Background

Several recent reports, including the Fair Society, Healthy Lives report (The Marmot Review) and World Health Organisation's Commission on Social Determinants of Health (CSDH) Closing the Gap in a Generation, have identified investment in strategies to promote 'a healthy start to life' as having the greatest potential to reduce health inequalities across the life course [1–3]. 'Giving every child the best start in life' was nominated by the Marmot Review as the number one priority for implementation [1]. Central to this recommendation is the accumulating evidence to show that brain development is highly sensitive to external influences in utero and in early childhood, with potential life long effects [4, 5]. Infants that are born preterm, small for gestational age and/or with a low birthweight have well documented increased health risks in childhood, with more recent evidence supporting effects continuing into later life [6]. This research provides a powerful argument for supporting early intervention to improve maternal health in pregnancy and pre-conception as part of a comprehensive strategy to improve maternal, newborn and child health.

The major contribution of the CSDH Report and the Marmot Review is to situate this strategy within the broader context of health equity and the social determinants of health [1, 2]. There is mounting evidence in both a developing and a developed country context, demonstrating higher prevalence of low birthweight and small for gestational age infants associated with individual and area-level measures of social adversity and deprivation [7–14]. Worse outcomes have been reported for 'blue collar workers' compared with 'white collar workers' [10, 12], lone parents [8, 11], low income households [11], and mothers living in deprived neighbourhoods [11, 13, 14]. A small number of studies have reported a positive association between low birthweight and stressful life events during or immediately preceding pregnancy [15, 16]. Intimate partner violence, trait anxiety in the second and third trimester, and maternal distress in pregnancy also show associations with low birthweight [19–23]. Two potential pathways have been identified to explain these findings. The first posits that psychosocial factors such as stress or lack of social support impact on birth outcomes via neuroendocrine, immune or vascular mechanisms [14, 16, 17]. The second, more well-established pathway is via health behaviours which are known to be associated with increased risk such as smoking, maternal nutrition, infection and substance use [24].

The overall aim of this paper is to examine social determinants of low birthweight in an Australian population-based cohort and consider implications for health policy and health systems. The paper draws on data collected in an Australian population based survey of 4,366 women who gave birth in South Australia and Victoria to examine (i) the prevalence of stressful life events and social health issues in the 12 months before the birth, and (ii) women's experience of discrimination in health care settings as factors which may potentially contribute to poor infant health outcomes. To our knowledge, discrimination in maternal health care has not previously been investigated as a factor which may contribute to poor maternal and/or child health outcomes. Post hoc analyses are reported examining associations with attendance at antenatal care and with infant birthweight.

Method

Sample

The 2008 Healthy Mothers Healthy Families (HMHF) Survey questionnaire was mailed to women giving birth in two Australian states (Victoria and South Australia) in September/October 2007 at five to six months postpartum, excluding those who had a stillbirth, or whose baby was known to have died. Public and private maternity hospitals and homebirth practitioners in both states distributed questionnaires to women who gave birth under their care in the study period. All hospitals with births in the study period (n = 110) agreed to participate, however one small public hospital with approximately 90 confinements in the study period subsequently withdrew. Together South Australia and Victoria accounted for greater than 31% of the 289,496 births in Australia in 2007 [25]. The inclusion of two states provided for greater diversity in the sampled population compared with previous Victorian surveys [26–28]. Compared with Victoria, South Australia has a higher proportion of births to women living in outer regional (11.8% versus 4.2%) and remote/very remote areas (3.9% versus 0%), a higher proportion of Aboriginal mothers (3.0% versus 1.0%) and women giving birth under 20 years (4.6% versus 2.6%), while Victoria has a higher proportion of births to women born overseas (26.6% versus 16.3%) [25].

Procedure

The primary aim of the study was to assess women's views and experiences of care received during pregnancy, birth and the postnatal period in representative samples of women giving birth in South Australia and Victoria in 2007. Questionnaires, together with a covering letter inviting women to take part, an explanation of the study in six community languages (Arabic, Vietnamese, Cantonese, Mandarin, Somali and Turkish) and a reply paid envelope for returning the questionnaire free of charge, were posted to women at six months postpartum. Two reminders were sent at two-week intervals with the second including a repeat copy of the questionnaire. The timing of the survey at five to six months postpartum matches previous Victorian surveys [26–28] and was decided taking into account the literature suggesting a 'halo' effect in relation to reporting of women's experiences of maternity care in the period immediately after childbirth [27] and the inclusion of questions on postpartum maternal health outcomes (e.g. maternal depression, urinary incontinence). A copy of the questionnaire is available on the study website: http://www.mcri.edu.au/HMHFSurvey. The study period - four weeks in September 2007 in Victoria, and eight weeks in September/October 2007 in South Australia - was chosen to ensure sub-groups of interest (e.g. women attending six main models of care, women <25 years, single women, and women whose infant was <2,500 grams at birth) were large enough for meaningful statistical comparisons.

