Mapping the trajectories for women and their babies from births planned at home, in a birth centre or in a hospital in New South Wales, Australia, between 2000 and 2012

While women usually choose where they would like to give birth at the beginning of pregnancy, the process is dynamic due to complications or risk factors that may develop, making the pathway or trajectory for women who plan to give birth at home or in a birth centre difficult to predict at a service level. A woman intending a homebirth, for example, may commence her pregnancy with no significant history of illness or pathology only to find her plans changed as the pregnancy continues and a complication arises. This may result in a change of birth setting, either during the pregnancy or in labour; the latter made sometimes more difficult due to a lack of integration between the providers of homebirth and hospital services [18]. In countries where homebirth and freestanding birth centres are well integrated into maternity services (UK, Netherlands), transfers between places of birth are facilitated by local policies and protocols which support the need to change location, including during labour, to the preferred or more appropriate birth setting [19]. By contrast, a maternity system lacking in integration between providers and places of birth, as is common across Australia, creates barriers for a smooth transition from home to hospital where indicated [18].

Transfer rates from planned homebirth to hospital vary by country as well as by parity, with predictably lower rates in multiparous women. The rates of intrapartum transfer from home to hospital in studies over the past 10 years from a number of high-income countries varied from 8.8 to 21.0% overall [4, 9, 20‐22]. When stratified by parity, the rates were 24 to 39.1% for nulliparous women and 4.8 to 12.3% for multiparous women. Transfer from a midwifery unit (either alongside or freestanding) to hospital were 12.4 to 33.9% overall [4, 9, 10, 22‐25] and by parity, 25.4 to 37.8% for nulliparous women and 5.3 to 14.0% for multiparous women. Reasons for intrapartum transfer range from request for analgesia and slow progress in labour (non-urgent) to fetal distress and haemorrhage (urgent) the latter being less common [21, 24, 26, 27].

While transfer rates in NSW have been reported overall, little is known about what happens to women who commence labour in their planned place of birth, and their babies during and after transfer. Anecdotally, support for the expansion of homebirth and birth centre services has been hampered by a belief that this intrapartum change of venue adds a layer of unnecessary risk to women and their babies [28, 29]. This study explores these events during labour, which include planned place of birth, transfer from home or a birth centre to hospital, actual place of birth, mode of birth and neonatal admission to special care nursery/neonatal intensive care unit (SCN/NICU), described as birth trajectories, for a low-risk cohort of women from NSW from 2000 to 2012. This information will aid in our understanding of the intrapartum transfer rate and subsequent interventions and assist with maternity service development and expansion of options for women interested in birth at home or in a birth centre. It will also inform understanding of the costs in different settings, because the costs of birth at home or in a birth centre should include the cost associated with transfer where applicable.

The aim therefore was to investigate the birth trajectories of women at low risk of complications who at the end of pregnancy plan to give birth at home, in a birth centre or a hospital labour ward. The development of this decision tree framework was also undertaken to inform a future costing of these birth settings.

A retrospective population-based cohort study using linked health data was undertaken. The study draws on the framework of decision analytic modelling to construct a decision tree. Typically, a decision tree model provides a simplified framework of decisions that are made at different points in a treatment schedule depending on outcomes or events at a given time, “under conditions of uncertainty” and are mutually exclusive [30, 31]. The decision tree developed for this study depicts the trajectories of women as their labour progressed by analysing linked health data, moving from their plans at the onset of labour to the birth of their child. We report probabilities at each ‘node’ of the decision tree, stratified by parity. We illustrated these trajectories in a decision tree (Fig. 1) with the events (branch) of the decision tree in Table 1.

A decision tree is interpreted left to right, on the left is the decision node representing the planned place of birth at the onset of labour for women with a healthy pregnancy at low risk of complications. The pathways or trajectories represent the events that occurred for these women and their infants and are defined at each ‘chance’ node moving right, from which a branch emanates. The alternative trajectories are mutually exclusive and the probability of each branch is calculated. While decision tree analysis is used for modelling options in terms of end-points and costs, we are using the framework to depict and quantify the trajectories of women by their planned birth setting. We populated the decision tree in this study with data analysed from a linked population-based data set obtained from NSW Ministry of Health.

