Our analysis of midwife-led care in alongside midwifery units in NRW revealed superiority of this model of care with respect to our primary endpoint mode of birth (spontaneous vaginal versus instrumental vaginal and CS). Since the recruitment took place at all but one AMUs in NRW, the control group consisted of matched, low-risk women, and an intention-to-treat protocol was applied we are confident our result is valid.
This result is in line with our previous investigation; there, we found a trend towards higher spontaneous and lower instrumental vaginal births (15). The comparison with other studies is difficult for various reasons; these pertain to differences in the study design, the choice of the control group, and variations in the provision of maternity care, among others. Three studies were restricted to low-risk women expecting an uncomplicated labor. Bernitz et al. 2011 in their randomized single-center study used operative birth as primary endpoint to examine three models of care (midwife-led unit, normal, and special unit, all located within one hospital); no difference was found (17). McLachlan et al. 2012 performed a randomized study and compared caseload with standard obstetric care. The spontaneous vaginal birth rate was higher in the former model (RR 1.13, 95%-CI 1.06-1.21) (2). The “Birthplace in England Study” analyzed the perinatal outcome according to intended place of birth; compared to obstetric units the adjusted odds ratio for spontaneous vaginal birth was higher in all models of care under investigation (18). Other investigations and reviews compared home birth or birth in FMUs with hospital birth (7, 3, 8–11).
Three systematic reviews investigated the association between various models of care, including AMU, and obstetric outcomes. Bohren et al. 2017 analyzed continuous support during childbirth versus standard care and found a positive effect on the spontaneous vaginal birth rate (RR 1.08, 95%-CI 1.04-1.12). Continuous support was not limited to midwives but included any labor companion (5). Sandall et al. 2016 in their meta-analysis compared midwife-led models of care - comprising continuity of care during pregnancy, birth and postpartum – with other models of care and found higher spontaneous vaginal birth rates (RR 1.05, 95%-CI 1.03-1.07). Pregnant women of low and high risk were included in the analysis (4). Scarf et al. 2018 restricted their analysis to high-income countries. Here, the odds ratio for spontaneous vaginal birth was higher in birth centers (encompassing AMUs and FMUs) compared to planned hospital births (estimated OR 1.92, 95%-CI 1.59-2.32). The majority of included studies were retrospective; additionally, parity was not accounted for (6).
The non-inferiority analysis of our secondary endpoints revealed varied results: For higher-order obstetric lacerations superiority could be established. In our previous, single-center retrospective investigation the rate of higher-order obstetric lacerations was higher in the study group. A higher mean birthweight and a higher number of newborns with birthweight >4500g in the study group may have contributed to this outcome (15). Other studies and reviews reported a similar trend to our present study: Bernitz et al. 2011 and McLachlan et al. 2012 did not find a difference in the rate of higher-order obstetric lacerations (17, 2). The “Birthplace in England study” revealed a lower adjusted odds ratio of third- and fourth-degree perineal tears for multiparous women giving birth at home or in FMU (18). The systematic reviews by Sandall et al. 2016 and Bohren et al. 2017 did not differentiate the extent of the perineal laceration (4, 5). No difference was found with respect to severe perineal trauma in the systematic review by Scarf et al. 2018 (6).
The perinatal outcome of births in midwife-led models of care has been thoroughly investigated. We defined a composite outcome since severe perinatal morbidity or mortality is a rare event (15). Our study was underpowered for this research question. Nevertheless, the confirmation of non-inferiority in AMU with respect to the perinatal outcome is reassuring. Bernitz et al. 2011 did not find a difference in the perinatal outcome parameters under investigation (17). A lower admission rate to special-care nursery with no difference in the admission rates to the neonatal intensive care unit (NICU) was reported by McLachlan et al 2012 (2). A composite primary perinatal outcome was chosen by the “Birthplace in England” authors. Here, for all but nulliparous women giving birth at home (adjusted odds ratio 1.75, 95% CI 1.07 – 2.86) the primary outcome occurred significantly less often in midwife-led models of care (1). The systematic reviews revealed similar results: No difference was present for selected perinatal outcome parameters (perinatal mortality >24 weeks of gestation plus neonatal mortality; 5’-Apgar score <7; neonatal convulsions; NICU admission) in the systematic review comparing midwife-led continuity models versus other models of care (4). A lower rate of low 5’-Apgar score (RR 0.62; 95% CI 0.46-0.85) was reported in births after continuous support in the systematic review by Bohren et al. 2017. For other selected perinatal outcome parameters (NICU admission, prolonged neonatal hospital stay) no difference was detected (5). Scarf et al. 2018 examined perinatal mortality and NICU admission; no difference was present for these outcome parameters between the different models of care (6).
