The Western Australian preterm birth prevention initiative: a whole of state singleton pregnancy cohort study showing the need to embrace alternative models of care for Aboriginal women

Background Preterm birth (PTB) is the greatest cause of mortality and morbidity in children up to five years of age globally. The Western Australian (WA) PTB Prevention Initiative, the world’s first whole-of-population whole-of-state program aimed at PTB prevention, was implemented across WA in 2014. Methods We conducted a prospective population-based cohort study using pregnancy data for singleton births in WA from 2009 to 2019. Logistic regression using the last full year before the Initiative (2013) as the reference, and run charts were used to examine changes in PTB rates compared to pre-Initiative levels, by gestational age group, hospital type, low and high risk of PTB in mid-pregnancy, and onset of labour (spontaneous/medically initiated). Analyses were stratified by Aboriginal and non-Aboriginal maternal ethnicity. Results Amongst non-Aboriginal women, there was initially a reduction in the PTB rate across the state, and in recent years it returned to pre-Initiative levels. Amongst Aboriginal women there was a small, non- significant reduction in the state-wide PTB rate in the first three years of the Initiative, followed by a rise in recent years. For non-Aboriginal women, the reduction in the rate of PTB at the tertiary centre was sustained and improved further for women of all risk levels and onsets of labour. This reduction was not observed for Aboriginal women giving birth at the tertiary centre, amongst whom there was an increase in the PTB rate overall and in all subgroups, with the exception of medically initiated PTB. Amongst Aboriginal women the PTB rate has also increased across the state. At non-tertiary hospitals there was a large increase in PTB amongst both Aboriginal and non-Aboriginal women, largely driven by medically initiated late PTB. Maternal risk factors cannot account for this increase. Conclusions The reduction in PTB rates amongst non-Aboriginal women at the state’s tertiary hospital demonstrates that with the right strategies, PTB can be reduced. A sustained collaborative model is required to realise this success in non-tertiary hospitals. The series of interventions was of limited use in Aboriginal women, and future efforts will need to be directed at strategies more likely to be successful, such as midwifery continuity of care models, with Aboriginal representation in the healthcare workforce. Supplementary Information The online version contains supplementary material available at 10.1186/s12884-022-05222-9.


Background
Preterm birth (PTB), defined as birth from 20 and before 37 completed weeks of pregnancy is the greatest cause of mortality and morbidity in children up to five years of age globally [1] and is increasing in high income countries [2], including Australia [3]. One in eleven Australians is born too early and among Aboriginal and/or Torres Strait Islander women (hereafter respectfully referred to as Aboriginal) the rate is almost double [3]. Perinatal outcomes are considerably worse for Australian Aboriginal infants, who experience higher rates of PTB, low birthweight and infant mortality both nationally [3], and in Western Australia (WA) [4][5][6]. The causes of these inequalities are complex and risk factors for poor perinatal outcomes are not necessarily the same in Aboriginal women as in the broader WA obstetric population.
The WA PTB Prevention Initiative (the Initiative) was implemented across WA in 2014 with the objective of safely reducing PTB. The WA Initiative is uniquely the world's first whole-of-population whole-of-state program aimed at PTB prevention, covering tertiary (one established and one evolving centre that opened during the implementation period), primary and secondary hospitals. In the first year of the Initiative (2015), PTB was reduced by 7.6% across the state and by 20% at the state's established tertiary centre [7]. A recent study revealed that the PTB reduction up to 2017 was sustained at the state's established tertiary centre, but not elsewhere in WA [8].
The purpose of this study was to evaluate the effects of the Initiative over the first 5 years (2015-2019), and for the first time, specifically assess outcomes amongst Aboriginal women in WA. It is essential that the Initiative is evaluated amongst Aboriginal women, to gain a better understanding of the major drivers of PTB in this population and thus enable the design of more appropriate and culturally responsive programs.

