This represents a large cohort of antepartum stillbirths using record linked data and is the first study to specifically examine risk factors for antepartum stillbirth by undelivered pregnancy in an Australian population. We have confirmed that increasing maternal age and nulliparity are important demographic risk factors for antepartum stillbirth in New South Wales. We have also been able to document the risk per undelivered pregnancy by gestational age week for maternal age strata and present absolute risks at term and beyond for older women. These findings will be important to both pregnant women and their clinicians in the provision of accurate information regarding stillbirth risk and timing.
In this setting smoking, pre-existing diabetes and pre-existing hypertension also remain important independent (and potentially modifiable) risk factors for antepartum stillbirth. Smoking rates are extremely high in the Indigenous population in Australia  and represent a particular opportunity for stillbirth prevention where evidence of benefit for cessation strategies exists . The increased risk often attributed solely to Indigenous status is not significant after adjusting for the other variables collected and is evidence that health care provision and access as well as health promotion strategies in pregnancy could be better targeted to these women. Further evidence of disparity appears to exist with our finding that women from non-English speaking backgrounds particularly the Middle East and Africa have higher independent risks of stillbirth in the Australian setting. Whilst we are unable to examine the underlying reasons for this further with this data this is an issue that is widely generalisable to similar high income countries with immigrant minority communities. In such settings it will remain important to ensure equal access and opportunities for the provision of maternity care to all women regardless of background.
The finding that gestational diabetes was apparently protective for antepartum stillbirth on adjusted analyses may relate to a number of factors. It is possible that identified women with this risk factor are receiving more regular antenatal care and monitoring and have therefore a reduced risk. They are also more likely to be induced and less likely to go post term. In this cohort 9.8 % of women with gestational diabetes delivered at 41 weeks gestation or beyond compared with 20.3% of the total sample. It may also be possible that we have been unable to adjust for other confounders not collected in population data. Recent population based studies have actually shown no increased risk for stillbirth for women with gestational diabetes [26, 27] with further evidence of low fetal death rates seen in the prospective Hyperglycaemia and Adverse Pregnancy Outcome (HAPO) cohort  (stillbirth rate of 3.8 per 1000) and no stillbirths in the intervention arm in the Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS) study compared with standard care arm (stillbirth rate of 5.7 per 1000) .
Strengths of our data include the large population base, the very small proportion of missing data (1.6%) and the standardised classification of cause of death. The Midwives Data Collection has previously performed very well in validation studies . We have ascertained timing of death from two independent data sources and are confident of the accuracy of the antepartum coding. This meant we did not have to rely on reporting of Apgar scores which are subjective, have high interobserver variability and are often subject to miscoding [31, 32]. It is possible however that our focus on accuracy of timing may have reduced the overall antepartum stillbirth numbers assessed as our rates of 3.4 per 1000 are slightly lower than other reported rates from similar populations of approximately 4.0 per 1000 .
The quantification of gestational-specific risk of antepartum stillbirth by maternal age has not previously been documented in an Australian setting and represents significant new information for maternity health care providers. Although absolute risks for older women are small, they are consistent with risks of chromosomal abnormalities in the magnitude of 1/192 – 1/378  within the same group of women for which they currently receive considerable counselling and potentially subsequent investigations. Women aged 35 and above and particularly those aged 40 or more should receive some information regarding stillbirth risk and be able to discuss potential options with their health care provider to assist with informed decision making. Ideally the question of balancing benefits and harms of induction of labour at term versus post term needs to be answered in a large randomised controlled trial. This may not be feasible as numbers needed would exceed 46,000 women in order to show a 1 per 1000 difference in late pregnancy stillbirth. Further, the balance of harms in the intervention group is complicated as the comparison of a short nursery admission with respiratory problems compared with a baby dying in utero are neither going to be comparable, nor acceptable to women. One randomised study has used statistical modelling of individual risk factors in women aged 35 or greater to estimate an optimal timing of delivery at term and induce women in the intervention group who had not entered spontaneous labour by this time . Although it was not powered to assess mortality there were reduced adverse outcomes including nursery admission and caesarean section in the intervention arm.
Having access to standardised classification of cause of death for 90% of the stillborn babies analysed is rare for large population based studies and this linkage has allowed us to both exclude deaths due to congenital anomalies and examine cause of death relative to the maternal perinatal risk factors of parity and maternal age. The classification of cause of death uses the PSANZ Perinatal Death Classification system which has been endorsed nationally and has now also been shown to perform well against other classification systems used internationally . Although small numbers did not permit adjusted analyses for the explained causes of death for women aged 40 and above it is interesting to note that deaths due to perinatal infection were more common in these women. This is consistent with recent literature documenting increased incidence of placental histological evidence of infection in term stillbirths [36, 37] and suggests a potential area for future research into underlying mechanisms.
There are however inherent shortcomings in using routinely collected population based datasets. The detail that can be obtained is limited to what information is collected and ascertainment of stillbirth is less accurate for extremely premature gestations . Importantly, maternal height and weight, or BMI are not collected in the perinatal population health databases in New South Wales so we are unable to assess the importance of this risk factor on antepartum stillbirth. As a significant increasing public health problem and independent contributor to poor pregnancy outcomes there is an argument to be made to add this information to population perinatal data collection throughout Australia. We also do not have information on the timing of death just the timing of delivery which may affect the proportion of deaths classified as due to fetal growth restriction.