Article Text

Download PDFPDF

Disparities in severe neonatal morbidity and mortality between Aboriginal and non-Aboriginal births in Western Australia: a decomposition analysis
  1. Akilew A Adane1,2,
  2. Helen D Bailey1,
  3. Rhonda Marriott2,
  4. Brad M Farrant1,
  5. Scott W White3,4,
  6. Carrington C J Shepherd1,2,5
  1. 1 Telethon Kids Institute, The University of Western Australia, Nedlands, Western Australia, Australia
  2. 2 Ngangk Yira Research Centre for Aboriginal Health and Social Equity, Murdoch University, Murdoch, Western Australia, Australia
  3. 3 Division of Obstetrics and Gynaecology, The University of Western Australia, Nedlands, Western Australia, Australia
  4. 4 Maternal Fetal Medicine Service, King Edward Memorial Hospital for Women Perth, Subiaco, Western Australia, Australia
  5. 5 Curtin Medical School, Curtin University, Bentley, Western Australia, Australia
  1. Correspondence to Dr Akilew A Adane, Ngangk Yira Research Centre for Aboriginal Health and Social Equity, Murdoch University, Murdoch, Western Australia, Australia; akilew.adane{at}murdoch.edu.au

Abstract

Background The health disadvantages faced by Australian Aboriginal peoples are evidenced in early life, although few studies have focused on the reasons for population-level inequalities in more severe adverse outcomes. This study aimed to examine the scale of disparity in severe neonatal morbidity (SNM) and mortality between Aboriginal and non-Aboriginal births and quantify the relative contributions of important maternal and infant factors.

Method A retrospective cohort study with singleton live births (≥32 weeks’ gestation) was conducted using Western Australia linked whole population datasets, from 1999 to 2015. Aboriginal status was determined based on the mothers’ self-reported ethnic origin. An Australian validated indicator was adapted to identify neonates with SNM. The Oaxaca-Blinder method was employed to calculate the contribution of each maternal and infant factor to the disparity in SNM and mortality.

Results Analyses included 425 070 births, with 15 967 (3.8%) SNM and mortality cases. The disparity in SNM and mortality between Aboriginal and non-Aboriginal births was 2.9 percentage points (95% CI 2.6 to 3.2). About 71% of this gap was explained by differences in modelled factors including maternal area of residence (23.8%), gestational age (22.2%), maternal age (7.5%) and antenatal smoking (7.2%).

Conclusions There is a considerable disparity in SNM and mortality between Aboriginal and non-Aboriginal births in Western Australia with the majority of this related to differences in maternal sociodemographic factors, antenatal smoking and gestational age. Public health programmes targeting these factors may contribute to a reduction in early life health differentials and benefit Aboriginal population health through the life course.

  • epidemiology
  • health inequalities
  • infant
  • newborn
  • perinatal epidemiology

Data availability statement

Data may be obtained from a third party and are not publicly available. All data relevant to the study are included in the article or uploaded as online supplementary information.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Severe neonatal morbidity (SNM), a composite term based on diagnoses and procedures such as respiratory distress syndrome, broncho-pulmonary dysplasia and resuscitation, affects a significant proportion of neonates, especially among those born preterm.1 In addition to short-term impacts, SNM is associated with a greater risk of poor development, childhood hospitalisation and mortality.2–4

Several factors are impacted in adverse neonatal outcomes, although a few studies have examined factors specifically associated with SNM.5–7 Misra et al proposed a multiple determinants framework for conceptualising the complex relationships that influence perinatal (including neonatal) health,8 and implicate inter-related distal social environmental factors (eg, socioeconomic status (SES), ethnicity), proximal biomedical (eg, chronic diseases) and behavioural (eg, smoking) responses and intermediate diseases and pregnancy complications in the pathways to poor outcomes.