Questionnaire

Information was collected on women's views and experiences of care, health service use, maternal socio-demographic characteristics, reproductive history, and maternal and infant health outcomes, including infant birthweight. The questionnaire also included questions on maternal smoking in pregnancy, maternal pre-pregnant weight and height, and stressful life events and social health issues in the 12 months before the index birth. Measures of socio-economic status included: pre-tax household income adjusted for household size and composition, possession of a current health care concession card and health insurance status (private cover/Medicare only). Women were classified as being at higher risk of complications in pregnancy if they had a previous low birthweight infant, preterm birth, or stillbirth, or if they had any substantial medical or obstetric risk factors in the current pregnancy (e.g. multiple pregnancy, gestational diabetes). Maternal pre-pregnant Body Mass Index (BMI) was calculated as the ratio of body weight in kilograms divided by the square of height in metres. Smoking status was assessed using a multiple response question enabling women to indicate if they had never smoked, quit smoking before their pregnancy, quit since the beginning of pregnancy, cut down cigarette smoking since finding out they were pregnant, continued to smoke at about the same level or increased their smoking since the beginning of pregnancy [29]. Women who said they had quit before pregnancy or never smoked were categorised as non-smokers. Women who said they quit when they found out they were pregnant were categorised as smoking in early pregnancy; women who cut down, increased or continued to smoke about the same number of cigarettes as before their pregnancy were categorised as continuing to smoke throughout their pregnancy.

Assessment of stressful life events and social health issues drew on items included in the PRAMS study [30] with additional items incorporated based on results of a pilot survey and consultations with Aboriginal community organisations and communities in the two states. Women were asked "Did any of the following things happen to you in the 12 months before your new baby was born?" and invited to tick 'yes' or 'no' to 23 items including major life events, such as separation and divorce, moving house, losing your job, death of a close family member or friend and social health issues, such as having a lot of bills you couldn't pay, not having enough money to buy food, legal troubles or being involved in a court case, serious family conflict or being homeless. A complete list is provided in an additional file (see Additional File 1). Perceived discrimination was assessed using five questions (see Additional File 1) adapted from the Measure of Indigenous Racism Experience [31] to elicit information about women's experience of being discriminated against by health professionals in the perinatal period. The questions were designed to capture the concept of being 'treated unfairly', that is being treated as inferior, with less respect or courtesy, or receiving poorer care than other people. Women were also asked if they had been insulted, stereotyped, talked down to or ignored. Although based on a measure originally designed to assess racial discrimination, the questions were framed to illicit experiences of perceived discrimination in a variety of other contexts, for example, in relation to maternal age, education and other social characteristics.

The study was approved by the Royal Children's Hospital Human Research Ethics Committee (HREC), Victorian Department of Human Services HREC, South Australian Department of Health HREC and a number of other university and hospital based HRECs.

Analysis

Data on the social and obstetric characteristics of study participants were compared with routinely collected Victorian and South Australian perinatal data for all women who gave birth in the study period in order to assess the representativeness of the sample. Reporting of stressful life events and social health issues was categorised according to whether women reported no social health issues, one or two issues, or three or more issues. Data on discrimination were dichotomised to distinguish women who reported any experience of discrimination from those who reported never experiencing any of the kinds of discrimination we asked about. Data were analysed using STATA version 11.0 [32] and involved the calculation of unadjusted and adjusted odds ratios and 95% confidence intervals. Multivariable logistic regression was used to assess the relationship between number of stressful life events and social health issues as the exposure of main interest (categorised as none, one to two issues, three or more issues) and low birthweight as the primary outcome variable. Four models are reported adjusting for: maternal smoking, social characteristics, number of antenatal visits and medical risk.

Results

Questionnaires were mailed to 8,597 women. The adjusted response fraction excluding a small number of questionnaires (n = 129) 'returned to sender', duplicate responses and women who gave birth outside the study period was 52% (4366/8468).