Methods
Based on existing evidence, the WA Initiative incorporated new clinical guidelines for singleton pregnancies. This included recommending: measurement of cervix length at all mid-pregnancy morphology scans and administration of vaginal progesterone and/or cervical cerclage for women with a shortened cervix, administration of vaginal progesterone for women with a history of spontaneous PTB, that no pregnancy should be ended prior to 38+ weeks without medical or obstetric justification, identification and provision of smoking cessation counselling to women who smoke, and a new PTB Prevention Clinic at the tertiary centre for high risk cases [7].
We conducted a prospective population-based cohort study using pregnancy data on all Western Australian births from 20 weeks gestation onwards between 2009 and 2019 obtained from the Midwives' Notification System (MNS) in which pregnancy and birth information is recorded by the attending midwife. Information included data on maternal characteristics (maternal age, health region of residence, socioeconomic status, smoking during pregnancy and ethnicity), medical history (preexisting diabetes and hypertension, asthma and other pre-existing conditions), obstetric history (parity, previous stillbirth and IVF conception, caesarean at last birth, number of antenatal visits), complications during pregnancy (placental abruption, antepartum haemorrhage, gestational diabetes, threatened miscarriage, threatened preterm labour) and labour and birth (delivery hospital, onset of labour, mode of delivery, gestational age at birth and admission to special care nursery). Socioeconomic status was calculated using the Index of Relative Socio-Economic Advantage and Disadvantage score derived by the Australian Bureau of Statistics [9], and was categorised into those in the lowest 40%, and those in the upper 60%. History of preterm birth was derived for each mother and delivery hospital was used to derive hospital type (established tertiary, evolving tertiary, non-tertiary). Aboriginal status was reported for each individual pregnancy, and mothers who reported being Aboriginal for the majority of their pregnancies were deemed to be Aboriginal for all of them. Given the WA Initiative targeted singleton pregnancies, only singleton pregnancies were included in the analysis. Inductions at the established tertiary hospital before 25 weeks gestation that resulted in intrapartum death were excluded on the assumption that they were terminations of pregnancy.
The PTB rate was examined overall and by gestational age (20-31 weeks and 32-36 weeks). Gestational ages 20-27 and 28-31 were combined due to small number of annual births amongst Aboriginal women. Analyses were performed for the state, for the established tertiary centre, and for non-tertiary hospitals (secondary and primary hospitals combined). Births from the evolving tertiary centre were only included in the state-wide PTB rates as the PTB rate at this centre was substantially different from the established tertiary centre (hereafter referred to as the tertiary centre). Subgroup analyses were also performed by onset of labour (spontaneous/medically initiated) and risk of PTB at the first antenatal visit (low/high).
Binary logistic regression was used to model the likelihood of PTB and stillbirth in each year from 2009 to 2018, relative to these outcomes in 2019, with alternative comparisons using year 2013, the last full year before the Initiative, as the reference. Analyses were performed on pregnancies in years 2009-2019 but only rates from 2013 to 2019 are presented. Supplementary evaluation of PTB incidence was conducted after classifying women into low or high risk of PTB based on information available at the first antenatal visit, using separate logistic regression models for nulliparous and parous women. This risk classification was based on all pregnancies and ethnicities combined, from 2009 to 2019 without year of birth as a predictor, and classification was therefore independent of year of birth. The variables used for classification of low and high risk are shown in Supplementary  Table 1.
Patterns of bi-monthly PTB incidence rates were investigated using run charts, which are used to assess temporal changes. Probability rules are used to detect shifts, defined as six consecutive points above or below the baseline rate, and this method is useful where small sample sizes create deficits in power [10], such as among Aboriginal women. For these analyses, the baseline PTB rate was calculated as the median bimonthly rate from January 2013 to June 2014, which was the 18 months immediately prior to the introduction of the WA Initiative. Run charts by Aboriginal status and hospital type were created by onset of labour and by risk category, for each gestational age group (20-31 weeks, 32-36 weeks, < 37 weeks). There were insufficient numbers of births among Aboriginal women to produce run charts by risk or onset of labour for the 20-31 week gestational age group. This was also the case for low risk Aboriginal women giving birth at the tertiary centre in all gestational age groups. The proportion of high risk women, as well as changes in the rates of risk factors over time were examined by ethnicity and hospital.
SAS statistical software (version 9.4, Cary, NC: SAS Institute Inc.) was used for data analysis with p-val-ues< 0.05 considered statistically significant.
The study was approved by the Women and Newborn Health Service Human Research Ethics Committee (RGS0000002677), the Health Department of Western Australia (RGS0000000704) and the Western Australian Aboriginal Health Ethics Committee (965). A summary of maternal characteristics of Aboriginal and non-Aboriginal women are presented in Table 2. Aboriginal women were younger, had more children, were more likely to smoke during pregnancy, reside in in rural areas, and fall in the two lowest socio-economic quintiles. Aboriginal women were also more likely to have pre-existing diabetes, hypertension and other pre-existing conditions, and to have a history of prior PTB or stillbirth. They were less likely to have conceived using IVF or to have had a caesarean at their last delivery. Aboriginal women were more likely to have spontaneous labour than non-Aboriginal women, and their babies were more likely to be admitted to the special care nursery.

