Contribution of psychopathology, psychosocial problems and substance use to urban and rural differences in birth outcomes

Abstract

Background: Urban residence contributes to disparities in preterm birth (PTB) and birth weight. As urban and rural pregnant populations differ in individual psychopathological, psychosocial and substance use (PPS) risks, we examined the extent to which PTB and birth weight depend on the (accumulative) effect of PPS risk factors and on demographic variation. Methods: Follow-up study from 2010 to 2012 among 689 urban and 348 rural pregnant women. Urbanity was based on the population density per ZIP code. Women completed the validated Mind2Care instrument questionnaire, which includes the Edinburgh Depression Scale, and demographic, obstetric and PPS questions. Pregnancy outcomes were extracted from medical records. With regression analyses we assessed crude and adjusted associations between residence and birth outcomes, adjusted for available confounding or mediating factors. Results: PTB was significantly associated with segregation, maternal age (<25 and ≥ 35 years old), primiparity, smoking during pregnancy and the accumulation of risks, but not with residence (urban, 4%; rural, 7%; P = 0.16). Mean birth weight was significantly lower for urban babies (crude β: −174; P < 0.001). Adjusting for potential confounders and mediators, non-Western ethnicity, parity and smoking during pregnancy significantly decreased birth weight besides residence. The accumulative effect of PPS risk factors significantly decreased birth weight (β: −58 g per risk factor; P < 0.001). Conclusion: PTB was not associated with residence. The lower birth weight of urban babies remains significant after adjusting for urban risks, such as non-Western ethnicity and the PPS risk factor smoking. The accumulation of multiple (moderate) PPS risks accounts partly for the urban effect.

Introduction

Preterm birth (PTB) and low birth weight (LBW) are two important adverse birth outcomes, affecting >10% of women and their newborns worldwide.^1,2 Both conditions are leading causes of perinatal and infant mortality,^3,4 chronic disease in later life⁵ and result in high health-care costs.^1,6

Among other well-known risks factors, such as pre-eclampsia and ethnicity contributing to adverse birth outcomes,^2,6 the influence of geographical disparities is relatively understudied.

Regarding geographical disparities in perinatal health, there are contrasting findings on the contribution of urban and rural residence.^7–13 In international literature, rural populations generally have the highest prevalence of adverse birth outcomes, of deprivation and of individual risk factors, such as a non-indigenous ethnicity, low socio-economic status and maternal depression.^8–10,13,14 Some studies, however, find the opposite, in particular, studies from the Netherlands.^7,15

One explanation for the contrasting findings is that they are confounded or mediated by underlying individual risks, as urban and rural populations differ in many respects, e.g. demographic composition and environmental exposures.^1,6,16,17 More specific and often co-occurring risk factors concerning psychopathology, psychosocial problems and substance use (PPS) are also associated with PTB and LBW,^14,15,18 in particular, in case of risk accumulation.¹⁹ Geographical disparities in PPS risk factors also exist. Therefore, urban–rural disparities in birth outcomes may not only be affected by geographic risks but also by (the accumulation of) PPS risks.

The observed perinatal health disparities urge for targeted improvement strategies. Increased insight into the contribution of specific (treatable) risk factors for adverse birth outcomes in combination with screening and subsequent referral to targeted interventions could reduce these disparities.

To our knowledge, there are no studies that systematically investigated the (accumulative) effect of PPS risk factors in relation to geographic disparities in birth outcomes.

We aimed to investigate the extent to which PTB and birth weight depend on the (accumulative) effect of individual PPS risk factors in urban and rural areas or on geographic and demographic variation.

Methods

Data were obtained from the validated Mind2Care screen-and-advice instrument (M2C, formerly known as the GyPsy instrument),²⁰ including the Edinburgh Depression Scale (EDS),²¹ and a set of risk factors for psychiatric disorders during pregnancy, including substance use, demographic, obstetric and psychosocial factors. M2C was specifically developed by the Erasmus Medical Center as a tool for screening and the subsequent treatment allocation for PPS risks during pregnancy, as these often escape detection.²²

In Rotterdam, one of the major cities in the Netherlands, M2C was implemented from June 2010 to June 2012, in three midwifery practices and a general hospital. All pregnant women with a booking visit at these clinics were approached for this study. The midwifery practices were located in a severely deprived, a less deprived and a non-deprived area. The hospital served a severely deprived area. Together, 40% of all urban Rotterdam pregnant women were served by a participating practice.

