Bibbulung Gnarneep (‘solid kid’): Causal pathways to poor birth outcomes in an urban Aboriginal birth cohort


Professor Sandra Eades, The Sax Institute, PO Box 123, Broadway, NSW 2007, Australia. Fax: +612 9514 5951; email:


Aim:  To explore the causal pathways leading to poor birth outcomes among a cohort of Aboriginal infants.

Methods:  A cross-sectional study was conducted. Data were collected via face-to-face interviews conducted 6–12 weeks post partum and links to the WA Midwives' Notification System. Two hundred and seventy-three Aboriginal infants and their families from Perth, Western Australia were recruited in the mid to late 1990s.

Results:  Poor birth outcome was defined as low birth weight and/or preterm birth. Criteria for poor birth outcome were met by 12.3% of the cohort. A history of maternal hypertension, vaginal bleeding and consumption of excess spirits in pregnancy were independent predictors of poor birth outcome. Mother being raised on a mission, maternal education, smoking during pregnancy and being exposed to passive smoke during pregnancy were also important risk factors.

Conclusions:  Results indicate that maternal social exposures, maternal ill-health before the index pregnancy and maternal ill-health during the index pregnancy are all important contributors to poor birth outcomes for urban Aboriginal infants. While the causes of poor birth outcomes are complex, the current study highlights several areas where preventive measures may be useful.

Numerous studies indicate that Aboriginal women are significantly more likely than other Australian women to experience poor pregnancy outcomes. While the causes of this inequality are undoubtedly complex, previous studies of obstetric outcomes among Aboriginal women in major urban centres1–3 have restricted their focus to conventional perinatal risk factors, providing limited insights into how best to target interventions to improve pregnancy outcomes for urban Aboriginal women.

Bibbulung Gnarneep (‘solid kid’), a cohort study of Aboriginal women and their babies, was designed to provide a greater understanding of the factors which influence the health of Aboriginal infants. Conducted in collaboration with the Perth Aboriginal community, Bibbulung Gnarneep is the only population-based study to explore the issue of poor birth outcomes among Aboriginal women. The study also represents a significant advance on previous designs because of its comprehensive exploration of proximal, medial and distal factors of unique relevance to Aboriginal women.

Bibbulung Gnarneep was driven by the idea that shifts in the systems which organise and systematically influence the health risks that arise for Aboriginal people need to be a major part of our response to improvement of Aboriginal health. As such, epidemiological modeling and other studies which enhance our understanding of the impact of systematic influences on obstetric and other health outcomes for Aboriginal women are urgently required to drive and inform these shifts.

With this in mind, a causal pathways approach was chosen for the Bibbulung Gnarneep study. A causal pathway is a sequence of events or conditions culminating in the outcome or disease of interest, in which the effect of any step is dependant on the presence of other steps. Causal pathways are particularly useful in helping us consider opportunities for prevention. By elucidating the whole causal pathway, the choice of preventive strategies is increased, enabling the most effective to be chosen. Causal pathways models also assist us to avoid concentrating on limited time periods when optimal time for successful prevention may have passed. This approach is advantageous because of its ability to provide the practical information which highlights broader associations of many diseases with social or economic factors, without then controlling for them with analyses which focus on more proximal factors in the causal pathway, some of which could be seen as early signs of the disease or outcome of interest. The benefit of this approach is to target the public health response to tacking individual behaviour change and to changing the societal situations which provide the environments in which risk arises.4

The current study thus aimed to explore the associations between a number of exposure variables and poor birth outcomes, specifically low birth weight and/or preterm birth, in the Bibbulung Gnarneep cohort using a causal pathways framework.

Key Points

  • 1Poor birth outcomes (low birth weight and/or preterm birth) occurred in 12% urban Aboriginal births in this cohort study.
  • 2Maternal pregnancy health issues including excess alcohol use, hypertension and vaginal bleeding predicted a poor birth outcome and provide opportunities for intervention and prevention.
  • 3Maternal education and smoking during pregnancy are also important risk factors for a poor birth outcome and provide opportunity for intervention and prevention.