Risk factors for PTB, by hospital and year, for Aboriginal and non-Aboriginal mothers
Across the study period, PTB rates were higher amongst Aboriginal women with an overall PTB rate of 14.1% compared to 6.8% amongst non-Aboriginal women. Pregnancies were classified as being at low or high risk of PTB based on information available at the first antenatal visit, and the clinical profile described above resulted in 82.9% of Aboriginal women being classified as high risk, compared to 16.3% of non-Aboriginal women. The PTB rate was 15.3% amongst high risk Aboriginal women compared to 13.3% amongst high risk non-Aboriginal women. Amongst low risk Aboriginal women, the PTB rate was 8.4% compared to 5.5% amongst low risk non-Aboriginal women. The PTB rates for the variables used for classification, by risk level are presented in Supplementary Table 1. Women classified as high risk of PTB also experienced higher rates of pregnancy complications including stillbirth, threatened preterm labour, gestational diabetes, pre-eclampsia, antepartum haemorrhage and preterm pre-labour rupture of membranes with delivery < 37 weeks, irrespective of ethnicity (Supplementary Table 2).  Table 3 shows the annual rates of PTB among Aboriginal and non-Aboriginal women in the tertiary centre, nontertiary hospitals combined and state-wide from 2013 to 2019.

PTB rates in Aboriginal and non-Aboriginal mothers
For non-Aboriginal women at the state's tertiary centre, the rate of PTB was significantly reduced from 20.0% in 2013 to its lowest rate of 14.8% in 2019 (p < 0.001). The significant reduction in the PTB rate occurred primarily in births between 32 and 36 weeks which decreased from 14.8% in 2013 to a minimum of 10.3% in 2019 (p < 0.001). Run charts of PTB incidence amongst non-Aboriginal women by gestational age group demonstrated a significant reduction in PTB for all years of the Initiative (Fig. 1). No similar improvements occurred amongst Aboriginal women for whom the PTB rate increased from 31.0% in 2013 to 35.3% in 2019 (p = 0.393) and run chart analysis showed an increase in PTB between 20 and 31 weeks during 2018 and 2019 (Fig. 2).
For non-Aboriginal women in non-tertiary hospitals, the PTB rate increased from 4.5% in 2013 to 5.  (Fig. 1). Amongst Aboriginal women in non-tertiary hospitals, the PTB rate was 9.2% in 2013 and in subsequent years ranged from a minimum of 8.0% in 2015 (p = 0.222) to a maximum of 9.4% in 2019 (p = 0.867). Run chart analysis showed an increase in overall PTB during 2018 and at the start of 2019 (Fig. 2).
For non-Aboriginal women across the state the PTB rate was 7.0% in 2013, and subsequently ranged from 6.5% in 2015 (p = 0.017) to 7.2% in 2017 (p = 0.372). In 2018 and 2019 the PTB rates were similar to pre-Initiative levels at 7.1% (p = 0.370) and 6.9% (p = 0.404) respectively. Run charts showed a reduction in PTB in all gestational age groups for the two years after the introduction of the Initiative. Subsequently, increases in PTB incidence were observed between 2016 and 2018 for the 32-36 week gestational age group, and overall ( Fig. 1). Amongst Aboriginal women across the state, the PTB rate was 14.2% in 2013, with the lowest rate occurring in 2015, the first full year of the Initiative (13.3%, p = 0.236) and the highest in 2019, the most recent year of data (15.4%, p = 0.337). Run charts showed increases in PTB incidence in all gestational age groups during 2018 and 2019, and also during 2016 and 2017 in births between 32 and 36 weeks (Fig. 2). Tables 4 and 5 show the annual rates of PTB among Aboriginal and non-Aboriginal women in the tertiary centre, non-tertiary hospitals combined and state-wide from 2013 to 2019, by low and high risk of PTB.