In Meppel, a small rural city, M2C was implemented from April 2012 to May 2012 in four midwifery practices and the local general hospital. All pregnant women residing in this city and the adjacent rural agricultural area who attended the clinics during the study period were approached for this study.

All participating women completed the M2C questionnaire on a personal digital assistant before their pregnancy check-up, as part of routine practice. In total, 1076 (88%) of the 1230 eligible women were screened with M2C (figure 1). Of these, 11% refused participation and 1% was excluded owing to a language barrier. M2C data were complete for 1037 women (96%), as data on residence or ≥ 10 M2C questions were missing for 39 women. Postpartum, follow-up information and birth outcomes were extracted from medical records. Birth outcomes were unavailable for 25% of women (n = 259). Main reasons included miscarriage or pregnancy termination (4%), referral outside the study region (4%) and lack of consent for follow-up (13%). Outcomes of 778 women were analysed. Women with incomplete data on birth outcomes were more often of rural residence, non-Western ethnicity and low educational level compared with women with complete data (42 vs. 31%, 33 vs. 23% and 28 vs. 21%; P < 0.05; data available on request).

Study approval was obtained from the medical ethical board of the Erasmus University Medical Center Rotterdam [MEC-2013-162].

Birth outcomes included PTB (<37 weeks of gestation), small for gestational age (SGA, <10th percentile based on national percentile distributions²³) and birth weight (in grams). The definition of urban and rural residence was based on the population density per ZIP code, as classified in the 2011 consensus of the Dutch Central Bureau of Statistics (CBS).²⁴ Urban residence was defined as residence in a ZIP code area with >1500 household per square kilometre, and rural residence was defined as residence in a ZIP code area with <1500 households per square kilometre. A dichotomous definition was used to ensure sufficient power for statistical analysis. ZIP code and individual risk factors were measured with M2C.

Individual risk factors included demographic factors (socio-economic status, segregation, ethnicity, educational level, maternal age and single status), obstetric factors (gestational age, gravidity and parity), psychosocial stressors (insufficient social support, relational problems, financial debts, housing problems and physical/sexual abuse), substance use (smoking, alcohol consumption and illicit drug use) and psychopathology-related factors (depressive symptoms and psychotropic medication use). Demographic and obstetric factors were regarded as individual risk factors related to geographical composition. Psychosocial stressors, substance use and psychopathology-related factors were regarded as individual PPS risks.

Socio-economic status was measured by proxy (z-score for socio-economic status for each ZIP code) and was obtained from the Dutch Social and Cultural Planning Office.²⁵ Segregation was qualified as the distribution of non-Western immigrants among all other residents per ZIP code within one borough. The index can be interpreted as the proportion of non-Western immigrants that would have to move from the ZIP code areas to achieve a uniform distribution over the borough. Segregation indexes were calculated conventionally by using data on population size and composition from the CBS website and the formula of Taylor et al.^26,27

Depressive symptoms during pregnancy were measured by the EDS, a validated 10-item self-report checklist with a total score range of 0–30.²¹ Women with EDS scores ≥10 were regarded as having clinically relevant depressive symptoms.²⁸

Risk accumulation was expressed through a weighted sum score of all separate PPS risks representing the overall PPS risk load. Based on the study of Timmermans et al.,¹⁹ PPS risk factors accounted for 1 point, except for the continuation of substances, which accounted for 2 points. The sum score ranged from 0 to 13, where 0 represented absence of any risk.

As a general reference for the prevalence of demographic risk factors, PPS risks and birth outcomes, we used two large Dutch cohort studies (Generation R and ABCD study)^29,30 and national reference data as stated on the websites of the Dutch Social and Cultural Planning Office and the Dutch Perinatal Registry.^23,25

Urban and rural differences in birth outcomes and in individual risks, and crude associations between individual risk factors and birth outcomes were assessed with chi-squared tests, student’s t-tests and Mann–Whitney U-tests. The crude association between residence and birth outcomes was examined with univariate logistic and linear regression analyses. Multiple stepwise regression analyses were conducted to assess the association between residence and adverse birth outcomes, adjusted for all available confounding or mediating individual risk factors. All birth outcomes and individual demographic risk factors with a P < 0.10 following descriptive statistics were included in the multiple stepwise regression model, with model 1 including demographic risk factors, model 2 including demographic and PPS risk factors and model 3 including the accumulative effect of PPS risks. All models were adjusted for gestational age at birth.