Social characteristics and risk factors

Table 1 shows the social and obstetric characteristics of participants compared with all Victorian and South Australian women who gave birth in the study period according to routinely collected perinatal data for both states. The mean age of women in the sample was 31 years (range 16 to 46 years). Comparisons with routinely collected data showed that women taking part in the survey were largely representative in terms of key obstetric characteristics including parity, method of birth, and infant birthweight, but included fewer women born overseas of non-English speaking background, Aboriginal and Torres Strait Islander women, single women and women under 25 years. Eighteen percent of women reported smoking in pregnancy, including women who said that they quit or cut down smoking while pregnant. One third of participants who provided data on height and weight prior to pregnancy were overweight (25.0-29.9) or obese (≥30). Two-thirds of women reported one or more stressful life events or social health issues during pregnancy, and 17% reported three or more issues.

Table 1 Social and obstetric characteristics of participants in the 2008 HMHF Survey and all women who gave birth in Victoria and South Australia during the study period*
Full size table

Stressful life events and social health issues

Table 2 summarises data on associations between maternal social characteristics and stressful life events/social health issues experienced in the 12 months before the birth. Maternal characteristics that were associated with substantially raised odds of reporting three or more stressful life events or social health issues were: being under 25 years; single, divorced or widowed; having a below average equivalised income; not completing year 12; smoking at any stage of pregnancy; being underweight (BMI <18.5); being at higher risk of obstetric complications, and being of Aboriginal and/or Torres Strait Islander origin.

Table 2 Relationship between stressful life events/social health issues, perceived discrimination and social and obstetric characteristics
Full size table

Experiences of discrimination in health care settings

Women were significantly more likely to report experiencing discrimination if they were: under 25 years, had a low income, had not completed year 12, were not married, had a very high (≥35) or a very low BMI (< 18.5), quit smoking in pregnancy or smoked throughout pregnancy and if they were at higher risk of medical complications. Women who reported three or more stressful life events/social health issues were significantly more likely to report discrimination (314/760, 41.3% versus 296/1429, 20.7%, OR = 2.69, 95% CI 2.2-3.3). Women who reported one or two stressful life events/social health issues also had significantly raised odds of experiencing discrimination (518/2123, 24.6% versus 20.7%, OR = 1.2, 95% CI 1.1-1.4) suggesting a 'dose response'.

Patterns of health service use

Despite the existence of universal health cover via Medicare, uptake of antenatal care varied in relation to maternal social characteristics and the number of stressful life events or social health issues women were coping with. Younger women, those who had below average household income, women who had not completed year 12, women without a partner and women who smoked throughout pregnancy were more likely to attend a first antenatal check-up at ≥13 weeks' gestation, and to attend less than five visits (Table 3). Women who reported stressful life events or social health issues were also more likely to attend late, and to attend fewer visits.

Table 3 Risk factors by pattern of health service use
Full size table

Associations with low birthweight

Univariable associations between maternal characteristics and low birthweight (< 2500) are shown in Table 4. Women at higher risk of obstetric complications in pregnancy had a fourfold increase in odds of having a low birthweight infant. Other maternal factors significantly associated with low birthweight were: not completing secondary education, being born overseas, and having a low pre-pregnant BMI (< 18.5). Women reporting three or more stressful life events or social health issues had a twofold increase in odds of having a baby with a low birthweight compared with women reporting no social health issues.

Table 4 Associations with infant birthweight
Full size table

In order to obtain a more precise estimate of the association between low birthweight and the number of stressful life events and social health issues, we developed the four models shown in Table 5. Maternal smoking status was fitted in all models for a priori reasons based on the known aetiological relationship between maternal smoking and infant birthweight. The first model adjusts for maternal smoking alone, and shows that the association between number of stressful life events and social health issues and low birthweight remains statistically significant after adjusting for maternal smoking in pregnancy. Model 2 adjusts for other socio-demographic factors associated with the exposure of main interest (three or more social health issues) and the outcome variable (low birthweight) in univariable analyses. Models 3 and 4 include smoking status and all of the socio-demographic variables included in model 2, and adjust separately for number of antenatal visits (model 3) and risk of complications in pregnancy (model 4). All models indicate a dose effect related to number of stressful life events and social health issues. All models that fitted maternal country of birth indicated that women born overseas in non-English speaking countries had significantly raised odds of having a low birthweight infant.

Table 5 Multivariable logistic regression analyses assessing associations with low birthweight (n = 3650)
Full size table

Discussion

Over 30 years ago Archie Cochrane noted that antenatal care had escaped critical assessment in terms of content [33]. There is still a paucity of research evaluating the effectiveness of antenatal interventions specifically targeting modifiable risk factors for poor maternal and/or child health outcomes [34–36]. Recent efforts to improve the evidence base for antenatal care have focused on clinical care, smoking cessation support and on assessment of recommended visit schedules [37–40]. The relationship between the content and organisation of antenatal care and subsequent maternal and child health outcomes has received scant attention [36]. For example, scientific knowledge about the extent to which different approaches to care (e.g. model of maternity care, pregnancy outreach workers) may influence nutrition, exercise habits, oral health, smoking, drug use, choices regarding infant feeding method, maternal depression or anxiety, intimate partner violence, inter pregnancy interval, and post pregnancy use of health services is negligible.

