• Open Access

Maternal smoking during pregnancy among Aboriginal women in New South Wales is linked to social gradient

Authors


Correspondence to: Prof. Adrian Bauman, Prevention Research Collaboration, Level 2, Medical Foundation Building K25, University of Sydney, Camperdown NSW 2006; e-mail: adrianb@health.usyd.edu.au

Abstract

Objective: Social gradients in Aboriginal health are seldom explored. This study describes social gradients and trends in smoking during pregnancy among Aboriginal mothers in NSW.

Methods: This was a secondary analysis of the NSW Midwives Data Collection (MDC) 1994–2007, covering all births in NSW. Analyses examined associations between socio-demographic characteristics and smoking during pregnancy.

Results: Data from 1,214,206 pregnant women showed that 17.4% smoked during pregnancy. The rate of smoking during pregnancy among all NSW women declined from 22.3% in 1994 to 12.8% in 2007; the rate among Aboriginal women remained high, declining from 61.4% in 1994 to 50.2% in 2007. Smoking was substantially higher among Aboriginal mothers compared to non-Aboriginal mothers. Socio-economic analyses showed that the smoking rate among low SES Aboriginal mothers was approximately two and a half times that of high SES Aboriginal women, a similar gradient to non-Aboriginal women.

Conclusions: Indicators of socio-economic position are a consistent, independent correlate of smoking during pregnancy for Aboriginal and non-Aboriginal women.

Implications: There is a need for a social inequalities approach to smoking during pregnancy, specifically targeting more disadvantaged Aboriginal mothers and all teenage mothers for smoking prevention. Strategies to access more disadvantaged mothers should not be missed through broadly focused Aboriginal tobacco control strategies.

Extensive efforts by government and healthcare providers have increased awareness of the consequences of smoking during pregnancy (SIP). However, it remains a significant public health issue and affects both maternal and infant health.1 The independent, adverse effects of SIP include increased risk of fetal mortality, spontaneous abortions, miscarriage, still birth, ectopic pregnancy, neonatal morbidity and mortality, pre-term birth, low birthweight, small head circumference, low APGAR scores, and sudden infant death syndrome (SIDS).2–9 However, the adverse effects of SIP extend beyond birth, including subsequent problems in childhood.10–13

Tobacco is the most prevalent substance used during pregnancy,14 and SIP is the leading preventable cause of infant mortality in most developed countries.4,15 Importantly, there is strong evidence that smoking cessation during pregnancy can prevent and reverse the adverse effects of smoking in the mother and child.16 However, SIP rates remain high with between one in seven and one in three women smoking during pregnancy in developed countries.4,14,17 For all NSW women, the correlates of higher SIP rates include younger maternal age, lower socio-economic status and Aboriginality of the mother.14 These subgroups report SIP at a rate up to four times that of the total NSW population.14 Importantly, it has been shown that SIP posed equivalent epidemiological risks on Aboriginal and non-Aboriginal mothers and babies, thus it is smoking specifically that causes adverse outcomes.18

Maternal socio-demographic factors, including education level, maternal age, socio-economic status (SES) and lack of private health insurance, are important correlates of SIP and adverse infant health outcomes.14,19–21 Smoking patterns among women show a well-established social gradient.1,22,23 The strength of the social gradient varies across populations, for example, in Sweden the gradient is less marked than in the USA or Australia.23 SIP exhibits the same social gradient,23,24 with previous studies suggesting a widening of socio-economic inequalities in rates of SIP.25–28

In Australia, tobacco use among Aboriginal women remains disproportionately high, with rates of SIP more than three times as likely as non-Aboriginal women.29,30,52 Between 2001 and 2004, the National Perinatal Data Collection reported that 51% of Aboriginal women smoked during pregnancy compared to 16% of non-Aboriginal women.30,31 Over the past decade, SIP has appeared to declined in the general Australian population overall, but still remains high among Aboriginal women.14,30–32,52,53 In addition, rates of health outcomes are related to smoking in pregnancy among Aboriginal mothers, with low birthweight 70% higher than among non-smoking Aboriginal mothers.53