Persistent marked racial/ethnic variations in preterm birth, low birth weight and perinatal mortality rates have been documented in the US.9 Similarly, in Australia, the rates of these adverse perinatal outcomes have been shown to be substantially higher for Aboriginal and/or Torres Strait Islander (hereafter Aboriginal) than non-Aboriginal populations, although the overall perinatal mortality rate in Australian is comparable to most high-income countries.10 The factors driving these disparities are not yet fully understood and they are likely to be complex and contextual. The limited available US-only literature on ethnic disparities in SNM is exclusively focused on very preterm (born before 32 weeks of gestation)11 or preterm births (born before 37 weeks of gestation),12 13 and features conflicting findings among this small subgroup of births (~10%).14

Accumulated data show that Aboriginal women in Australia are more likely to have greater risks to their pregnancy health than non-Aboriginal women.15 For instance, they are more likely to be younger at childbirth, live in remote and socioeconomically disadvantaged areas, smoke during pregnancy, have prenatal chronic conditions and pregnancy complications. Consequently, they are also at greater risk of having preterm and small for gestational age births.10 To date, there is a paucity of evidence to ascertain the degree to which these risk profile differences contribute to the disparities in SNM and mortality.

The most recent US study11 found that maternal and infant characteristics did not make a significant contribution to the ethnicity gap in SNM, but the analysis was limited to very preterm births. In contrast, other US studies16–19 have found that differences in maternal factors contributed significantly to the ethnic gap in preterm birth and low birth weight, although the relative contribution varies across studies. Notably, there are considerable differences in ethnic composition and healthcare between the US and Australia, which means these findings may not be applicable in Australia. Moreover, numerous factors are likely to be involved in perinatal morbidity/mortality, which impedes targeted interventions.20 Therefore, to better understand the underlying causes of the disparity in SNM and mortality between Aboriginal and non-Aboriginal births in the Australian context, and to reduce the disparity, comprehensive context-based studies are urgently required. This study examined the scale of disparity in SNM and mortality between Aboriginal and non-Aboriginal births and calculated the relative contributions of key maternal and infant factors to the gap.

Methods

Study design and population

A whole population retrospective cohort study with singleton live births of ≥32 weeks’ gestation in Western Australia (WA) from July 1999 to December 2015 was conducted. Neonates with major congenital anomalies (n=19 879), identified in the Western Australian Register of Developmental Anomalies (WARDA), were excluded. The scope of the sample was restricted to ≥32 weeks as all earlier births are, by definition, classified as having SNM by the composite indicator used.1

Datasets and measures

Data were obtained from several datasets including the Midwives’ Notification System (MNS), Hospital Morbidity Data Collection (HMDC), Birth and Death Registers, WARDA and an Aboriginal status flag dataset. The MNS, the primary data source, records the circumstances of all births (≥20 weeks’ gestation) in WA since 1980, while the HMDC provides comprehensive information for all admitted patients in the state from 1970. Both datasets are periodically checked for data quality, accuracy and validity.21 22 In the rare instance where any of the key variables (such as dates of birth) varied between the administrative datasets, the value from the MNS was used. Similarly, the Birth and Death Registers collect information on all births and deaths across WA from the 1970s. The WARDA routinely records developmental anomalies in WA for children from birth to 6 years of age. These datasets were linked by the WA Department of Health’s Data Linkage Branch using a probabilistic linkage method,23 with data securely transferred (with identifiers removed) to the researcher team.

As part of the MNS, women declared their ethnic origin (Caucasian, Aboriginal and/or Torres Strait Islander, Asian, Indian (subcontinent), African, Polynesian, Maori and Other). For this analysis, Aboriginal and/or Torres Strait Islander women were collapsed together while all other ethnic groups were classified as non-Aboriginal.

The SNM outcome variable was adapted from a validated composite outcome indicator which was developed to identify infants with severe adverse outcomes (includes death within 28 days of birth or before first discharge home from hospital) in routinely collected population health datasets in New South Wales.1 The indicator includes conditions and procedures based on the International Statistical Classification of Diseases, Tenth Revision, Australian Modification (ICD-10-AM codes). In consultation with the WA Clinical Coding Authority, we made changes to some components of the indicator. For example, ‘other’ brachial plexus birth trauma (ICD-10-AM code P14.3) was added while intravenous fluids (ICD-10-AM code 96199) was excluded (online supplemental table S1). Data on SNM and mortality were complete for 99.6% of singleton births, but 4.7% (n=21 005) of eligible births were excluded because of missing data on area-based socioeconomic measures (n=18 983), gestational age (n=1960) and other covariates (n=62).