PTB rates in Aboriginal and non-Aboriginal mothers by risk
For non-Aboriginal women at the state's tertiary centre, who were classified at low risk, the PTB rate declined from 16.7% in 2013 in every subsequent year, with the reduction reaching statistical significance in years 2015-2019. The lowest PTB rate occurred in 2019 (11.7%, p < 0.001) and the reduction over time was primarily due to a reduction in PTB between 32 and 36 weeks. Run charts of PTB incidence, by gestational age group demonstrated a significant reduction in PTB overall and from 32 to 36 weeks for all years of the Initiative, and from 2016 to 2019 for PTB between 20 and 31 weeks (Supplementary Fig. 1). For high risk non-Aboriginal women who gave birth at the state's tertiary centre, the PTB rate decreased from 29.7% in 2013, in every subsequent year, with the reduction reaching statistical significance from 2015 to 2019. The PTB rate of 23.9% was at its minimum in 2019 (p < 0.001), although this was a smaller percentage decrease (19.5%) than that in low risk births (30.1%). Run charts showed a reduction in PTB incidence overall and from 32 to 36 weeks for all years of the initiative (Supplementary Fig. 1).
For non-Aboriginal women at non-tertiary hospitals who were classified at low risk, the PTB rate increased from 3.9% in 2013, to 4.2% in 2017 (p = 0.011) and 2018 (p = 0.013). This was due to an increase in PTB from 32 to 36 weeks from 3.7% in 2013 to 3.9% in 2017 (p = 0.027) and 2018 (p = 0.029). Run charts by gestational age group demonstrated a significant increase in PTB incidence from 20 to 31 weeks in 2018 and 2019 ( Supplementary  Fig. 1). For non-Aboriginal women at non-tertiary hospitals who were at high risk, the PTB rate was higher than   Fig. 1).
For Aboriginal women at the tertiary centre who were classified at low risk, evaluation of the PTB rate is limited by small numbers ranging from just 23 births in 2013 to 56 births in 2019. This is reflected in the unstable PTB rate that was 8.7% in 2013 rising to a maximum of 28.6% (p = 0.116) in 2016 (Table 4). These low numbers also prevented meaningful run chart analysis. For Aboriginal women at the tertiary centre who were classified at high risk, the PTB rate was 32.3% in 2013 and then ranged between 30.8% in 2014 (p = 0.585) and 36.8% in 2019 (p = 0.611). Run charts by gestational age group amongst high risk Aboriginal women at the tertiary centre demonstrated a reduction in the overall PTB rate from 2014 to 2015 ( Supplementary Fig. 2).
For low risk Aboriginal women at non-tertiary hospitals, the PTB rate was 7.3% in 2013, then ranged between a minimum of 4.0% in 2016 (p = 0.222) and a maximum of 7.2% in 2015 (p = 0.792) and 2019 (p = 0.516). For Aboriginal women at non-tertiary hospitals who were classified at high risk, following a rate of 9.6% in 2013, the PTB rate was at its lowest at 8.1% in 2015 (p = 0.154) and highest at 10.4% in 2018 (p = 0.884). Run charts by gestational age group amongst high risk Aboriginal women at non-tertiary hospitals demonstrated an increase in PTB overall and among births between 32 and 36 weeks in years 2018 and 2019 ( Supplementary Fig. 2). *Only term births from 2013 are included + Column percentages don't add up to 100 due to missing data  Tables 6 and 7 show the annual rates of PTB among Aboriginal and non-Aboriginal women, by onset of labour, in the tertiary centre, non-tertiary hospitals combined and state-wide from 2013 to 2019. For non-Aboriginal women at the tertiary centre, there was a significant decrease in the rate of medically initiated PTB in every year following 2013 (12.4%), reaching its lowest rate in 2015 (8.5%, p < 0.001), and the second lowest in 2019 (9.6%, p < 0.001). The decrease from 9.1% in 2013 was also statistically significant in the 32-36 week gestation group for every year of the Initiative, with the lowest rate occurring in 2015 (6.0%, p < 0.001) followed by 2019 (6.9%, p < 0.001). Run charts by gestational age group demonstrated a reduction in PTB overall and in both gestational age groups, for all years of the initiative ( Supplementary Fig. 3). For non-Aboriginal women at the tertiary centre, there was also a significant decrease in the spontaneous PTB rate compared to 2013 (7.6%) in 2016 (6.6%, p = 0.036), 2017 (5.8%, p < 0.001) and 2019 (5.2%, p < 0.001). Run charts by gestational age group demonstrated a significant reduction in PTB overall and between 32 and 36 weeks from 2015 to 2019, and in PTB between 20 and 31 weeks from 2017 to 2018 (Supplementary Fig. 3).

PTB rates in Aboriginal and non-Aboriginal women by onset of labour
For non-Aboriginal women at non-tertiary hospitals, there was a significant increase in the rate of medically initiated PTB in every year following the introduction of the Initiative, from 2.0% in 2013 to a peak of 2.8% in 2018 (p < 0.001). Run charts by gestational age group demonstrated a significant increase in the incidence of PTB among births between 20 and 31 weeks in year 2018 and 2019 and overall and between 32 and 36 weeks in years Table 3 Rates of preterm birth in singleton pregnancies stratified by gestational age, ethnicity and hospital, 2013-2019 -PTB rates are compared using logistic regression analysis with year 2019 as a reference. Statistical adjustments were made for all maternal characteristics known at the time of the first antenatal visit used to derive low risk of PTB including maternal age, pre-existing diabetes, pre-existing hypertension, other pre-existing maternal medical conditions, asthma, smoking during pregnancy, ethnicity (Aboriginal, Caucasian, other), socioeconomic status in the lowest two quintiles, IVF conception, history of stillbirth, history of PTB and caesarean at last delivery. Rates that were significantly different from 2019 after adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates    Fig. 4).
For Aboriginal women at non-tertiary hospitals, the rate of medically initiated PTB was 2.0% in 2013 and in subsequent years ranged from a minimum of 1.9% in 2015 (p = 0.742) to its highest level of 3.6% in 2019 (p = 0.020), the most recent year of our evaluation. Run charts by gestational age group demonstrated increases in the incidence of medically initiated PTB overall and in births between 32 and 36 weeks from 2016 ( Supplementary Fig. 4). For Aboriginal women at nontertiary hospitals, the rate of spontaneous PTB was 7.2% in 2013, decreasing to a lowest rate of 5.5% in 2017 (p = 0.062)  Fig. 4). Table 8 shows the annual rates of stillbirth among Aboriginal and non-Aboriginal women, in the tertiary centre, non-tertiary hospitals combined and state-wide from 2013 to 2019, while Supplementary Table 3 shows the adjusted and unadjusted odds of stillbirth compared to 2019.