Although there is clear evidence of a social gradient in health-compromising risk factors,30 few studies have considered the existence of social gradients in health within the Aboriginal population. Generally, Aboriginal populations are classified together as a single group in considering health risks and outcomes, compared to the non-Aboriginal population. Social determinants research has noted that the most socio-economically disadvantaged Aboriginal people are most likely to smoke.33 There is a strong age-related gradient in Indigenous smoking, with teenage Indigenous mothers showing the highest smoking rates.31,53

This study examines the social gradient in Aboriginal SIP using the New South Wales Midwives Data Collection.45 An improved understanding of the social gradient in SIP in the Aboriginal population may enable policy makers to more effectively target sub-groups within the Aboriginal population, and enable researchers to better understand the social and economic causes of sustained Aboriginal smoking rates.

Methods

Data Source

The study population included women who gave birth in New South Wales from 1 January 1994 to 31 December 2007. Data were obtained from the New South Wales Midwives Data Collection (MDC) held by the Centre for Epidemiology and Research, Population Health of the NSW Health Department.34,45 The MDC is a legislated State-wide surveillance system that covers all births in NSW public and private hospitals, and home births of at least 20 weeks gestation or at least 400 grams birthweight.34 The NSW MDC comprises all births in NSW and around 30,000 Aboriginal births between 1994 and 2007. The NSW MDC has shown good agreement between MDC data and information obtained directly from medical records for the majority of data items, with high specificity, sensitivity and kappa coefficients 0.70 or higher.35,36

Measures

Pregnant women were classified as smokers or non-smokers according to whether or not they self-reported smoking at any time during their current pregnancy. Validation of self-reported smoking status in the Aboriginal population and in pregnant women in particular show reasonable measurement properties and are reported elsewhere.35,37–39 However, one recent validation paper suggests that passive smoking exposure, in particular, may be more under-reported among pregnant Aboriginal women, suggesting caution in self-reported smoking exposure measurement.50

The socio-economic status (SES) of pregnant women was determined from the Australian Bureau of Statistics SEIFA Index of Relative Socio-economic Disadvantage (IRSD). SES was assigned based on the mother's postcode of residence and was calculated using information of the proportions of households with low income and low educational attainment, high unemployment, rented dwellings and individuals not fluent in English.35,40 SEIFA was categorised into the highest SES quintile, the middle two quintiles, and the lowest two SES quintiles. SEIFA is an ‘area’ or aggregate measure of socio-economic disadvantage, and not an individual-level measure [which were not available in the dataset]. The ARIA [Accessibility remoteness index of Australia] was used to assess urban and rural differences.56

Aboriginal and Torres Strait Islanders were defined in the MDC by self-reported status and were counted as one group in the MDC up to 1997, and are combined in this paper. Under-enumeration of Aboriginal people in NSW health datasets can introduce bias in the study results.34 The reliability of self-reported Aboriginal status of the MDC was assessed in 1998, at which time there was excellent agreement between MDC data and information obtained directly from medical records (98%).36 However, enumeration of Aboriginal status may not be as good in some settings51 and the accuracy of reporting may be better in rural than urban maternity data.44,45

Data analyses

We assessed trends in SIP in NSW during the period 1994 to 2007. First, we determined the crude prevalence of SIP and the distribution of the study population by Aboriginality and socio-demographic factors, including maternal age and socio-economic status. Second, multivariate logistic regression was used to examine the relationship between socio-economic status and self-reports of SIP while controlling for year of delivery and an individual's demographic characteristics of maternal age and Aboriginal status. Adjusted odds ratios (OR) and 95% confidence intervals (CI) were used to examine the social gradient in SIP in Aboriginal women. Data were analysed using SAS Version 9.1 (SAS Institute Inc, Carey, North Carolina), with the regression analyses performed using PROC LOGISTIC.