Supplemental material

Other variables obtained as part of the above datasets were categorised as follows. Maternal age at childbirth (<20, 20–34, 35–39 and ≥40 years), parity (0, 1, 2–4, ≥5), antenatal smoking (yes/no), pre-existing condition (yes/no), pregnancy complication (yes/no), onset of labour (spontaneous, induced, no-labour), method of birth (spontaneous vaginal, assisted vaginal, elective or emergency caesarean), infant sex (boy/girl), gestational age at birth (32–36, 37–38, 39–40 and ≥41 weeks), appropriateness for gestational age defined using Australian national birth weight centiles (severe growth restriction (<3rd percentile), small for gestational age (3rd to <10th percentile), appropriate for gestational age [10th–90th percentile), large for gestational age (>90th percentile))24 and year of birth (1999–2003, 2004–2007, 2008–2011, 2012–2015). The MNS and HMDC data were used to define a pre-existing condition (pre-existing diabetes and/or hypertension) and a composite pregnancy complication variable (gestational hypertension, pre-eclampsia, gestational diabetes, placenta praevia, placental abruption or other antepartum haemorrhage).

We used an area-level SES measure as no individual data were available. The area-level SES quintiles (most disadvantaged (first quintile) to least disadvantaged (fifth quintile)) were created based on the Socio-Economic Indexes for Areas; a product created by the Australian Bureau of Statistics.25 Similarly, area of residence was determined by the Australian Bureau of Statistics Remoteness Area classification which divides Australia, based on a measure of relative access to services, into five classes: major cities, inner regional, outer regional, remote and very remote.26 For this analysis, we have combined inner and outer regional areas (‘regional’) and remote and very remote areas (‘remote’) to establish categories with a sufficient sample size.

Statistical analysis

All statistical analyses were performed in STATA V.15 (StataCorp. 2017). χ2 tests were used to compare the percentage of maternal and infant characteristics by Aboriginal status. Log-binomial or Poisson (when the former does not converge) model with cluster-robust standard errors (accounting for multiple births per mother; maximum 10 births per mother) was used to calculate risk ratios (RRs) and 95% CIs for the association of maternal ethnic origin and other factors, and the risk of SNM and mortality. Model factors were selected based on the existing literature and the perinatal health framework developed by Misra et al.8 The available data included distal (area-level SES and area of residence), proximal (maternal age, parity, smoking during pregnancy and pre-existing condition) and intermediate (pregnancy complication, the onset of labour, appropriateness for GA, method of birth and GA) factors as well as covariates (infant sex and year of birth). Gestational age is a well-known predictor of SNM and mortality,1 and Aboriginal women are more likely to have preterm births.10 Our preliminary analysis also showed a significant interaction between gestational age and ethnicity on SNM, particularly at 32–36 weeks of gestation (p<0.001). Hence, we repeated the log-binomial model by gestational age group.

The Blinder-Oaxaca decomposition method was employed to estimate the percentage contribution of each maternal and infant variable to the disparity in SNM and mortality between Aboriginal and non-Aboriginal births. It partitions the ethnic gap in SNM and mortality into ‘explained’ and ‘unexplained’ parts. While the explained component shows the amount of the disparity accounted for by group differences in factors included in the model, the unexplained part represents the remaining gap in the outcome (ie, SNM and mortality) which is not attributable to the distribution variations in those variables.27 28 A variable can contribute positively or negatively to the inequality (ie, a negative contribution indicates that the predictor contributed towards decreasing the disparity).

Finally, to check consistency and robustness of results, we repeated both the log-binomial and decomposition models using an alternative Aboriginal status indicator, derived using an algorithm based on several administrative datasets held by the Department of Health.29 The decomposition analysis was also performed in subsamples restricted to one birth per woman (the most recent birth was included when two or more births were available) and births with spontaneous onset of labour and was replicated in a stratified analysis based on the gestational age group.