Rates of stillbirth
Across the study period stillbirth rates were higher amongst Aboriginal women (11.4 per 1000) compared to non-Aboriginal women (4.7 per 1000). Annual variations were much greater amongst Aboriginal women than amongst non-Aboriginal women due to these women having a lower number of births.
There was no significant change in the stillbirth rate amongst non-Aboriginal women who gave birth at the tertiary centre, following the introduction of the Initiative. The stillbirth rate amongst non-Aboriginal women who gave birth at non-tertiary centres was significantly higher in 2017 compared to 2013 (1.96 per 1000 in 2013 vs 2.92 per 1000 in 2017, p = 0.039). The change in the stillbirth rate cannot be explained by any increase in maternal characteristics or obstetric risk factors. There was no significant change in the stillbirth rate amongst non-Aboriginal women who gave birth across the state.
Amongst Aboriginal women who gave birth at the tertiary centre, the stillbirth rate was significantly higher in Amongst these women, the maternal risk factor that was higher in 2018 than previous years was caesarean at last birth (18.5% in 2013 vs 25.9% in 2018). Obstetric complications that were slightly higher in 2018 than previous years were threatened preterm labour (10.5% in 2013 vs 12.5% in 2018) and gestational diabetes (11.2% in 2013 vs 14.5% in 2018), although this variation may be able to be explained by small numbers. Examination of the gestational age at which the stillbirth occurred revealed that almost all of these stillbirths occurred at early gestations. Amongst Aboriginal women who gave birth at non-tertiary centres or across the state overall there was no significant change in the stillbirth rate following the introduction of the Initiative.

Discussion
The current study is the first to examine the effect of the Initiative on PTB amongst Aboriginal women specifically, for whom the PTB rate is higher than the population average. Amongst Aboriginal women there was a small, non-significant reduction in the state-wide PTB rate in the first three years of the Initiative. However, the rate of PTB amongst Aboriginal women has been above pre-initiative (2013) levels since 2017, and reached its highest level in the most recent year of available data (2019). A similar reduction in PTB was observed amongst Aboriginal women giving birth at the tertiary centre in the first 4 years of the Initiative, but this reduction was also not statistically significant, and not maintained to the end of the study period. Amongst non-Aboriginal women, the significant reduction in Table 4 Rates of preterm birth in low risk singleton pregnancies, by gestational age, ethnicity and hospital level -PTB rates are compared using logistic regression analysis with year 2019 as a reference. Statistical adjustments were made for all maternal characteristics known at the time of the first antenatal visit used to derive low risk of PTB including maternal age, pre-existing diabetes, pre-existing hypertension, other pre-existing maternal medical conditions, asthma, smoking during pregnancy, ethnicity (Aboriginal, Caucasian, other), socioeconomic status in the lowest two quintiles, IVF conception, history of stillbirth, history of PTB and caesarean at last delivery. Rates that were significantly different from 2019 after adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates. PTB cases were supressed when < 5 cases occurred and the upper bounds for these rates were calculated based on 5 cases PTB following the introduction of the WA Initiative has been maintained and improved upon at the tertiary centre, but in non-tertiary hospitals the rate of PTB has increased in recent years.