Bibbulung Gnarneep is a prospective cohort study of Aboriginal mothers and their infants residing in Perth, Western Australia. Mothers were identified as Aboriginal and Perth residents from the WA Midwives' Notification System.5 All eligible mothers (who could be contacted) who had given birth during a 15-month period in the mid to late 1990s were invited to participate in the project. Fully informed consent was obtained from each mother and approval for the study was given by the Institutional Ethics Committee of Princess Margaret Hospital for Children and the Confidentiality of Health Information Committee at the Health Department of Western Australia. The data collection phase of the project comprised five face-to-face interviews with mothers. All survey data presented in the current study was collected as part of the first interview, occurring when the infant was 6–12 weeks old. In addition, birth weight and gestational age data were collected from the WA Midwives' Notification System form for each infant.5 Information was collected on a wide range of medical, demographic, social, economic, behavioural and family variables. A community reference group was established to guide the study6 and the questionnaires were developed following a series of community consultations. Further information on data collection can be found elsewhere.7

Birth outcomes

A poor birth outcome was the outcome selected for analysis in this paper. A poor birth outcome was defined as an infant of low birth weight (<2500 g) or preterm (<37 weeks of gestation) or both of these. Birth weight and gestational age were obtained from the WA Midwives Notification System birth record for each infant (Table 1).

Table 1.  Infant birth outcomes
 n (%)
Poor birth outcome
 Low birth weight only8 (3.0)
 Gestational age < 37 weeks only9 (3.4)
 Both low birth weight and preterm16 (6.0)
Normal birth outcome
 Neither low birth weight or preterm234 (87.7)
 Total267 (100)

Exposure variables

All information on exposures was collected from the face-to-face interview conducted 6–12 weeks postpartum. The interviews were all conducted by Aboriginal women from the community involved. Most of the interviewers had Health Worker qualifications. Measurement of alcohol intake during pregnancy was assessed with a table which asked women to indicate their usual alcohol intake during pregnancy for beer, wine, fortified wine and spirits. The table assessed frequency of alcohol use and the number of standard drinks consumed on any occasion. Women who consumed more than two standard drinks on any occasion were classified as consuming excess amounts of the specified drink. Measurement of smoking during pregnancy asked women if they smoked cigarettes during pregnancy and the number of cigarettes they usually smoked each day. Questions regarding the women's experience of health problems during pregnancy asked women to indicate if they experienced bleeding or repeat bleeding in the first 20 weeks of pregnancy and the second 20 weeks of pregnancy. The numbers of women in each of the four categories were small, so a new variable called ‘any bleeding in pregnancy’ was reported. Initial descriptive analysis using χ2 tests8 showed that a number of these exposure variables were potentially associated with the selected birth outcomes. These variables were then analysed by univariate logistic regression9 and results from these analyses were used to construct multivariate logistic regression models. There were minor differences found between univariate and multivariate analyses adjusted for maternal smoking and maternal education; therefore, the multivariate models are not quoted in reporting the findings in this paper.10 The univariate and multivariate steps in causal pathway modeling analysed risk factors for specified outcomes based on knowledge of the literature and integrated risk factors Indigenous community members perceived to be important determinants of these outcomes. SAS (Statistical Analysis Software, Version 6.09, SAS Institute Inc. Cary, NC, USA) was used for all data entry, manipulation and analysis.

Study cohort

Using the WA Midwives Notification System, characteristics of the mothers and infants who enrolled in this cohort study were compared with the total number of Aboriginal mothers who lived in the selected urban geographic region and gave birth during the recruitment period for the cohort. χ2 tests were used for comparisons.8


Of the total eligible population, 273 (53%) mothers who gave birth during the recruitment period enrolled in the Bibbulung Gnarneep project and completed the first interview. The cohort characteristics have been reported previously.7 However, compared with all Perth Aboriginal women who gave birth during the same period, a smaller proportion of the mothers studied were teenagers (15% compared with 21%) and a smaller proportion had just given birth to their first infant (20% compared with 23%). The study infants were also less likely to have been born preterm (9% compared with 13%) and less likely to have been of low birth weight (10% compared with 14%).

Unadjusted logistic regression models of the factors influencing the occurrence of a poor birth outcome showed similar results as logistic regression models adjusting for maternal smoking in pregnancy and maternal education. Therefore, the unadjusted odds ratios (ORs) are reported in Table 2.

