The NSW State Health Plan ‘Towards 2021’ aimed to reduce smoking rates among pregnant women in NSW by 0.5% per year, to 7.5% in 2015 [1]. Whilst smoking rates among pregnant women in NSW declined from 22.1% in 1994 to 8.3% in 2016 [2], the target of 7.5% has not been met and there are concerns that the rates have plateaued.

Although the risks of smoking during pregnancy are well established, 8.8% of pregnant women in NSW reported smoking in 2019 [3]. A recent study clearly demonstrated the benefits of not smoking during pregnancy among NSW Aboriginal women and showed that babies born to Aboriginal mothers who did not smoke during pregnancy were at a significantly reduced risk of adverse perinatal outcomes compared to infants born to similar mothers who did not smoke [4]. Results from that study are currently being used to inform smoking cessation materials for Aboriginal women and their families. As there are large differences in smoking rates between Aboriginal and non-Aboriginal mothers, smoking cessation strategies which may be effective for Aboriginal women may have little or no effect in non-Aboriginal women. A need for similar evidence on benefits of not smoking during pregnancy among the remainder of the NSW population has been identified. This evidence is needed in both system level planning and individual patient counselling. Hence, this study aimed to quantify the benefits of not smoking during pregnancy for non-Aboriginal NSW mothers and their babies.

The study population was all singleton babies born to non-Aboriginal NSW mothers residing in NSW between 1 January 2012 and 31 December 2016, and their mothers. Births were identified from the NSW Perinatal Data Collection (birth data), which is a statutory record of all livebirths and stillbirth of at least 20 weeks gestation or 400g birthweight in NSW. Women who were recorded as Australian Aboriginal in the birth data or who were assigned Aboriginal status according to the Enhanced Reporting of Aboriginality algorithm used in the previous study [5] were excluded from this study.

The birth data were probabilistically linked with the Admitted Patient Data Collection (hospital data) and the Registry of Births, Deaths and Marriages deaths data (death data). Record linkage was performed by the NSW Centre for Health Record Linkage using personal identifiers, with de-identified data provided to researchers. The rate of false links was low (5 per 1000) [6], meaning it was rare that records belonging to different people were wrongly assessed as belonging to the same person. The hospital data contain information on diagnoses and procedures for all inpatient admissions to public and private hospitals for both mothers and infants coded according to the International Classification of Diseases version 10-Australian modification and the Australian Classification of Health Interventions [7]. The death data, recording fact of death for deaths registered within NSW, was used in conjunction with birth and hospital data to identify neonatal deaths.

The exposure of interest was absence of maternal smoking throughout the pregnancy (‘Non-Smokers’), as opposed to any smoking during pregnancy (‘Smokers’). Smoking was identified through self-report in the birth data and/or a diagnosis code indicating current smoking (Z72.0, F17) in the hospital record associated with the delivery. The sensitivity of current smoking from the most recent separation in the hospital data is estimated to be 58.5% and the specificity 98.4% [8].

Two maternal outcomes of interest were identified from the birth data and the hospital record(s) related to the delivery. Outcomes considered were a composite indicator of severe maternal morbidity which includes transfusion, assisted ventilation and organ failure (Supplementary Table 1 [9]) and inter-hospital transfer (reflecting the need for higher level care). Both these outcomes were binary.

Perinatal outcomes included those occurring at birth and within the first 28 days of life, and were identified from the hospital, birth and death data. Perinatal outcomes were preterm birth (<37 weeks gestation), birthweight less than the 3^rd and 10^th centiles for gestational age and sex [10], severe neonatal morbidity, and perinatal death (stillbirth and neonatal death) and its components. Severe neonatal morbidity was measured using a validated composite indicator [11] containing procedures and diagnoses associated with severe morbidity and was calculated amongst live births only (Supplementary Table 2).

Maternal age was obtained from the birth data. Other covariates included any hypertension and any diabetes and were obtained from the birth and hospital data. Socioeconomic status and remoteness were assigned based on the statistical local area of residence of the mother using the NSW ranking of the Australian Bureau of Statistics 2011 Socio-Economic Index for Areas (SEIFA) Index of Relative Socio-Economic Disadvantage and the 2011 Remoteness Areas. Hospitals were grouped according to birth volume, location and ownership [12].