Our findings from a large Australian population-based survey of recent mothers attest to the high prevalence of potentially modifiable risk factors such as smoking and overweight and obesity. They also highlight a concerning level of social adversity associated with stressful life events and social health issues co-occurring with pregnancy. One in six women reported three or more stressful life events or social health issues in the 12 months before the birth. These women were much more likely to perceive that they were discriminated against in health care settings, to attend later in pregnancy and to have fewer antenatal visits. Women coping with multiple life events or social health issues remain significantly more likely to have a low birthweight infant when maternal smoking, number of antenatal visits, and other covariates are taken into account. Interpretation of the relationship between timing and number of antenatal visits and infant birthweight is complex. Our data show that some groups of women are less likely to access antenatal care early, and to attend regularly during pregnancy. The same groups of women are more likely to have adverse outcomes. This does not signify causation, but it does suggest that women who most need care and support during pregnancy are more likely to not to engage with services.

Several study limitations need to be considered. First, we used a modified life events scale originally developed for use in pregnancy [41] that has also been used in large scale maternity surveys in the United States and Canada [16, 30, 42]. Modifications made based on piloting and consultations preceding implementation of the HMHF Survey have not been subject to psychometric testing. Nor have we sought in this paper to differentiate different types of stressors. Our primary aim was to assess the association between low birthweight and number of stressful life events and social health issues and consider potential implications for the care of women during pregnancy. Consideration of broader questions regarding the relationship between different types of stress in pregnancy and area level deprivation are beyond the scope of analyses presented in the paper.

Second, it is plausible that women having a low birthweight infant may be more inclined to recall adverse life events in pregnancy than women who have an infant of higher birthweight, although this potential source of bias may have been lessened by the time elapsed between the birth and completing the survey at around six months postpartum. Third, we were not able to validate maternal recall of infant birthweight or maternal smoking in pregnancy. Validation studies comparing mothers' accounts of infant birthweight with medical records have shown high levels of agreement for this outcome as opposed to gestation, where there are more mixed findings [43, 44]. This was the reason for reporting infant birthweight, rather than small for gestational age. We acknowledge the potential for misclassification of maternal smoking status in early and late pregnancy. However, our finding that 17.6% of study participants reported smoking in pregnancy is in accord with routinely collected data available for seven Australian states and territories in 2007 which indicate that 16.6% of women smoked in pregnancy [25]. We used a multiple response format for seeking information about maternal smoking which has been shown to improve disclosure of maternal smoking in self-administered questionnaires [29]. Fourth, the sample while relatively large, provided low numbers in sub-groups for some comparisons (e.g. single women, Indigenous women).

Strengths of the study include a population-based sample drawn from two out of the eight Australian states and territories. The sampled population for the survey comprised 15% of confinements in South Australia and 8% in Victoria in 2007. While it is not possible to determine causal pathways in an observational study, the findings draw attention to social circumstances in women's lives that may place them at higher risk of poor perinatal and longer term outcomes. They also draw attention to vulnerable population groups who may be at additional risk.

Conclusions

Developing the infrastructure to support an integrated primary care approach to antenatal care in which services have the capacity to respond effectively to social circumstances in women's lives and to vulnerable populations is long overdue, and requires far reaching changes involving both hospital and community based services [45]. The recently released NICE clinical guideline on service provision for pregnant women with complex social factors provides an excellent starting point and practical recommendations focusing on four vulnerable population groups: pregnant women who misuse substances, young pregnant women, pregnant women experiencing intimate partner violence and pregnant women who are recent migrants, asylum seekers or refugees [46]. Another important area for intervention is pre-conception, especially during adolescence and during early adulthood. There are major challenges for policy makers and health services in developing an appropriately skilled workforce; development of a life-course approach to promotion of maternal, newborn and child health that includes a focus on public health intervention during pregnancy and prior to conception; development of team work and collaboration within and across agencies; facilitating professional, consumer and community participation; and investment in monitoring and evaluation to ensure outcome based accountability. A shift in this direction requires fundamentally different thinking and co-ordinated action across sectors to bring about system wide changes [47, 48]. Only if this challenge is embraced will health care services be capable of developing effective proactive approaches to potentially modifiable risks and health needs in diverse populations.