Associations following univariate and multiple regression analyses were expressed through adjusted odds ratios or betas [95% confidence interval (CI)] and P-value, respectively. Associations with a P < 0.05 were regarded as statistically significant. All analyses were performed using the Statistical Package for Social Science, version 20.0.

Results

Table 1shows birth outcomes differed across residence, with significantly lower birth weights in urban areas (mean weight 3385 vs. 3533 g, P < 0.01). Prevalence of PTB and SGA were comparable between urban and rural babies (urban PTB 4% vs. rural PTB 7%, P = 0.16; urban SGA 7% vs. rural SGA 5%, P = 0.42). As SGA is a standardized measure of birth weight, SGA was not included in further analysis.

A number of demographic factors were also associated with residence. Urban pregnant women were more likely to be of low socio-economic status, non-Western ethnicity and high educational level, and were more often single (all P < 0.01). The distribution of non-Western immigrants was higher among urban residents (median segregation index 0.30 vs. 0.17, P < 0.001). Urban women were also more often of primigravidity and primiparity, and had, owing to the study design, a lower gestational age at screening (median 9 vs. 26 weeks, P < 0.05).

Of the PPS risk factors, financial debts and alcohol consumption during early pregnancy had a higher prevalence among urban women (9 vs. 5% and 20 vs. 11%, P < 0.05). Urban women had more depressive symptoms (17 vs. 13%), and rural women were more likely to continue psychotropic medication use during pregnancy (3 vs. 1%, both P < 0.05). Accumulation of PPS risk was highest among urban women (urban 12% sum score ≥ 5 vs. 3% rural, P < 0.001). The demographic profiles, the prevalence of PPS risk factors and the prevalence of adverse birth outcomes were representative for Dutch urban and rural populations.

Table 2 shows that a lower birth weight was associated with low socio-economic status, non-Western ethnicity, primigravidity, multigravidity, primiparity, financial debts, physicial and sexual abuse, continued smoking during pregnancy and the accumulation of PPS risks (all P < 0.05). Maternal age was associated with LBW at a P < 0.10 level and therefore included in the multivariate regression model too.

PTB was associated with segregation, low and high maternal age (<25 years and ≥35 years), primiparity, smoking and accumulation of PPS risks (all P < 0.05, Supplementary table S1). Because the prevalence of PTB did not significantly differ between urban and rural women, PTB was not included in further analysis.

Crude analysis (table 3) showed that urban babies had a 174-g lower birth weight than rural babies (95% CI: −240, −107 g; adjusted for gestational age at delivery). This significant effect decreases to a 130 - and 138-g lower birth weight (P < 0.01) after adjusting for individual demographic factors in model 1 and PPS risk factors in model 2, respectively. Non-Western ethnicity and primiparity significantly decreased birth weight (−83 g, 95% CI: −161, −4; −135 g, 95% CI: −200, −70). Both factors remained significantly associated with birth weight after adjusting for PPS risk factors. Of the PPS risk factors, only smoking significantly contributed to a lower birth weight (−230 g, 95% CI: −351, −109). The accumulation of PPS risks contributed to a decrease in birth weight with 58 g for each risk factor (95% CI: −106, −10; P < 0.05).

Discussion

This multicentre follow-up study showed that urban babies have a 174-g lower birth weight than rural babies. After adjusting for available mediators and confounders, non-Western ethnicity, primiparity and the PPS risk factor smoking decreased birth weight besides urban residence. However, the accumulation of multiple PPS risk factors contributed to the urban effect with a 58-g lower birth weight for each risk factor.

The strength of this study is the fairly large population-based sample comparing a small rural area with a large urban city, with representative prevalence of risk factors and perinatal outcomes for Dutch large rural and urban populations. However, while our results may be generalizable in this context, more data are needed to explore whether the same observation holds for medium-size cities. The second strength is the use of prospective data derived from routine practice settings.

Besides residence and demographic factors, a comprehensive set of PPS risk factors were included in the analysis, allowing for an examination of the contribution of PPS risks to disparities in birth outcomes.