Unadjusted and adjusted relative risks were calculated using modified Poisson regression with robust error variances. All analysis was performed in SAS [13]. Given the established causal relationship between smoking and adverse perinatal outcomes, we also quantified the proportion of adverse perinatal outcomes that would not have occurred in this population if all the mothers had been non-smokers during pregnancy. We used the formula: PAF = [Ps(RRs-1)]/RRs, where Ps is the proportion of babies with the given outcome whose mothers smoked and RRs is the adjusted RR for smokers. The RRs is the inverse of the RR for non-smokers.

Between 2012 and 2016 there were 488,768 babies born to 382,268 mothers in NSW. Of these, 20,961 (4.3%) babies were identified as having Aboriginal mothers (15,438 mothers). After restricting the population to singletons and NSW residents there were 449,590 babies born to 358,308 non-Aboriginal mothers (Figure 1).

Most (92%) mothers reported not smoking during their pregnancy and this proportion increased slightly over time, from 90.8% in 2012 to 92.8% in 2016. Mothers who reported not smoking in pregnancy were more likely to be older, be having their first baby, live in an area with the least disadvantage (i.e. more likely to be high socioeconomic status), live in a city, and not suffer from any chronic conditions (Table 1). The same proportion (8%) of smoking and non-smoking mothers suffered from hypertension in their pregnancy, and a slightly greater proportion of non-smoking mothers had a diagnosis of diabetes than smoking mothers (12.4% vs 10.5%).

Overall rates of severe maternal morbidity and transfer to another hospital during the birth admission were low (<3%) and both outcomes were lower among non-smoking mothers than mothers who smoked (Table 2). These differences remained statistically significant after adjustment (Table 2). Not smoking during pregnancy was associated with a 13% reduction in risk of severe maternal morbidity (adjusted Relative Risk, aRR: 0.87 (0.81,0.93)) and 8% lower risk for transfer during the birth admission (aRR 0.92 (0.86,0.99)).

Babies born to non-smoking mothers had substantially lower risks of all adverse perinatal outcomes, compared with babies born to mothers who reported smoking during their pregnancy (Table 3). These differences remained statistically significant after adjusting for maternal age, socioeconomic status, parity, any hypertension and any diabetes. Adjusted relative risks varied from as low as 0.36 for being born with a birthweight lower than the third percentile for gestational age and sex, to 0.69 for being stillborn (Table 3). As indicated by the PAFs (%) in Table 3, between 3.7 and 11.4% of all these adverse perinatal outcomes were attributable to smoking in this cohort of babies.

This study quantifies the benefits of not smoking during pregnancy for non-Aboriginal mothers and their babies in NSW. The reduction in risk of all adverse perinatal outcomes for babies whose mothers did not smoke during pregnancy was considerable. After adjusting for the effects of maternal age, socioeconomic status, parity, any hypertension and any diabetes, babies born to mothers who reported not smoking during pregnancy had a 31% lower risk of being stillborn, 34% less risk of dying in the first 28 days of life, a 42% lower risk of being born preterm, 52% less risk of being born small for gestational age (< percentile) and a 64% lower risk of being born with a birthweight lower than the third percentile for gestational age and sex. The PAFs for the adverse perinatal outcomes highlight the potential for the reduction in the rates of these adverse events in NSW if smoking rates during pregnancy could be reduced. Currently there is a focus in Australian maternity care on reducing the rates of stillbirth (The Safer Baby Bundle) [15] and preterm birth (the focus of the Australian Preterm Birth Prevention Alliance) [16]. Our findings show that among singleton babies born to non-Aboriginal women in NSW, 5.6% of preterm births and 3.7% of stillbirths are attributable to maternal smoking during pregnancy. These fractions are likely to be higher in areas with higher smoking rates. Addressing maternal smoking is an important contributor to reducing both stillbirth and preterm birth rates. Across Australia, rates of smoking during pregnancy range from 5.6% in the Australian Capital Territory to 20.7% in the Northern Territory, with an overall rate of 10.2% in 2019 [17] Although all states and territories have seen a reduction in smoking over the last decade, in some areas smoking rates have started to increase. Reducing smoking rates in regions with higher smoking rates could have greater even returns in stillbirth and preterm birth prevention.