Changes in the PTB rate amongst non-Aboriginal women
Amongst non-Aboriginal women, the reduction in PTB at the tertiary centre occurred overall and in the 32-36 weeks gestational age group, with run charts showing a decrease in all years, in all gestational age groups. The initial reduction has not only been maintained, but has been improved upon, with PTB rates in all gestational age groups reaching their lowest incidence since the introduction of the Initiative, in 2019. The reduction among births between 20 and 31 weeks was moderate and is likely due to the introduction of routine cervix length screening at mid-pregnancy morphology scans, followed by vaginal progesterone for women with a shortened cervix or a history of PTB. The major reduction occurred in births between 32 and 36 weeks which may have additionally benefited from a decline in unnecessary iatrogenic delivery, which has resulted from a persistent cultural change amongst clinicians working within the hospital. This hypothesis is consistent with the result that the reductions occurred in women at both low and high risk of PTB, but these reductions were greater in women at low risk amongst whom early iatrogenic delivery is less likely to be required. It is further supported by the fact that births following both spontaneous and medically initiated onset or labour were reduced in this gestational age group, indicating that all of the aforementioned measures likely had an impact. In contrast to the results observed at the tertiary centre, the PTB rate has risen sharply amongst non-Aboriginal Table 5 Rates of preterm birth in high risk singleton pregnancies, by gestational age, ethnicity and hospital level -PTB rates are compared using logistic regression analysis with year 2019 as a reference. Statistical adjustments were made for all maternal characteristics known at the time of the first antenatal visit used to derive low risk of PTB including maternal age, pre-existing diabetes, pre-existing hypertension, other pre-existing maternal medical conditions, asthma, smoking during pregnancy, ethnicity (Aboriginal, Caucasian, other), socioeconomic status in the lowest two quintiles, IVF conception, history of stillbirth, history of PTB and caesarean at last delivery. Rates that were significantly different from 2019 after adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates women who gave birth at non-tertiary hospitals in WA, particularly between 32 and 36 weeks. This increase has been driven entirely by iatrogenic PTB, and the increase has been greater in women at high risk of PTB. The Australian Preterm Birth Prevention Alliance observed a similar increase in planned birth, as well as a trend toward these births occurring earlier in pregnancy, with elective caesarean sections performed between 34 and 36 week's being associated with severe newborn morbidity [11]. Furthermore, New South Wales data from years 2001-2009 showed an increase in planned birth before the due date and hypothesised that the rise in planned births may be due to clinicians taking an increasingly conservative approach for women with pre-labour rupture of membranes, diabetes, low-lying placenta and advanced maternal age [12]. They also concluded that the increase could be driven by clinicians perceiving less risk associated with early delivery due to advancements in neonatal care [12]. Early delivery is indicated where the risks of continuing the pregnancy outweigh the risks associated with prematurity [13], and examination of rates of maternal conditions amongst non-Aboriginal women giving birth at non-tertiary hospitals showed an increase across the study period in advanced maternal age, other pre-existing medical conditions, gestational diabetes and preterm pre-labour rupture of membranes followed by delivery < 37 weeks. However, there were also decreases in rates of smoking and hypertension and these changes in maternal conditions cannot explain the increase in iatrogenic deliveries before 37 weeks gestation.