Results

Of the 1,234,777 women confined in NSW between 1994 and 2007, smoking status, Aboriginality and maternal age was recorded for 1,214,206. Among them, 211,482 were smokers (17.4%). The demographic characteristics of these women are presented by age, socio-economic status and Aboriginality, with 2.4% of confinements being among Aboriginal women. The relationships between various maternal characteristics and SIP are shown in Table 1. Between 1994 and 2007, the SIP rate among all NSW women declined from 22.3% in 1993 to 12.8% in 2007. Rates were much higher among teenagers (42.7%), compared to older mothers. High rates of smoking were reported by Aboriginal women (57.3%), who were more than five times as likely to smoke compared to non-Aboriginal women (OR=5.10, 95% CI: 4.97–5.22). Low socio-economic status also increased the risk of maternal smoking during pregnancy.

Table 1.  All NSW women: socio-demographic correlates of smoking during pregnancy.
Socio-demographic characteristicsTotal number of women (%)% of smokers (n=211,482)Adjusteda odds ratios (95% CI)
  1. Notes:

  2. (a) Dependent variable: Smoking status (smoker = 1, non-smoker = 0), Explanatory variables: Year of delivery, maternal age, SES and Aboriginal status.

  3. * p < 0.05

Year of delivery
1994–20071,214,20617.42-
199485,19922.281.00
199585,70221.410.95 (0.92–0.97)*
199684,80620.970.92 (0.90–0.95)*
199786,52520.500.90 (0.88–0.92)*
199884,70219.700.86 (0.84–0.88)*
199986,55818.830.79 (0.78–0.81)*
200087,17317.210.71 (0.70–0.73)*
200185,14916.960.70 (0.69–0.72)*
200285,22516.250.67 (0.65–0.69)*
200385,62415.060.61 (0.60–0.63)*
200484,79014.700.61 (0.59–0.62)*
200589,34414.270.58 (0.57–0.60)*
200689,48213.680.55 (0.54–0.57)*
200793,92712.780.51 (0.50–0.52)*
Maternal age (years)
Under 2053,363 (4.39)42.723.84 (3.76–3.93)*
20–34942,259 (77.60)17.441.42 (1.40–1.44)*
35+218,584 (18.00)11.131.00
SES
Lowest SES489,833 (40.34)23.053.33 (3.27–3.39)*
Moderate SES483,300 (39.80)16.952.44 (2.39–2.48)*
Highest SES241,073 (19.85)6.911.00
Aboriginal status
Non-Aboriginal1,184,584 (97.56)16.421.00
Aboriginal29,622 (2.44)57.325.10 (4.97–5.22)*

Aboriginal-specific data are shown in Table 2. The reported number of Aboriginal mothers giving birth increased from around 1,700 in the mid-1990s to more than 2,500 in most recent years. Part of this increase could be due to increased identification as Aboriginal. Although Table 1 shows a 50% decline in the likelihood of SIP among non-Aboriginal women between 1994 and 2007, we also note a 36% decline among Aboriginal mothers (Table 2). These declines in smoking prevalence seemed to start around 2000–2002 among Aboriginal women, with non-Aboriginal women showing declines around five years earlier, starting in 1995 (interaction between year and Indigenous status, p<0.001). Although there was a strong age-related gradient in SIP among non-Aboriginal women, this age-relationship was much less apparent among Aboriginal women, where SIP rates ranged from 55.7% to 59.2% by age.

Table 2.  Aboriginal women: socio-demographic correlates of smoking during pregnancy.
Socio-demographic characteristicsTotal number of women (%)% of smokers (n=16,979)Adjusteda odds ratios (95% CI)
  1. Notes:

  2. a. Dependent variable: Smoking status (smoker = 1, non-smoker = 0), Explanatory variables: Year of delivery, maternal age and SES.