This study investigates the trajectories of women in New South Wales who planned to give birth in the birth settings described above. During the study period (2000–2012), there were six alongside BCs and three freestanding BCs in NSW. Freestanding BCs accounted for approximately 15% of BC births between 2000 and 2012. The data did not permit separate analysis by type of BC.

The number of maternity hospitals in NSW has remained constant over the period. The majority of homebirth services were and still are provided by midwives in private practice who are employed directly by women. There are a small number of publicly funded homebirth services which are staffed by midwives employed by public hospitals [32, 33].

Data for all women who gave birth in NSW between January 2000 and December 2012 and all babies born between January 2000 and December 2012 of greater than 400 g and 20 weeks gestation were included. Four datasets were linked:

The cohort was derived from the Perinatal Data Collection (PDC) which records all births in NSW from public and private maternity service providers, including homebirths [1]. Women were included if they were at low risk of complications, that is:

Given that this study aimed to examine the trajectories of women who planned to give birth in the three available settings in NSW, we classified the women according to planned place of birth as recorded in the PDC. This dataset was obtained for the Birthplace in Australia Study, a national data linkage study of maternal and perinatal outcomes by place of birth (home, birth centre or hospital) [12]. A detailed description of the methods for selecting the women included in this study is described in Cheah et al. [30]. Briefly, women were excluded if they had any identified pregnancy complication (Table 2). For the remainder who laboured spontaneously between 37 and 41 completed weeks, we assumed their place of birth at the onset of labour was as planned.

We stratified the decision tree by parity to investigate the impact and events related to planned birth settings, as the demographic details are significantly different for nulliparous women compared to multiparous women. Women who have an unplanned homebirth (born before arrival (BBA)) and those who free-birthed (that is, gave birth without a registered health provider present) were not included in this cohort. The age range indicated here corresponds with the age range categorised as ‘A’ in the Australian College of Midwives Consultation and Referral Guidelines [34]. Category ‘A’ refers to women at low risk of complications who fall under the scope of practice of a midwife. If a variance occurs, the Guidelines recommend the midwife consult either another midwife, a medical practitioner or refer the women to be overseen by a medical practitioner for secondary or tertiary care, depending on the significance of the variance.

Data were received and analysed in SPSS V24. Groups were established according to the women’s intended place of birth as recorded in the PDC. The trajectories were determined using descriptive statistics to map the events that occurred throughout the labour, birth and postnatal period. These events represent the intended place of birth at the onset of labour, transfer to hospital (in labour or post-partum), mode of birth, and neonatal events including admission to special care nursery and neonatal intensive care. Data indicating mode of birth were missing in both nulliparous (120 cases) and multiparous (110 cases) hospital groups, therefore these cases were not included in the trajectories. Demographic data were stratified by parity; we used Chi Square test to compare grouped categorical data and univariate general linear model analysis of variance (ANOVA) to determine the differences in the means.

When allocating women who “transferred to HLW” from a BC for the decision tree, interventions such as epidural analgesia and instrumental birth were taken into account as some women who were recorded in the PDC to have given birth in a birth centre had received one or more of these interventions. These women were considered to have had a planned birth centre birth but were transferred to a hospital labour ward. Given these rooms are commonly adjacent to or near the labour ward for an alongside BC, the ‘transfer’ is assumed in this analysis. Freestanding BCs in NSW are not located near obstetric and neonatal services and as such, these women would have physically changed location. The proportion calculated in each branch are conditional on the number in the previous event (to the immediate left), adding up to 100%.

A total of 496,387 women were included in the decision tree. The majority of women (464,630 93.6%) had their intended place of birth recorded as hospital, 29,951 (6.0%) intended a birth centre birth and 1824 (0.4%) intended a homebirth. There were differences in the demographic characteristics of the three groups with women intending a homebirth being older (32 years; standard deviation (SD) 4.7) than those in the birth centre group (30 years; SD 5.1) and hospital group (29; SD 5.3). A higher proportion of women in the hospital group were giving birth to their first baby (nulliparous) (45.1%) compared to the birth centre and homebirth groups (42.7 and 29.9% respectively) and the highest proportion of women with a gestational age of 40 weeks and over were women in the homebirth group (67.1%) compared with women in the birth centre (59.1%) and the hospital group (54%) (Table 3).