Table 2.  Logistic regression models of factors among mothers influencing the occurrence of low-birth-weight births and/or preterm births
 n (%)Unadjusted OR (95% CI)
  • Other forms of care excluded.

  • ‡Mother, partner or other adults living in the household smoked after baby's birth.

  • §

    §Mother usually consumed more than two standard drinks containing spirits on any occasion. CI, confidence interval; OR, odds ratio.

Maternal social exposures
 Raised in mission/institution versus own parents care8 (4.8)5.0 (1.1–22.7)
 Completed less than year 10 at secondary school62 (24.3)1.8 (0.8–3.9)
 Has a relatively young mother (aged <41 years old)42 (24.4)1.6 (0.6–4.1)
 Alcohol use of family members caused hardship56 (22.1)2.1 (0.9–4.6)
Maternal health before pregnancy
 Not aware of the use of periconceptional folate112 (43.9)1.8 (0.8–3.7)
 High blood pressure before pregnancy10 (3.9)8.4 (2.3–30.9)
 Any medical problem before pregnancy128 (50.2)1.8 (0.8–3.8)
Maternal health during index pregnancy
 Mother smoked164 (64.3)2.2 (0.9–5.2)
 High blood pressure needing treatment10 (3.9)1.8 (0.4–8.9)
 Any medical problem213 (83.5)3.3 (0.8–14.2)
 Any vaginal bleeding35 (13.7)2.4 (1.0–6.0)
 Lived in a house where others smoked205 (83.3)3.2 (0.7–14.0)
 Consumed excess§ spirits during pregnancy33 (13.4)4.2 (1.7–10.1)
 Late (>20 weeks) or no antenatal care42 (17.1)1.4 (0.6–3.6)
 Prolonged rupture of membranes54 (21.3)1.3 (0.6–3.2)

The OR for a poor birth outcome among mothers who were raised in a mission/institution compared with mothers raised by their own parents was 5.0 (95% confidence interval (CI) 1.1–22.7). Mothers from circumstances where family members' alcohol use caused hardship were twice as likely to have a poor birth outcome (OR 2.1, 95% CI 0.9–4.6). Maternal illness before pregnancy increased the risk of a poor birth outcome in mothers who reported having high blood pressure (OR 8.4, 95% CI 2.3–30.9) and who reported any medical problem (OR 1.8, 95% CI 0.8–3.8). Similarly, maternal illness during the index pregnancy increased the risk of a poor birth outcome to three times the risk (OR 3.3, 95% CI 0.8–14.2). Maternal smoking during pregnancy (OR 2.2, 95% CI 0.9–5.2) and maternal consumption of more than two standard alcohol drinks containing spirits on any occasion (OR 4.2, 95% CI 1.7–10.1) were also important risk factors for a poor birth outcome. Passive exposure to tobacco smoke during pregnancy was also important with living in a house where others smoked increasing the risk of a poor birth outcome (OR 3.2, 95% CI 0.7–14.0).


This study is notable for the extreme level of social disadvantaged experienced by participating mothers and infants, with 24.3% completing less than year 10 at secondary school, 22% reporting family members' use of alcohol caused them hardship and 24% of infants having relatively young maternal grandmothers. In addition, the study is notable for providing quantitative evidence about the social and economic factors, as well as medical and obstetric, that have a substantial impact on the birth outcomes of Aboriginal children in a large urban community. Such data are rare, the majority of studies in this area of research being qualitative or anecdotal.11,12 Consequently, while poor obstetric outcomes have been repeatedly described among Aboriginal children in both Western Australia5,13,14 and other states,1–3,15 this information has rarely been linked to the broad ranging disadvantage known to be experienced by the Aboriginal population in other areas.16–18 Being raised in a mission or institution, high blood pressure before the index pregnancy, vaginal bleeding during the pregnancy and consumption of alcohol containing spirits at hazardous levels, were all shown in unadjusted analyses to be significantly related to birth outcome among the Bibbulung Gnarneep cohort. For other factors such as level of high school education completed, other medical problems before the index pregnancy, passive household smoke exposure during pregnancy, maternal smoking and prolonged rupture of membranes, this study did not establish a statistical association with the outcomes of interest. However, they have a number of theoretically plausible causal associations with the outcomes of interest and could be re-examined in future studies.