Results

We included 425 070 births (from 259 080 women), of which 23 665 (5.6%) births were to Aboriginal women. Overall, 15 967 (3.8%) neonates had SNM, including 184 (122 with SNM) deaths within the first month of life or before first discharge home. Aboriginal neonates had the highest SNM rate, while neonates born to other ethnic minority women (included as non-Aboriginal) had a similar SNM rate to the births to Caucasian women (79% of total births) (online supplemental table 2). Compared with non-Aboriginal women, Aboriginal women were more likely to be younger at birth, be in the highest quintile of socioeconomic disadvantage, live in remote areas, multiparous, have smoked during pregnancy, have a pre-existing condition, have spontaneous onset of labour at lower gestational age groups (32–36 and 37–38 weeks) and have preterm (32–36 weeks’ gestation) and severely growth restricted or small for gestational age births. However, they were less likely to have births at advanced maternal age (≥40 years) and assisted vaginal or elective caesarean births (table 1).

Table 1

Maternal and infant characteristics of live births ≥32 weeks’ gestation by Aboriginal status, Western Australia, July 1999–December 2015

Table 2 shows the association between maternal and infant factors and SNM and mortality. In the unadjusted model, Aboriginal births were 1.81 (95% CI 1.72 to 1.90) times more likely to have SNM or dies in the neonatal period, compared with non-Aboriginal births. The association remained in a model adjusted for a range of sociodemographic, medical and other factors, although the effect size was considerably attenuated. The largest effect size was observed for preterm births (at 32–36 weeks’ gestation; RR=8.44, 95% CI 8.13 to 8.76). Other notable maternal factors associated with a higher risk of SNM and mortality were younger maternal age, remote residence, grand multiparity, pre-existing medical condition and pregnancy complications. Additionally, birth-related and neonatal factors including the onset of labour, method of birth and appropriateness for gestational age were significantly associated with SNM and mortality (table 2).

Table 2

Relative risk of severe neonatal morbidity (SNM) and mortality among live births ≥32 weeks’ gestation, Western Australia, July 1999–December 2015 (N=425 070)

A stratified analysis based on gestational age revealed that the risk of SNM and mortality for Aboriginal births increased with gestational age. At 32–36 weeks of gestation, we observed no difference between Aboriginal and non-Aboriginal births in the risk of SNM and mortality, but the adjusted RR (95% CI) increased from 1.35 (1.20 to 1.53) among neonates born at 37–38 weeks of gestation to 1.60 (1.32 to 1.96) among those born at 41 or more weeks of gestation (table 3).

Table 3

Relative risk of severe neonatal morbidity (SNM) and mortality, by Aboriginal status—stratified by gestational age group, Western Australia, July 1999– December 2015

The results of our supplemental analyses using the alternative Aboriginal status indicator (derived from algorithms) were similar to the main analysis, although there were slightly smaller effect sizes among births to Aboriginal mothers across all gestational age groups (online supplemental tables S3 and S4).

As shown in table 4, the disparity (in absolute scale) in SNM and mortality between Aboriginal and non-Aboriginal births was 2.9 percentage points (95% CI: 2.6 to 3.2), and remained relatively stable over the study period (online supplemental table S5). About 71% of the gap was attributable to factors included in the model and most of these variables contributed positively to the disparity. Overall, differences in sociodemographic factors (including age, area-level SES, area of residence and parity) and antenatal smoking accounted for the greatest proportion of the explained gap (53.3%), with area of residence being the biggest contributor (23.8%). The differences in birth-related and neonatal factors also made a considerable contribution to the SNM and mortality gap. While the onset of labour and method of birth contributed towards reducing the disparity, differences in gestational age and appropriateness for gestational age between Aboriginal and non-Aboriginal births were important drivers of the inequality. In other words, the onset of labour and method of birth were negative contributors to the disparity in which Aboriginal women had more favourable characteristics (a greater proportion of spontaneous onset of labour and spontaneous vaginal births). Without these differences in rates of spontaneous onset of labour and spontaneous vaginal births between Aboriginal women and their non-Aboriginal counterparts, the gap would have increased. Maternal pre-existing medical conditions and pregnancy complications did not contribute appreciably to the gap in SNM and mortality.