  3. *p<0.05

Year of delivery
1994–200729,62257.32
19941,48961.381.00
19951,71861.180.99 (0.86–1.15)
19961,69561.831.02 (0.88–1.18)
19971,82761.411.00 (0.87–1.15)
19982,02558.670.90 (0.79–1.03)
19992,05558.880.89 (0.78–1.03)
20002,10056.190.80 (0.70–0.92)*
20012,11558.960.90 (0.78–1.03)
20022,15957.990.87 (0.76–0.99)*
20032,17256.950.83 (0.73–0.95)*
20042,31556.630.83 (0.72–0.94)*
20052,48155.300.78 (0.69–0.89)*
20062,58454.020.75 (0.66–0.85)*
20072,88750.230.64 (0.57–0.73)*
Maternal age (years)
Under 206,164 (20.81)59.211.08 (0.98–1.19)
20–3421,353 (72.08)56.921.00 (0.91–1.09)
35+2,105 (7.11)55.721.00
SES
Lowest SES20,029 (67.62)59.272.64 (2.22–3.14)*
Moderate SES9,017 (30.44)54.402.17 (1.82–2.59)*
Highest SES576 (1.94)34.901.00

By socio-economic status, there was an expected dose response gradient in SIP among non-Aboriginal women (Table 1), but there was also evidence of a gradient among Aboriginal women, with a greater relative difference observed among Aboriginal women. The prevalence of SIP ranged from 34.9% to 59.3%, with low SES Aboriginal women over two and a half times more likely to smoke during pregnancy compared to high SES Aboriginal women (OR=2.64, 95% CI: 2.22–3.14). Interestingly, this social gradient was stronger for Aboriginal mothers resident in urban centres [ARIA≤0.2], with smoking in pregnancy across low, mid and high SES (56.8%, 48.8% and 32.3%, respectively), compared to SES gradients in rural areas [ARIA >0.2; SIP rates 60.1%, 57.3% and 48.3% respectively]. In adjusted analyses, low SES rural Aboriginal women more likely to smoke compared to high SES rural Aboriginal women (OR=1.54, 95% CI: 1.03–2.32), and low SES urban Aboriginal women even more likely to smoke compared to high SES urban Aboriginal women (OR=2.69, 95% CI: 2.20–3.29).

The prevalence ratio, defined as the ratio between the prevalence of SIP in low SES women and the prevalence of SIP in high SES women, has increased in both Aboriginal and non-Aboriginal women over time. For Aboriginal women, the prevalence of SIP was 1.55 times higher for low SES women than high SES women for the period 1994 to 1998, increasing to 1.82 times higher for the period 2004 to 2007. For non-Aboriginal women, the prevalence of SIP was 2.46 times higher for low SES women than high SES women for the period 1994 to1998, increasing to 4.75 times higher for the period 2004 to 2007, suggesting widening inequalities among all women.

Finally, these data indentified an increased risk of SIP for teenagers, regardless of their Aboriginality or socio-economic status (rates of 59% and 41%). Stratifying by socio-economic status, smoking prevalence remained high for teenage Aboriginal mothers (60.0%, 57.3% and 59.5% for low, moderate and high SES mothers, respectively). For teenage non-Aboriginal mothers, prevalence rates were 40.3%, 41.9% and 34.3% for low, moderate and high SES mothers, respectively. Figure 1 shows the social gradient in SIP for Aboriginal and non-Aboriginal women, with high rates and less social gradient in SIP evident for teenagers, regardless of Aboriginality.

Figure 1.

Smoking during pregnancy by Aboriginality, maternal age and socio-economic status.

Discussion

Poor maternal and infant health among Aboriginal mothers who smoke is a pressing public health problem.31 Similar to other studies, we found that the prevalence of SIP was higher among Aboriginal women than non-Aboriginal women,14,29–33 but these data showed a delayed decline among Aboriginal compared to non-Aboriginal women. While it is encouraging to see long-term reductions in smoking among Aboriginal women, rates remain higher, especially among low SES Aboriginal women, and there is ongoing evidence of widening socio-economic inequalities. The reasons behind these declines and behind the five-year delay in the onset of the decline among Aboriginal mothers are not clear. It seems that mainstream tobacco prevention efforts, rather than any specific major smoking in pregnancy initiative, led to gradual adoption of new social norms (about not smoking in pregnancy), which diffused into Aboriginal culture and health services slightly more slowly.