Our causal pathway model using appropriate, relevant and valid data clarifies the actions required if we aim to improve infant and child outcomes for the population studied. The more distal social factors must be targeted where possible. This paper highlights the importance of care arrangements for children not raised by their own parents, the importance of high school level educational outcomes for Aboriginal women and the possible impact of the hardships caused by teenage parenting for mothers in this cohort whose mothers were relatively young themselves. This study also highlights the important impact of medial factors such as the health of women during child-bearing years on poor birth outcomes. The impact of many of the health problems women experienced was amenable to effective health interventions, such as those women who reported high blood pressure during pregnancy suggesting the potential value of strengthening intervention efforts for this group. Further, this study demonstrates the importance of proximal factors with health status of the mother during the index pregnancy having an impact on birth outcomes.

The findings of this study have direct relevance to the Productivity Commission's report on ‘Overcoming Indigenous Disadvantage’ which provides indicators of Indigenous disadvantage that are of relevance to all governments and that can demonstrate the impact of programme and policy interventions.19 The Productivity Commission report 2003 highlights a number of areas for strategic action as important in changing headline indicators such as life expectancy at birth and year 10 and 12 retention and attainment over time. However, the major gap in these strategic action areas as highlighted by this study is the provision of access to effective primary health-care systems. For example, until education improvements occur, intermediate indicators identified by the Productivity Commission report such as birth weight can be modified by health interventions which treat existing health issues such as infections, high blood pressure and anaemia, and the introduction of interventions to reduce the harmful impact of tobacco and alcohol consumption in pregnancy.

There is a paucity of research investigating health outcomes for Aboriginal children residing in metropolitan areas,15 although the need for such studies has been recognised.19,20 Hence, the Bibbulung Gnarneep study has special value, particularly as the study was based on the total population of births to Aboriginal mothers residing in the Perth metropolitan area and the final sample provided an adequate representation of families from all suburbs.6 Another major strength of this study was the fact that it was driven by extensive community input. The establishment of the community reference group6 was particularly valuable as it allowed us to translate issues of community concern into the standardised questionnaires from the early planning stages and provided practical guidance throughout the project.

There are some limitations to the study: first, the sample size is not large enough to achieve statistical significance to establish the presence or absence of an association between the outcomes of interest and the risk factors in the postulated causal pathways; second, most of the information obtained in this study was obtained directly from the mothers and not validated by accessing other material. However, with regard to the latter point, internal validity of the questionnaires was demonstrated by having several questions relating to the same individual variables. In addition, for many questions, it is a strength of the study that the mothers' knowledge and perceptions have been used11,21,22 given the nature of the issues explored. For example, mothers are the best judges of the daily difficulties they face, including those related to alcohol.

The other major limitation of this study involves the definition of poor birth outcomes. Specifically, because of the survey methodology applied to this study, we were not able to define which infants had intrauterine growth restriction. Intrauterine growth restriction has an array of causal pathways resulting in the foetus not achieving optimal growth. To achieve good estimates of intrauterine growth, it is essential to have accurate dating of the gestation of the foetus/infant, preferably by early and serial ultrasound dating. As this was not possible in this study, the use of low birth weight and preterm birth as proxy measures for intrauterine growth restriction may have resulted in the effect of relevant exposures being under-estimated. On the other hand, studies rigorously designed to collect data enabling accurate definition of intrauterine growth are rarely able to collect the breadth of community level exposure data that has been outlined in this study. An important future development may be to combine these two study methods to improve our understanding of the causal pathways for intrauterine growth restriction and preterm birth.

Finally, if we return to our idea of what makes a ‘solid kid’ (the premise of the Bibbulung Gnarneep study), we must turn all the risk factors around. So our ‘solid kid’, in relation to birth outcomes, was born of a well-educated mother who had no medical or obstetric problems. Their mother had not been raised on a mission or in an institution and our ‘solid kid’ was not exposed to cigarette smoke or the effects of alcohol in utero. This picture does not seem too complicated, but because of the long history of disempowerment and removal from family, creating the solid kid will not be achieved without maximum effort from all involved.