Table 4

Factors contributing to the gap in severe neonatal morbidity (SNM) and mortality between Aboriginal and non-Aboriginal live births ≥32 weeks’ gestation, Western Australia, July 1999–December 2015

While we found similar results in supplemental analyses using the alternative Aboriginal status indicator (online supplemental table S6), although a higher proportion of the overall disparity was explained by modelled variables in the analysis restricted to one birth per woman (online supplemental table S7) and a lower proportion was evident in the analysis limited to births with spontaneous onset of labour—largely because of the changes in the contributions of method of birth (ie, negative contributor) (online supplemental table S8). In the stratified analysis, based on gestational age, we observed smaller disparities in SNM and mortality between Aboriginal and non-Aboriginal births at 32–36 and 39–40 weeks of gestation. Moreover, the total inequality explained by included factors varied from 20.1% (≥41 weeks) to 60.4% (37–38 weeks), which were lower than that of the main analysis results (online supplemental table S9).

Discussion

In this population-based cohort study, the rate of SNM (including mortality) was significantly higher in newborns of Aboriginal women than those of non-Aboriginal women, with the major proportion (70.8%) of this disparity explained by maternal sociodemographic factors (including maternal area of residence) along with gestational age. Our results indicate that a reduction in the rate of neonatal morbidity/mortality among the Aboriginal population is possible with effective interventions and programmes.

Our finding that newborns of Aboriginal women are more likely to have SNM than the non-Aboriginal population accords with the existing evidence of widespread inequalities in perinatal morbidity/mortality in Australia—this includes disproportionately greater rates of low birth weight, preterm birth and perinatal mortality among Indigenous populations.30–32 Our evaluation of the relative contributions of salient maternal and infant factors to disparities in neonatal morbidity and mortality is a novel contribution to the literature. We highlight that the gap in SNM and mortality is largely attributable to differences in maternal sociodemographic factors (eg, area of residence, area-level SES and maternal age), antenatal smoking and gestational age. Although maternal pre-existing medical conditions and pregnancy complications were important predictors of SNM and mortality, they did not contribute meaningfully to the disparity in these outcomes. This underscores the common observation that inequalities in perinatal outcomes are primarily driven by social and behavioural health determinants.9

The available limited literature on the factors contributing to ethnic disparities in perinatal outcomes has exclusively focused on preterm birth and/or low birth weight outcomes.16 18 19 To our knowledge, only one recent US study,11 limited to very preterm infants, has evaluated this issue and found that infant factors (birth weight, gestational age, sex and multiple births) accounted for about 49% of the disparity in neonatal morbidity and mortality between non-Hispanic black and non-Hispanic white populations as well as between Hispanic and non-Hispanic white populations. However, in contrast to our study, they demonstrated no significant contributions by maternal sociodemographic (age, parity, educational level and delivery type) and lifestyle (body mass index and antenatal smoking) factors. This may be because of differences in the socioethnic composition of the population (and their characteristics) and healthcare systems between Australia and the US, which reinforces the notion that ethnic-based analyses and interventions should be population and context specific.33

Because of the long-term sequelae and economic consequences,2 3 34 the findings of the current study underscore the urgent need to close the gap in perinatal morbidity/mortality between Aboriginal and other population groups in WA. Our findings also suggest that comprehensive population-based interventions that aim to: improve maternal area-level SES; access to services and reduce teenage pregnancies, as well as preterm births, may narrow the ethnic disparities in severe neonatal outcomes. Remoteness is the single most important contributor to the disparity and may reflect the risks associated with a comparatively lower level of access (in terms of physical proximity and financial wherewithal) to culturally appropriate and high-quality antenatal and obstetric services in these settings.35 These issues are acknowledged in the most recent pregnancy care guidelines,36 and therefore, improvement in these services and implementation of Aboriginal-led programmes may meaningfully reduce the gap.37 This, in turn, may help to further reduce the overall SNM and mortality in the State. We also found that difference in the distribution of gestational age is the major infant factor responsible for SNM disparity. Interlinked factors such as maternal socioeconomic disadvantage, antenatal smoking and urogenital infections are more prevalent in the Aboriginal population and are strongly linked to preterm birth.36 38 Therefore, to further reduce the gap in SNM, effective spontaneous preterm birth prevention strategies and interventions that improve preterm birth outcomes should be implemented.39–41