We could not confirm non-inferiority for our secondary composite endpoint maternal outcome, comprising postpartum hemorrhage and other abnormalities of the third stage of labor. This result is in contrast to our retrospective analysis. There, no difference was present in the postpartum hemorrhage rate between study and control group (15). Likewise, above mentioned studies and reviews did not find a difference in the maternal outcome (17, 2, 18, 4). The result of our composite maternal outcome analysis was mainly determined by a higher postpartum hemorrhage rate in the study group. An explanation for our result may be the overall high postpartum hemorrhage rate (13.5% in the study group, 11.5% in the control group, corresponding to a difference of eight cases). Postpartum hemorrhage was defined as blood loss > 500 ml after vaginal birth (1000 ml after CS, respectively). The diagnosis did not require any quantitative confirmation. We assume that midwives at the study sites had an overall low threshold to diagnose postpartum hemorrhage. Another explanation for our finding may be the eschewal of prophylactic administration of oxytocin. This intervention is known to reduce blood loss >500 and >1000 ml after vaginal birth (19).
We additionally compared selected interventions. We found a significantly lower epidural analgesia rate in the study group. This result is in line with all published studies (17, 2, 18, 4, 5). There was no difference in the episiotomy rate, which was low in both, study and control group (9.8 and 9.3%, respectively). Episiotomy rates vary greatly between countries and healthcare systems, and even within countries and models of care (20, 21).
With respect to hospital stay our results indicated that women in the study group favored early hospital discharge. McLachlan et al. 2012 in their study found a reduction in the length of postpartum hospital stay (55.4 hours, SD 0.97 versus 60.5 hours, SD 0.78, p<0.001) in their caseload group (2). No difference in postpartum hospital stay was detected in the systematic review by Sandall et al. 2016 (4). Our findings may indicate that women opting for care in AMU may have different values with respect to their childbirth and aim for an experience without interventions and minimal contact time with a hospital environment.
More than half (51.2%) of the parturients were transferred to standard obstetric care. Nulliparous women constituted the majority of transfers, and request for epidural analgesia was the most common cause. In Germany, midwife-led models of care do not allow for interventions like oxytocin augmentation or administration of i.v.-opioid analgesia. These factors may have contributed to the high transfer rate. Additionally, the fact that standard obstetric care is available within the same premises; possible without delay; and with continuing care of the parturient by the respective midwife before and after transfer may have lowered the threshold for a decision in favor of transfer.
An explanation for our finding of a predominance of nulliparous women requesting analgesia in transfer rates and causes may be owed to the fact that during the study period a severe shortage of hospital-based midwives was present. Pregnant women were anxious about the quality of care they would receive for their labor. Since care in AMU included continuous one-to-one care, women may have registered for birth in AMU with the major intention to get high-quality care for their birth, and less with the aim to give birth without interventions. Our results confirm our previous retrospective data; here, the transfer rate was 50.3% (15). Transfer rates and causes from midwife-led models of care to standard obstetric care have been extensively examined; in the majority of reports transfers pertain to hospital from home. For all investigations, an association between parity and transfer rate was established.
A German single-center analysis reported 14.6% transfers from home to hospital (David 2006). Transfers were separately analyzed for the “Birthplace in England” study. Here, transfer from AMU occurred in 27.0% (21.0% from FMUs, respectively), with an adjusted odds ratio for transfer from AMU of 2.6 for nulliparous women. Prolonged labor was the most common cause (35%) (22). Blix et al. 2014 in their systematic review reported overall transfer rates between 9.9 and 31.9% (nulliparous: 23.4 – 45.4%; parous 5.8 – 12.0%) (23). In a study from Oregon / USA 16.5% of women were transferred from home to hospital during labor (7). In a New Zealand study, transfer from FMUs occurred in 53.1%; these included transfers before the onset of birth (24, 8). In a study from Denmark, transfer rates of 28.4% from home to hospital were reported (9). Seijmonsbergen-Schermers et al. 2020 described a transfer rate of 55-68% for nulliparous women in the Netherlands (20-32% for parous women, respectively) (20).
Strengths of our study include the prospective design; the conduction in obstetric departments of all levels of care; the meticulous selection of cases and controls – only women entering labor after uneventful pregnancy with a high chance for an uncomplicated vaginal birth were recruited; the analysis according to the intended place of birth; the predefined transfer criteria; and the reporting of the outcome of transferred cases.
Limitations of our study include the size of our study group. The reduction in recruitment was mainly owed to the Covid pandemic which forced all participating study sites to change the booking procedures for birth. This included either a switch from personal to electronic booking or abandoning booking procedures for low-risk women altogether, thereby reducing the chances of recruitment. Additionally, as a result of shortage of staff and raising number of births, midwives of all study sites reported a very high workload during the study period. These time constraints resulted in women not being invited to participate despite their eligibility.