We collected data on 79
774 eligible women, of whom 64
538 were low risk, from 142 (97%) of the 147 trusts providing home birth services, 53/56 (95%) of freestanding midwifery units, 43/51 (84%) of alongside midwifery units, and a sample of 36 obstetric units (figure). Of the initial sample of 37 obstetric units, five did not agree to participate and were replaced by resampling from within the same stratum, and one failed to establish data collection successfully. Overall 74% (203/274) of participating units or trusts achieved the target response rate of 85% or more. More than 96% of records had complete data relating to the primary outcome and confounder variables (see appendix 7 on bmj.com). Based on data recorded on the initial forms, neonatal morbidity data were requested for 3.5% of births, and 94% (2615/2770) of these forms were returned; maternal morbidity data were requested for 1.9% of births, and 93% (1388/1490) of these forms were returned.
Flow of participants through study
The characteristics of women and their babies varied by planned place of birth (table 1). Compared with the obstetric unit group, women planning to give birth at home were more likely to be older, white, have a fluent understanding of English, and live in a more socioeconomically advantaged area. The characteristics of women in the freestanding midwifery unit and alongside midwifery unit groups tended to fall between the obstetric unit and home birth groups, with women in the alongside midwifery unit group generally more similar to the obstetric unit group. The biggest difference between the groups was for parity: 27% of the planned home birth women were nulliparous compared with 46% of the freestanding midwifery unit women, 50% of the alongside midwifery unit women, and 54% of the obstetric unit women.
Table 1 Characteristics of healthy women with low risk pregnancies by their planned place of birth at start of care in labour. Values are numbers (percentages) of women unless stated otherwise
There were marked differences between planned places of birth in the proportion of women with complicating conditions identified by the attending midwife at the start of care in labour (table 1). Almost 20% of women in the obstetric unit group had at least one complicating condition noted at the start of care in labour, compared with ≤7% in each of the other settings. This finding was unexpected and suggested that the risk profile of the “low risk women” varied between the different groups. Before the analysis of the outcomes, the co-investigators and independent advisory group agreed to modify the analysis plan to include additional analyses of outcomes restricted to women without complicating conditions at the start of care in labour.
For the three non-obstetric unit settings, transfer rates were much higher for nulliparous women (36% to 45%) than for multiparous women (9% to 13%) (table 2). The timing of transfer, before or after birth, also varied by planned place of birth and parity (table 2).
Table 2 Transfers during labour or immediately after birth among healthy women with low risk pregnancies by their planned place of birth at start of care in labour. Values are numbers (percentages) of women
There were 250 primary outcome events and an overall weighted incidence of 4.3 per 1000 births (95% confidence interval 3.3 to 5.5) (table 3). Intrapartum stillbirths and early neonatal deaths accounted for 13% of events, neonatal encephalopathy for 46%, meconium aspiration syndrome for 30%, brachial plexus injury for 8%, and fractured humerus or clavicle for 4% (see appendix 8 on bmj.com for distributions by planned place of birth).
Table 3 Primary outcome* for babies of heathy women with low risk pregnancies by their planned place of birth at start of care in labour. Categorised by parity for all women and restricted to those without complicating conditions at start of care in (more ...)
Overall, there were no significant differences in the odds of the primary outcome for births planned in any of the non-obstetric unit settings compared with planned births in obstetric units (table 3). For the restricted sample of women without any complicating conditions at the start of care in labour, the odds of a primary outcome event were higher for births planned at home compared with planned obstetric unit births (adjusted odds ratio 1.59, 95% confidence interval 1.01 to 2.52) but there was no evidence of a difference for either freestanding or alongside midwifery units compared with obstetric units.
In the subgroup analysis by parity, the odds of the primary outcome for nulliparous women was higher for planned home births than for planned obstetric unit births (adjusted odds ratio 1.75, 1.07 to 2.86; table 3). The strength of this association was increased when the sample was restricted to women with no complicating conditions at the start of care in labour (adjusted odds ratio 2.80, 1.59 to 4.92). There were no significant differences in the odds of the primary outcome for nulliparous women in the freestanding midwifery unit or alongside midwifery unit groups compared with the obstetric unit group. For multiparous women there was no evidence of a difference in the primary outcome by planned place of birth. The overall test for interaction (heterogeneity) was of borderline statistical significance for all women (P=0.06), and was significant for women with no complicating conditions at the start of care in labour (P=0.03). The pairwise tests for each non-obstetric unit birth setting versus the obstetric unit group showed that this interaction was only statistically significant for the home birth group (all women P=0.01, no complicating conditions P=0.006), indicating that the differences seen are unlikely to be due to chance variation.
Most individual perinatal outcomes were rare, and adjusted odds ratios could not be estimated because of the small numbers of events (see appendix 8 on bmj.com for individual perinatal outcomes). Babies were significantly more likely to be breast fed at least once for planned births at home and at freestanding midwifery units compared with planned obstetric unit births.
The odds of receiving individual interventions (augmentation, epidural or spinal analgesia, general anaesthesia, ventouse or forceps delivery, intrapartum caesarean section, episiotomy, active management of the third stage) were lower in all three non-obstetric unit settings, with the greatest reductions seen for planned home and freestanding midwifery unit births (table 4). The proportion of women with a “normal birth” (birth without induction of labour, epidural or spinal analgesia, general anaesthesia, forceps or ventouse delivery, caesarean section, or episiotomy9 10
) varied from 58% for planned obstetric unit births to 76% in alongside midwifery units, 83% in freestanding midwifery units, and 88% for planned home births; the adjusted odds of having a “normal birth” were significantly higher in all three non-obstetric unit settings (table 5). For other maternal outcomes (third or fourth degree perineal trauma, maternal blood transfusion, and maternal admission to higher level care), there was no consistent relation with planned place of birth, although these adverse outcomes were generally lowest for planned births in freestanding midwifery units (table 4 and appendix 8 on bmj.com).
Table 4 Interventions for healthy women with low risk pregnancies by their planned place of birth at start of care in labour
Table 5 ”Normal births”* for healthy women with low risk pregnancies by their planned place of birth at start of care in labour. Results for all women and restricted to those without complicating conditions at start of care in labour
When the analysis was restricted to units or trusts with a response rate of at least 85%, the higher odds of the primary outcome for nulliparous women in the planned home birth group remained, and the strength of this association increased (appendix 5 on bmj.com). The odds of the primary outcome were also higher for nulliparous women in freestanding midwifery units compared with obstetric units for the subgroup of women without any complicating conditions at the start of care in labour (adjusted odds ratio 2.29, 1.17 to 4.47; test for heterogeneity P=0.07).
The propensity score analyses did not affect the interpretation of the results and are described in detail in appendix 6 on bmj.com.