To our knowledge, this is the first study to estimate individual variable contributions to the disparity in SNM and mortality between Australian Indigenous and non-Indigenous populations. We were able to include important variables that explained over two-thirds of the disparity. The use of a large total population cohort with linkages to detailed information on prenatal and early life circumstances is a considerable strength. Nevertheless, this study has some limitations. We used administrative data that are susceptible to errors and usually lack some salient covariates. For instance, data were not available or only available recently on antenatal care, obesity, alcohol consumption and mental health problems and thus some part of the disparity remained unexplained. Future studies may identify additional important factors for targeted interventions. Additionally, Aboriginal women could have been misclassified,42 although we found similar results to the main analysis when we employed an Aboriginal indicator variable derived using the algorithms developed by the Getting our Story Right project—suggesting a minimal impact of any identification errors. The other limitation of this study is that we have excluded extremely preterm births (20–31 weeks’ gestation) since the indicator we used classed all as being severely morbid. As these made up only 2.2% and 0.7% of total Aboriginal and non-Aboriginal live singleton births respectively during the time period of interest, they would have had a limited effect on the generalisability of our findings.

In conclusion, the findings of this large population-based study demonstrate that Aboriginal births are at higher risks of SNM and mortality, compared with non-Aboriginal births. The majority of this ethnic disparity was explained by maternal sociodemographic and antenatal smoking, along with gestational age. Reducing the gap will require a sustained focus on well resourced, Aboriginal-led public health interventions and programmes for the benefit of infant health.

What is already known on this subject

  • Aboriginal women are more likely to have poorer perinatal outcomes than non-Aboriginal women. This issue requires urgent and effective intervention. However, the evidence is scarce regarding which factors contribute to the gap in severe adverse health events in the neonatal period between Aboriginal and non-Aboriginal populations.

What this study adds

  • We demonstrate that the rate of severe neonatal morbidity and mortality was substantially higher for newborns of Aboriginal than non-Aboriginal women. Over two-thirds of this disparity was explained by maternal sociodemographic factors, such as maternal area of residence, along with gestational age. Public health interventions addressing these factors are likely to reduce this gap.

Data availability statement

Data may be obtained from a third party and are not publicly available. All data relevant to the study are included in the article or uploaded as online supplementary information.

Ethics statements

Patient consent for publication

Ethics approval

Ethics approvals were obtained from the Western Australian Department of Health Human Ethics Research Committee (2016/51) and the Western Australian Aboriginal Health Ethics Committee (797).

Acknowledgments

We would like to acknowledge the Data Linkage Branch (Western Australian Government Department of Health), the Midwives’ Notification System, Birth and Death Registrations, the Hospital Morbidity Data Collection and the Western Australian Register of Developmental Anomalies for providing data for this project. In particular, we acknowledge D. Rosi Katich, Senior Coding Consultant, WA Clinical Coding Authority (WA Department of Health) who advised us on the ICD-10-AM classification in relation to the severe neonatal morbidity index.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors AAA, CCJS and HDB conceived and designed the study. BMF, RM and SWW participated in the design of the study. AAA conducted all statistical analysis and drafted the manuscript. All authors critically reviewed and approved the final manuscript.

  • Funding This research was supported by funding from an Australian National Health and Medical Research Council (NHMRC) Project Grant (GNT1127265) which funded AAA, HDB and CCJS. BMF is funded by the Australian National Health and Medical Research Council (Project Grant 1098844).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.