Figures 1 and 2 depict the decision tree constructed for this study. The decision node is the planned place of birth, separately for nulliparous and multiparous women. The trajectories the women take from the start of labour are represented by the ‘branches’ which emanate from the chance nodes named at the top of the figure: actual place of birth, mode of birth, neonatal location and length of time in special care nursery/ neonatal intensive care unit (SCN/NICU). Each branch extending from a chance node is given a probability of that event occurring.

Of the nulliparous women, 0.2% planned to give birth at home, 5.7% planned a birth centre birth and 94% planned to give birth in a hospital labour ward. Of the women planning a homebirth in this group, 77.0% remained at home and had a normal vaginal birth (NVB). Of the 23% of women who transferred to hospital during labour, more than half (55.0%) went on to have an NVB. The rates of instrumental birth and caesarean section for nulliparous women planning a homebirth who were transferred to hospital were 22 and 23% respectively (see Fig. 1). These rates are 5.1 and 5.3% respectively when all nulliparous women planning a homebirth are taken into account. Of the women planning a BC birth, 66% remained in the BC and had an NVB. Forty-six percent of the women who transferred to the hospital labour ward had an NVB. The NVB rate for women in the planned hospital group was 62%. Of the women who transferred to the hospital from a BC, the rates of instrumental birth and caesarean section were 37 and 17% respectively. Overall, women in the planned BC group had lower rates of instrumental birth and CS compared with those in the planned hospital group (12.5 and 6.1% versus 23 and 15% respectively).

Multiparous women planning a homebirth had a 92.5% rate of NVB compared with 88.0% in a BC and 93.6% in the planned hospital group. Even following transfer, over 88% of women planning a homebirth had an NVB in hospital. In total, the vaginal birth rate in the multiparous birth centre group was 98.3%. Instrumental birth and CS rates were in the planned homebirth group were 12 and 1% respectively, following transfer (see Fig. 2).

Infants of nulliparous women had higher rates of admission to NICU/SCN than multiparous women, with the largest proportion originating from the women who planned a hospital birth (10.1%). The smallest proportion of neonates admitted to SCN/NICU were admitted following a homebirth (1.7%). Of the planned BC group overall, 7.1% of neonates were admitted to the SCN/NICU. Infants of women who were transferred from home to a hospital in labour had a 16.7% NICU/SCN admission rate however as a proportion of all planned homebirths, the overall SCN/NICU admission rate was 5.1%.

Overall, fewer infants of multiparous women were admitted to the SCN/NICU with total SCN/NICU admission rates as follows: planned homebirth 1.6%, planned BC birth 3.6% and planned hospital birth 6.3%. The highest proportion of infants of multiparous women who were admitted to SCN/NICU were in the planned hospital group, following a CS birth (21%).

While data linkage is a powerful means to examine perinatal outcomes at a population level, there are limitations to the granularity of the data making close investigation of specific events challenging. Transfer from a birth centre to hospital is a good example. This study intended to highlight the trajectories of healthy women who could be reasonably compared across the three birth settings. Transfer from one setting to another is sometimes not recorded in the PDC, particularly when a BC is located within a maternity unit. For this reason, we used interventions such as epidural block and instrumental birth to indicate a transfer from BC to hospital. The numbers of women choosing a homebirth in NSW is very small and the probabilities associated with each trajectory in this group are less certain. With the benefit of linking data from across health datasets from one state, we were able to develop a cohort of women with comparable observable characteristics. It is difficult, however, to account for the unmeasurable or unobserved characteristics of women which fundamentally influence their choice of place of birth. Cases of stillbirth and neonatal death have been retained within the corresponding trajectories as these eventualities contribute to the trajectory of the mother and baby. This framework forms the foundation of a future cost analysis of place of birth using Australian Refined Diagnosis Related Groups. Closer investigation of morbidity and mortality was not within the scope of this paper. However, these outcomes have been reported on a national level in the Birthplace in Australia Study [12].