Changes in the PTB rate amongst Aboriginal women
Overall, no reduction in PTB rates was observed amongst Aboriginal women giving birth at the tertiary hospital, Table 6 Rates of medically initiated preterm birth in singleton pregnancies, by gestational age, ethnicity and hospital level -PTB rates are compared using logistic regression analysis with year 2019 as a reference. Statistical adjustments were made for all maternal characteristics known at the time of the first antenatal visit used to derive low risk of PTB including maternal age, pre-existing diabetes, pre-existing hypertension, other pre-existing maternal medical conditions, asthma, smoking during pregnancy, ethnicity (Aboriginal, Caucasian, other), socioeconomic status in the lowest two quintiles, IVF conception, history of stillbirth, history of PTB and caesarean at last delivery. Rates that were significantly different from 2019 after adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates for whom the PTB rate followed a slowly increasing trend with a sharp rise in 2019. The rise occurred in both the 20-31 and 32-36 week gestational age groups. This increase in the PTB rate that occurred between 2018 and 2019 may be partially due to the observed increase in the rates of pre-eclampsia (6.0% in 2018, 7.4% in 2019) and preterm pre-labour rupture of membranes followed by delivery < 37 weeks (8.0% in 2018, 10.9% in 2019) between 2018 and 2019. As the major tertiary maternity hospital in the state, it is possible that a large proportion of the preterm births occurring at this hospital were to women who were transferred just prior to the birth of their baby, after the possibility of prevention of the PTB had passed. Compared to non-Aboriginal women, a larger proportion of Aboriginal women live remotely and therefore a disproportionately large number of them may have transferred to the tertiary hospital just prior to birth. Unfortunately, data on where women were receiving their antenatal care prior to the birth of their baby are unavailable so this explanation for the lack of improvement in the PTB rate among Aboriginal women who gave birth at the tertiary hospital cannot be explored. Furthermore, a high proportion of Aboriginal women birthing at the tertiary hospital have risk factors for PTB such as smoking, diabetes and a history of negative obstetric outcomes, with 91.4% of the Aboriginal women who gave birth at the tertiary centre being deemed at high risk of PTB, compared to only 25.1% of non-Aboriginal women. While it may be interpreted as a positive sign that in association with reduced smoking rates, the proportion of high risk women decreased across the study period, unfortunately, this did Table 7 Rates of spontaneous preterm birth in singleton pregnancies, by gestational age, ethnicity and hospital level -PTB rates are compared using logistic regression analysis with year 2019 as a reference. Statistical adjustments were made for all maternal characteristics known at the time of the first antenatal visit used to derive low risk of PTB including maternal age, pre-existing diabetes, pre-existing hypertension, other pre-existing maternal medical conditions, asthma, smoking during pregnancy, ethnicity (Aboriginal, Caucasian, other), socioeconomic status in the lowest two quintiles, IVF conception, history of stillbirth, history of PTB and caesarean at last delivery. Rates that were significantly different from 2019 after adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates not translate to a reduction in risk factors for PTB occurring later in pregnancy. It seems more likely that, as was demonstrated with reporting of birth defects in Aboriginal women [14], ascertainment of risk factors is poor amongst Aboriginal women and therefore risk classification lacks the precision to be of use. Interestingly, the increase in PTB amongst Aboriginal women at the tertiary hospital occurred in preterm births of spontaneous onset indicating that measures such as cervix length measurement, prescription of progesterone and prevention of infection may have been ineffective in this group. This could also be due to a larger proportion of Aboriginal women living outside of the metropolitan region and therefore having less access to mid-pregnancy scans and antenatal care to identify shortened cervix and potential infections. Furthermore, the cost of progesterone as a preventative measure may have been prohibitive for women living in areas assigned lower socioeconomic status, where Aboriginal women are over-represented. In hot weather progesterone also needs to be refrigerated which could pose an additional challenge in the heat of northern Western Australia. Conversely, until 2019, iatrogenic PTB amongst Aboriginal women giving birth at the tertiary hospital was lower than pre-intervention rates, and run chart analysis showed reductions in the 32-36 week age group through 2014-2016 and again in 2018-2019. These results indicate that the cultural change preventing unnecessary medically initiated births that was so effective in non-Aboriginal women at the tertiary hospital, may in fact also have had some impact on the timing of birth amongst Aboriginal women.

Gestational age/year Non-aboriginal or Torres Strait Islander women Aboriginal and Torres Strait Islander women
Outside of the tertiary centre there appeared to be a small non-significant reduction in PTB amongst Aboriginal women in the four years following the introduction -Stillbirth rates are compared using logistic regression analysis with year 2019 as a reference, univariately and after adjustment for maternal age, nulliparity, grandmultiparty, ethnicity, smoking during pregnancy, maternal asthma, low socioeconomic status, history of stillbirth, placental abruption, antepartum haemorrhage for reasons other than placental abruptions and placenta praevia, gestational diabetes, pre-existing diabetes or hypertension, threatened abortion, threatened preterm labour, IVF conception, caesarean at last delivery and history of PTB. Rates that were significantly different from 2019, both univariately and with adjustments for maternal characteristics are shown by ↓ for lower rates and ↑ for higher rates relative to the 2019 rates -Stillbirth numbers were suppressed when < 5 cases occurred and the upper bounds for these rates were calculated based on 5 cases Non-tertiary centres of the Initiative. However, in 2018 and 2019 PTB rates were slightly elevated from pre-Initiative rates. Unlike the results from the tertiary centre, the increases in recent years were driven by an increase in medically initiated births between 32 and 36 weeks. This increase in medically initiated late PTB was similar to that observed at non-tertiary hospitals among non-Aboriginal women. At a state level, the results indicate that the interventions of the Initiative had minimal impact on Aboriginal women, with the exception of the prevention of some late iatrogenic PTB at the tertiary centre, likely resulting from the aforementioned cultural shift at this hospital. The question remains as to why the Initiative was less effective amongst Aboriginal women than amongst other women giving birth in WA. In a study of Aboriginal women giving birth in Perth, a history of maternal hypertension, vaginal bleeding and consumption of excess alcohol were identified as independent predictors of low birth weight or PTB, with maternal education and smoking also being important risk factors [14]. While smoking cessation campaigns have demonstrated success in the broader population, focus groups have revealed that due to social and economic pressures, smoking cessation is not a priority for many pregnant Indigenous women [15]. Other publications have identified the higher rates of inadequate antenatal care [16] and socioeconomic disadvantage [17] as contributing to the higher rates of PTB in Indigenous women.
Data from New South Wales showed that with extensive efforts from government and healthcare providers, maternal smoking rates declined substantially from 1994 to 2007 across the state. However, the reduction amongst Aboriginal women was minimal and was strongly linked to socioeconomic position [18]. While a reasonable reduction in smoking rates was observed in Aboriginal women across the study period, this reduction was minimal amongst Aboriginal women who gave birth at the tertiary hospital, where the reductions in PTB were observed amongst non-Aboriginal women. This discordance in the response to public health messaging could explain the lack of uniform effect of the Initiative, particularly at the tertiary hospital and highlights the need for specific strategies targeting PTB in Aboriginal women. Community-led programs have shown promise in reducing reported alcohol use during pregnancy [19], while services co-designed with Aboriginal Community Controlled Health such as 'Birthing on Country' have been shown to reduce PTB rates in community trials [20]. It is positive that the reduction in medically initiated late PTB observed at the tertiary hospital was also beneficial to Aboriginal women and their babies. However, the outcomes across the state demonstrate that appropriately designed and culturally sensitive programs such as midwifery continuity of care models with Aboriginal representation will be required to improve perinatal outcomes more broadly, and that more targeted and collaborative approaches must be employed.