Consistent with the widely-documented association between smoking and socioeconomic status, non-smoking mothers tended to be less disadvantaged, older, reside in cities and have had fewer previous pregnancies than mothers who smoked during their pregnancy. Almost one third of the mothers who smoked lived in an area classified as the most disadvantaged SEIFA quintile and/or were aged less than 25 years. Mothers who are young and/or of low socioeconomic status are known to be at higher risk of smoking and less likely to quit, both in Australia and overseas [2, 18–20]. However, similarly to the findings of a Victorian study which considered absolute and relative risk reduction in tobacco control policy [18], each high risk group comprised only a small proportion of mothers, with the greatest number of smokers in the 25–29 year age group and resident of a major city. The authors of the Victorian study commented that high risk group approaches only have the potential to make very small reductions in overall smoking during pregnancy rates, and argue that although these priority groups should not be forgotten, they must not detract from population-wide and cost-effective policies that have been shown to reduce the prevalence of antenatal smoking [18]. Elsewhere this is referred to as the ‘Prevention Paradox’ [21]. A recent study in NSW also illustrated the benefit of targeting groups with higher numbers, rather than rates, of smokers [2]. The same study also highlighted that smoking rates in NSW are not distributed evenly across the 15 Local Health Districts and showed that over half the mothers who smoked during their pregnancy lived in just four Local Health Districts [2]. Targeting these four Local Health Districts with an effective program to reduce smoking in pregnancy has the greatest potential to reduce adverse perinatal outcomes, including stillbirth and preterm birth.

As the costs of any intervention need to be balanced against the potential benefits, the results of this study are potentially important inputs into modelling the impact of smoking cessation interventions in NSW and broader economic analyses that can inform strategic decision making. Births with adverse events such as preterm birth and stillbirth, are associated with increased healthcare costs [22, 23] and savings through avoiding these via reductions in smoking can be balanced against the costs associated with an intervention. A recent American study by Bacheller et al used similar estimates from the American population to assess the cost-effectiveness of a hypothetical smoking cessation intervention [24]. As there are differences between smoking rates, demographics and healthcare provision and access between Australia and the United States, it is important that these data be available on a local population.

The results of this study could be used to enhance current training for NSW Health staff on delivering smoking cessation support during pregnancy, including online modules offered by The Health Education & Training Institute [25]. Although the risks of smoking during pregnancy are well documented, presentation of the benefits of not smoking may be more beneficial in encouraging pregnant women to stop smoking. A recent review of attitudes towards smoking cessation programs found that healthcare workers found it difficult to communicate health advice on smoking during pregnancy without making the pregnant woman feel guilty and damaging the relationship with the pregnant woman [26]. The same study reported pregnant women feeling pressured and stigmatized for smoking. Positive reframing of the situation to present the expected benefits of not smoking may be more effective in prompting behavior change [27]. Highlighting the proportion of small for gestational age and other adverse outcomes that would be avoided if antenatal smoking rates were negligible could be particularly motivating at both the health service and individual levels.

Results might also inform an NSW-wide plan for enhancing clinician training, clinical care standards to improve the management of smoking before, during and after pregnancy, and monitoring and management of the performance of NSW Health services. Reducing antenatal smoking and increasing quitting during pregnancy are key performance targets for Local Health Districts [28]. The findings of this study highlight the flow on benefits to the health service of lower rates of smoking during pregnancy in terms of adverse outcomes and associated healthcare burden avoided, which underscores the public health significance of the afore-mentioned performance targets and may provide an additional incentive to change.

In addition to providing system level insights, this study provides local information, which can be used by health professionals to further engage the community on the benefits of not smoking for mothers and their babies. A similar study [4], focusing on the benefits of not smoking in Aboriginal women is being used to inform culturally relevant educational material for that population. The data from the current study can likewise be used to tailor advice given to local women. Studies have found that policy makers, practitioners and researchers tend to value locally generated evidence over studies conducted abroad [29, 30].

A key strength of this study is that it was co-produced by academic researchers at Women and Babies Research and policy makers and practitioner-scholars at the NSW Ministry of Health, from conception and planning through to results dissemination and translation. This way of working has been shown to increase the policy-relevance of research, the translation of results into practice, and the exchange of knowledge and skills [31–34]. Another strength of this study is that it was a large population-based cohort study capturing data from almost half a million babies. The main limitation is the lack of information on variables of interest such as heaviness of smoking as well as potential confounders such as alcohol consumption. In addition, smoking status was based on self-report supplemented by diagnosis codes within the medical record. As such, this might underestimate the smoking rate and bias effects towards the null.

Babies born to mothers who reported not smoking during their pregnancy were at a significantly reduced risk of all adverse maternal and perinatal outcomes compared with those born to mothers of similar demographics who reported smoking during their pregnancy. Mothers who reported not smoking during pregnancy had a small reduction in their risk of morbidity and of being transferred to another hospital.