deprivation pathway, because the stratified ORs for SGA and preterm birth are higher individually than LBW in which preterm birth and SGA are combined. Stratification seems to distinguish both the risk pathways of preterm birth and SGA more than LBW alone. The etiologies of the two outcomes are not always related to each other (preterm birth could be induced by infectious diseases, whereas placenta insufficiency primarily induces SGA). Furthermore, preterm birth often has a strong care effect (iatrogenic preterm birth), which is absent in SGA. Stratification may therefore clarify the neighborhood effect. Although such excess risks may seem small, they suggest the high attributable risk impact of deprivation through the high prevalence of deprivation.

    The association between many adverse perinatal outcomes and low socio-economic status is known. It is thought to be induced through multiple pathways, most importantly low education and low income levels 52. Although the present analysis did not focus on etiological factors that could explain the relation between non-medical risk factors and adverse perinatal outcomes, it is known from previous studies that decreased wealth and living conditions increase physiological stress 53, and low income levels and deprivation are associated with poor housing, nutrition and health care access 3. If we assume that adverse perinatal outcomes are the result of the interrelation between individual, environmental, and care-related factors 3, 11, deprived areas could contribute to adverse outcomes in several ways. In deprived areas, so-called non-medical risk factors (such as lifestyle and social risk factors) are much more common, especially in urban areas 3, 23, 54.

    Within a patient care perspective it is important to acknowledge that risk accumulation not only includes the commonly measured standard risks, but also many unmeasured disadvantages or risk factors, which often escape standard epidemiologic research. The physiological (air pollution, noise) and psychological environment (safety) are also part of the usually unmeasured risk burden in deprived neighborhoods 55.

    The greatest challenge in the meta-analysis was to overcome heterogeneity. We believe this is mostly due to the variable cut-offs used in forming neighborhood quintiles. In our meta-analysis, the results for preterm birth in our primary analysis did not differ from our subgroup analysis for income level. We approached all authors of potentially eligible studies to cooperate in our meta-analysis in order to pool pregnancy outcomes on individual level, so that we could investigate the comparability of the different indices (= 11). Unfortunately, most authors did not respond, so we were unable to perform the analysis. Another reason for this heterogeneity might be the variety of used definitions to indicate neighborhood deprivation. Some of these indices were compared in previous studies, and the use of area-based deprivation indices is an accepted method for measuring social inequality in neighborhoods 18. Moreover, income seems to be a good proxy for area-based deprivation related to health 19. However, it is still unclear whether area-based measures reflect the cumulated impact of individual socio-economic status, or represent the crude neighborhood effect or a combination of both 19, 49. Despite the high heterogeneity, we believe that pooling of the available data provided valuable information about neighborhood deprivation and the risk of unfavorable pregnancy outcomes, and we used random-effect models to calculate pooled risk estimates.

    Strengths and limitations

    A major strength of this study is that we performed the first meta-analysis in which preterm birth, SGA and stillbirth were analyzed separately. In addition, we included all study types in the initial search to be able to identify the (cluster) randomized controlled trials. This did not result in inclusion of randomized controlled trials. However, this is a strength because if we had ignored this study type, we could have overlooked studies investigating the effect of deprivation on, for example, neighborhood level.

    Our analysis has some limitations. First of all, we were unable to answer our research question with only multilevel studies, as in the previous meta-analysis referred to 23. Some effect of clustering may be present if data were retrieved from true cluster designs in which clusters involve, for example, schools, hospitals or communities, and ignoring this effect (if present) might lead to some overestimation of the precision and statistical significance. However, in our study, neighborhoods were overall large and defined by different principles and therefore unequally sized. We assumed that under these conditions no additional measures were needed to account for the study effect beyond the per study multilevel term.

    Secondly, as we decided to use data presented in quintiles for the reasons mentioned above, we had to exclude six studies from meta-analysis because they did not report neighborhood deprivation in quintiles. Risk estimates from these studies were mostly within the range of the risk estimates of studies pooled in the meta-analysis. In a meta-regression analysis we did not find an empirical relation for cut-off points. Another limitation was that the definition of fetal growth restriction as “LBW”. This was a reason for exclusion because we were unable to categorize findings into the LBW related to preterm birth outcome or the outcome of LBW babies at term (>37 weeks of gestation). We only included results from studies when growth restriction was defined as “small-for-gestational age below the 10th percentile”. We advocate separation of SGA and preterm birth because both etiology and long-term consequences differ considerably between these outcomes 56-58. For the association between neighborhood deprivation and LBW (as a broad category), we refer to the meta-analysis by Metcalfe et al. 23.

    Thirdly, we were not able to rank the included countries according to perinatal outcomes. The Euro-Peristat committee was able to make such a comparison. In their most recent published report, they compared perinatal outcomes in 29 European countries 59. They reported marked differences between countries. However, such a report was not available for the non-European countries included in our study. Due to the absence of this information, we were not able to make an “over countries” comparison and thus to relate our findings to the ranking of countries.

    Lastly, we were not able to stratify for ethnicity. Although the included studies were adjusted for ethnicity, we missed information about other ethnic groups living in these neighborhoods which were not included in their analysis. Therefore these outcomes were not representative for the whole neighborhood population. Confounding by ethnic differences is therefore unavoidable in this study. This is important because neighborhood effects might not be consistent across ethnic groups 24. Deprivation could have a stronger negative effect on Western women than non-Western women 60. It seems that simple aggregation of particular individual effects does not explain our findings at neighborhood level. Other partly unknown underlying mechanisms may influence both perinatal risk factors and outcomes at neighborhood level.

    Implications

    Since poor maternal circumstances during pregnancy have both short- and long-term consequences, it makes sense to organize tailor-made antenatal healthcare that is responsive to women's needs by taking into account deprivation notions, preferably in combination with preconception care. In particular in large cities, this implies involvement of local initiatives and engagement of public health services 61. A systematic approach in antenatal risk selection for both medical and non-medical risk factors with subsequent continuity of care might support early detection of potential high risk. More awareness regarding the medical impact of the non-medical domain should be advocated among healthcare professionals, but also among public health workers and policymakers. It may seem challenging in practice to reach women in deprived neighborhoods for specific intervention programs, but research has shown that specific recruitment strategies can be used to achieve participation among these women 62.

    Conclusions

    This systematic review and meta-analysis suggest that neighborhood deprivation is associated with SGA, preterm birth and stillbirth. However, more methodological research is necessary to determine the comparability of several neighborhood deprivation indices in relation to these perinatal outcomes. The included studies were not designed to explore mechanisms, so more etiological studies at a neighborhood and individual level are necessary to gain understanding of the effect of “neighborhood deprivation” on adverse perinatal outcomes. In the meantime this should not prevent us from designing new policies and programs for women living in deprived neighborhoods where both social and medical risk factors are present to a great extent.

    Funding

    No special funding.