PTB was slightly more prevalent in the rural setting, which is in line with previous findings.^8–10 The Canadian Study by Lisonkova et al.,¹¹ however, found rural women to be less likely to deliver preterm. In contrast to a number of previous studies,^8–10,12 but in line with the study of Hillemeier et al.,¹³ crude mean birth weight appeared lower in the urban area. Five explanations possibly contribute to these contrasting findings.

First, positive selection could have introduced bias to our study. Data on birth outcomes were more often incomplete for rural women and low-educated non-Western women. As non-Western ethnicity and rural residence were adversely associated with birth weight, we expect our findings to be an underestimation of the true urban–rural disparity in birth weight rather than an overestimation. In addition, rural women had a substantially higher gestational age during screening, which could have decreased the proportion of (early) preterm deliveries in our study. As the prevalence of (early) PTB and mean birth weights in our urban and rural population were comparable with national and European outcomes, we assume selection bias to be limited.^23,25,31

Second, the causal pathway of risks leading to PTB and LBW is likely to be equal across countries. International differences in the prevalence of the underlying demographic and PPS risks in urban and rural areas could, however, explain the urban–rural disparities in PTB and LBW prevalence. Risk factors for LBW are most prevalent in our urban sample, which is in line with previous Dutch studies.^7,19 In non-Dutch studies, however, these risk factors are most prevalent in rural samples.^8,9,11

Third, our results can be distorted by the use of a dichotomous definition of residence. The rural area includes rural areas adjacent and rural areas not adjacent to urban areas. In this context, we must be aware of heterogeneity of rural communities. Most remote rural areas currently are low-income and low-educated, as the better educated population tends to migrate to urban environments for better perspectives. However, rural areas adjacent to larger cities may profit from ‘back migration’ of people taking the disutility of commuting for living in a healthier environment. This may result in better birth outcomes in rural areas adjacent to an urban area than in rural areas not adjacent to an urban area.^10,12

However, as recommended by Strutz et al.,⁸ we choose the geographical area-level definition best suited to the population under study.

Fourth, most studies on the urban–rural effect were conducted in the 1980s and 1990s. It is likely that urban and rural populations have been changed over the past three decades in terms of poverty, infrastructure, prenatal care and population composition. Changes in this respect occurred in both the rural area and the metropolitan cities.

Finally, comparability with other studies is limited if selective sampling has been applied (e.g. non-Hispanic white women only) and PPS risks were not of primary interest.

This study is subject to several limitations. First, individual risk factors were derived from the self-report M2C questionnaire. Self-reportage can introduce misclassification in our results. However, the same is true with professional-based screening. Because of lack of an ethically approved gold standard, e.g. for financial debts and sexual or physical abuse, self-report is the most feasible alternative for those risk factors. Second, because we focused on PPS, a major limitation of this study is the lack of some important predictors for PTB and LBW, such as pre-eclampsia, previous PTB, maternal body mass index and infectious disease.^2,6 This could have led to omitted variable bias and may decrease the generalizability of our results in routine obstetric care.

In the future, community service factors, such as access to care, and the quality of the physical environment, such as noise and air pollution, could be added to the set of risk factors, as the individual residence effect appeared strong in this study. A multilevel study could be beneficial for an exploration of the underlying mechanism through which residence and individual risk factors are linked, and lead to adverse birth outcomes. Our data set was too small to investigate such effects.

While we started with the assumption of individual PPS risk factors to explain geographical inequalities in perinatal outcomes, this study shows that the strong residence effect is partly explained by the accumulation of PPS risks and, except for smoking, not by individual PPS risk factors. In addition, rural areas seem, not per definition, at low risk for adverse birth outcomes. Following this argument, the M2C instrument should routinely be implemented in antenatal care. This has large implications for prenatal care, as risk management starts from a uniform approach, rather than based on clinical assumptions.

We believe that the sacrifice of time for screening and health promotion targeted on PPS risks leading to adverse birth outcomes, in particular smoking, outweighs the benefits of detecting and possibly preventing adverse birth outcomes.

Funding

This work was funded by Stichting Achmea Gezondheidszorg [grant number z-282].

Conflicts of interest: None declared.

Acknowledgements

We acknowledge all participating obstetricians, midwives and research assistants who provided the opportunity for this study. Stichting Achmea Gezondheidszorg is acknowledged for providing funding for this study (grant no. z-282). The funders did not participate in any part of data collection, data analysis or interpretation of the data, nor in the writing or approval of the manuscript. We also thank Lauren Capron for critical revisions to the English language in this manuscript.

References