There are 2 million deaths each year resulting from childbirth - 814,000 intrapartum related neonatal deaths, over 1 million intrapartum stillbirths and a significant proportion of the world’s 352,000 maternal deaths. Skilled childbirth care is recommended as a universal right to reduce these deaths, yet there is limited mortality evidence of the effect of childbirth care packages. The mismatch between the size of the problem and the quality of the useable evidence is stark. Our primary finding, and the main limitation of our review, is the lack of high or even moderate quality evidence of the effect of childbirth care on neonatal mortality, particularly in low and middle-income countries where the impact would be the greatest. There are a number of reasons for this low level of evidence including the challenges of ethical approval for RCTs testing care that is already considered standard, variations in obstetric packages evaluated, and inconsistencies in outcome measurement.
The variation in terminology surrounding “birth asphyxia” is a key limitation. Consistent case definitions are required; we have used the terminology “intrapartum-related” to classify neonatal deaths due to childbirth-related complications in term infants, however despite recent improvements in clarity, many of the studies identified were older and outcome definitions varied. Furthermore, in settings where the majority of neonatal deaths occur in homes, and outside of vital registration, ascertaining cause of death must often rely on verbal autopsy, which varies with respect to tools , definitions, and hierarchies used. Consistent use of such verbal autopsy tools, and more importantly the hierarchies, is critical . Finally there is a paucity of data from resource-limited settings on intrapartum-related neonatal morbidity, such as neonatal encephalopathy, which requires regular neurologic assessment and is not possible for the majority of newborns in LMIC who are born at home.
The skilled birth attendance studies which we identified were heterogeneous with varying coverage and provider skill levels, and likely underestimated the effect for several reasons. First, the results for the before-after studies reflect that of additional midwife training, since at baseline midwives were already conducting deliveries in the community and attending deliveries, so the baseline effect is not zero. In Matlab, Bangladesh, the magnitude of the effect in the intervention vs. comparison villages was diluted by the low coverage of midwives at birth (only 25%). Furthermore, in many communities, formally trained midwives are only sought for complicated deliveries where the baby is already compromised and could only have been saved by emergency obstetric care, which may not be available.
Given the lack of cause-specific mortality evidence, we followed the LiST rules based on GRADE, and the effect of the 3 obstetric care packages was estimated using Delphi expert consensus . We included a variety of experts with wide geographic representation (geographic region, low-middle and high income settings) and range of expertise and background (clinical, epidemiology, obstetrics, neonatology). Consensus was reached within an IQR of 30%. However, any expert opinion process is clearly limited, and far from ideal.
Nonetheless, the potential for major mortality reductions with skilled intrapartum care, particularly due to intrapartum-related neonatal deaths, is widely accepted and consistent with historical data from the UK, Finland, and Malaysia. Whilst the lack of RCT evidence for the provision or non-provision of childbirth care is understandable, given that it would be unethical to conduct such trials, the dearth of observational studies of quality improvement of childbirth care assessing its effect on neonatal mortality is disappointing and a clear priority for more research. The few significant, large intervention trials of direct relevance for establishing mortality effect estimates were those of community midwife training, EmOC training, and individual interventions to improve labor monitoring and interventions (such as use of the partograph or fetal monitoring) that are reviewed in detail in two other publication supplements [10, 20, 21]. In some studies, there were specific missed opportunities to collect relevant perinatal outcome data. The QUARITE trial, a cluster-randomized trial of quality improvement in obstetric care via emergency obstetric care training (ALARM) and maternal death reviews, is presently underway and has perinatal-neonatal mortality as a secondary outcome . This, and hopefully many more such evaluations, will help to fill a critical information gap.
For the 60 million women who deliver at home world-wide, achieving universal skilled birth attendance may require decades, and in the meantime many preventable deaths occur each year, primarily at community level . TBAs attend up to 40% of births in South Asia, while the majority of home births in Africa are unattended . The evidence for TBA training programs is of low quality and heterogeneous [60, 61, 72, 73] and their role remains controversial. However one recent cRCT which emphasized partnership of TBAs with community health workers and links with the formal health system yielded promising reductions in stillbirth and neonatal mortality . Early recognition of obstetric complications, including obstructed labor, and higher referral rates for emergency obstetric care were observed in this trial, and would presumably be associated with reductions in intrapartum-related injury. Several other studies have evaluated the impact of TBA training on obstetric danger sign recognition and referral [66, 74] ; a meta-analysis reported a small, positive association between training and TBA referral-maternal health service utilization . Given that the skills, role and training of the TBA may vary widely between regions and communities, and that the quality of evidence regarding training effectiveness is low and heterogeneous, the GRADE recommendation for implementation is presently conditional  and we did not attempt to estimate the effect size. However, the potential for TBAs to integrate and partner with the formal health system is promising, and requires further evaluation at scale and in varying contexts.
During the 1990s, the coverage of skilled birth attendance in Sub-Saharan Africa and South Asia increased little, but recent years have seen increases in a few countries. A contributor to the increasing coverage has included demand-side financing (eg voucher schemes or conditional cash transfers in India [75, 76]), eliminating user fees (eg Ghana  and South Africa ) and the introduction of health insurance schemes (eg, Mauritania ), as reviewed recently . Furthermore, innovative strategies to increase the supply of obstetric care have emerged, including task-shifting and the use of non-physician clinicians . In Mozambique, assistant medical officers (técnicos de cirurgia) perform Caesarean section with no difference in complications or mortality rates compared to obstetricians [81, 82]. Training of non-physician clinicians has been prioritized in Ethiopia, Malawi, Zambia and Mozambique, in order to fill the human resource gap. In South Asia, task shifting has involved training general practitioners, nurses and medical officers in obstetrics and anesthesia to expand coverage of EmOC . Increasing the coverage of skilled obstetric care, particularly to reach the poorest, requires creative demand and supply side strategies, with sustained political and financial commitment by governments.