While SIP and socio-economic differences in smoking prevalence have long been observed, it is only recently that the influence of socio-economic status and other factors affecting SIP have begun to be observed.14,41 This study's unique feature was to identify that Aboriginal mothers are not a homogenous group. Different groups within the Aboriginal population have different smoking behaviour during pregnancy, although the prevalence of SIP remains high in all groups. The most socially disadvantaged Aboriginal women are the most likely to smoke during pregnancy and these inequalities in Aboriginal women have widened over the past decade. As socio-economic disadvantage decreases, prevalence of SIP decreases from 59% in low SES Aboriginal women to 35% in high SES Aboriginal women, an absolute difference of 24%. From this, we can conclude that variables reflecting social disadvantage contribute to SIP rates among both Aboriginal and non-Aboriginal women. The presence of a social gradient in Aboriginal mothers draws attention to the need to consider the diverse nature of Aboriginal populations who have far less favourable socio-economic and health experiences.

While other reports document remoteness of residence as a predictor of smoking for Aboriginal people,42 we showed that indicators of socio-economic position are a consistent and powerful predictor of SIP for Aboriginal women, and that inequalities in SIP have widened more among urban than rural Aboriginal mothers. We recognise the potential limitation of our aggregate area-level measure of socio-economic status, which does not consider individual-level measures of deprivation, poverty or isolation. However, a recent Australian paper showed that area-level disadvantage independently contributed to explaining smoking and other risk factors.46 Aggregate SES measures assess a local area's characteristics, which are consistently related to health outcomes,48 and assessment of ‘place’ factors often contribute substantially more to health outcomes than individual-level SES measures.47,49

The findings of smoking in pregnancy among teenage mothers is consistent with other studies.43 Teenagers who are pregnant show higher levels of risk taking, including smoking, across cultural and social groups. With Aboriginal women are more likely to have children at a younger age than non-Aboriginal women, and to show earlier onset of smoking irrespective of pregnancy status,30,53 this is cause for considering age-specific programs for this at-risk group.

Our study has several limitations, including using cross-sectional data that can only identify associations, not describe the socio-economic causes of smoking in pregnancy. Next, self-report of Aboriginality and smoking status may underestimate true rates. Aboriginal mothers who were not enumerated (that is, did not disclose Aboriginal status in MDC data) may influence the true social gradient in Aboriginal smoking rates. This misclassification bias is likely to be differential by cultural grouping, suggesting these data may under-estimate true associations. This issue of under-reporting of Aboriginality in the NSW Midwives data set has been recently explored44,45 for NSW mothers, under-reporting of Aboriginality seems more likely in urban settings [across socio-economic regions], and less of a problem with better enumeration in rural NSW regions. A more important problem, that cannot be overcome in these data, is under-reporting of Aboriginal fathers, where the mother is non-Aboriginal.50 Both these factors could lead to under-enumeration, but the relationship of this to the affluence of defined regions is complex and its influence on SIP rates is not known.

Another methodological limitation of the MDC is the de-identification of mothers; therefore mothers could appear twice or more over several years, a bias which would be more frequent among Aboriginal mothers as they show higher fertility rates. Despite these limitations, our findings are important as they provide new information on social gradients in smoking among Aboriginal mothers in NSW.

This study confirms the importance of addressing the social determinants of smoking during pregnancy, and the need for socio-economic factors, based on an area-level analysis to be considered. For example, the social gradient appears greater among urban Aboriginal mothers, so the targeting of programs may be more required, whereas in rural regions, more universal rates of smoking in pregnancy may suggest more homogeneous approaches to Indigenous mothers. The smoking in pregnancy issue has been identified in policy-relevant planning53 and embodied in national prevention plans.54,55 These may be expressed at a macro level, or as local community level programs such as ‘Strong mothers, Strong babies’.54 These approaches need to be consistent with, but a focused part of the National preventative partnership work, which has both tobacco and Indigenous health priorities.55 Finally, this research also identified the special problem of smoking among teenagers who are pregnant, with higher rates of smoking irrespective of Aboriginality, and needing age-targeted solutions.

Acknowledgements

Aaron Thrift was employed by the NSW Department of Health on the NSW Biostatistical Officer Training Program at the time this work was conducted. Thanks to the NSW Health, Centre for Epidemiology and Research, Population Health, for providing data access. This paper was approved for submission by the AHMRC.

Ancillary