Implementation of the initiative at non-tertiary hospitals
In non-tertiary hospitals the PTB rate rose amongst both Aboriginal and non-Aboriginal women, primarily driven by an increase in iatrogenic late PTB. Unfortunately, despite the increased prematurity aimed at reducing the chances of stillbirth, there was no reduction in the stillbirth rate at non-tertiary hospitals. The success in WA at reducing PTB rates at the tertiary hospital demonstrates that we have the knowledge and capability to lower the rate of PTB. The factor limiting effectiveness is the implementation of these strategies in the non-tertiary environment. The recipe to change culture and make it happen in the non-tertiary environment requires a sustained and collaborative model, with ongoing engagement and advocacy, such as the work that the Australian Preterm Birth Prevention Alliance engages in. As a result of their efforts, and in recognition of the disparity in results between the tertiary and non-tertiary hospitals, the lessons of the Initiative have now lead to embracing additional methodologies funded by the federal government, based on the breakthrough collaborative model (BTS). In the coming years, 52 hospitals from around Australia will participate in the BTS, a program which will bring together teams seeking reductions in PTB based on existing best practice already implemented in the tertiary hospital. The BTS model has demonstrated previous success in improving quality healthcare, including in the maternity environment with substantial reductions in caesarean section rates being observed in participating hospitals following the implementation of the model [21]. In time this model may be used to reduce the PTB rate in the non-tertiary environment.

Changes in the stillbirth rate
The higher stillbirth rate amongst Aboriginal women compared to non-Aboriginal women is likely due to their higher burden of obstetric conditions and pre-existing maternal conditions. While there was some variation in the stillbirth rate between years, at the tertiary hospital where the major reductions in PTB were observed amongst non-Aboriginal women, there was no associated increase in the stillbirth rate.
Increases in the stillbirth rate were observed amongst Aboriginal women giving birth at the tertiary hospital in 2018, and amongst non-Aboriginal women giving birth at non-tertiary hospitals in 2017. Compared to the years prior, neither of these increases were associated with a significant reduction in medically initiated preterm births or the overall preterm birth rate, and therefore are unlikely to be associated with any interventions of the Initiative. Furthermore, the stillbirths amongst Aboriginal women at the tertiary centre in 2018 were almost entirely in the 20-31 week gestational age group, indicating that they could not have resulted from delaying medically indicated deliveries. Further investigations did not reveal any major increases in risk factors in these years that could explain the elevated rates.

Strengths and limitations
The size of the study population was the major strength of this study, with 11 years of data for the whole state being included. Our large sample allowed us to separately investigate the impact of the Initiative on PTB rates amongst Aboriginal women for the first time. It also allowed us to conduct subgroup analyses by hospital, gestational age group, onset of labour and risk of PTB, which revealed valuable insights which may otherwise have been concealed. The use of a population-based prospective cohort study design based on administrative data is appropriate, but comes with the limitation of a lack of concurrent controls which could introduce bias. Our introduction of low and high risk of PTB status mitigates this limitation. Another limitation is that the ethnicity variables used to derive Aboriginal and Torres Strait Islander ethnicity are known to be underreported in population-based health and health related data collections [22]. It is also generally accepted that ascertainment of risk factors amongst this group is less robust. This source of bias was minimised by removing the inconsistencies and assigning Aboriginal status when the majority of a woman's pregnancies were reported as Aboriginal ethnicity. Finally, previous analyses have shown that rates of PTB were higher in private hospitals [8]. In this study we were not able to separate our analysis of non-tertiary hospitals into private and public, which may have concealed differences between these two types of hospitals.