Despite major global improvements in maternal and neonatal health during the Millennium Development Goals era, unacceptably high levels of preventable morbidity and mortality remain in many areas of the world . Reductions in neonatal mortality (deaths within the first 28 days of life) have lagged behind those of overall mortality in children under 5 years of age, and as a result neonatal mortality accounts for 45% of total under-5 mortality worldwide . Prematurity and associated complications are now the most frequent cause of death in all children younger than 5. Improvements in newborn survival have been slower in sub-Saharan Africa with current trends suggesting it will take over a century to achieve rates of newborn survival comparable to North America or Europe . Sub-Saharan Africa has the highest stillbirth rates of any region, an under-recognised and neglected global public health issue responsible for 2.6 million third trimester fetal deaths worldwide in 2015 .
Despite evidence suggesting that 71% of neonatal deaths could be averted through scale-up of existing interventions, interventional studies in low-resource settings are limited . This is reflected in the WHO guidelines on maternal and newborn care, which are derived primarily from studies done in high-income settings [6–8]. Extrapolation of efficacy estimates between settings is problematic as seen in the Antenatal Corticosteroids Trial (ACT)  and Fluid Expansion As Supportive Therapy (FEAST) trials [10,11]. There is therefore a major requirement for further efficacy studies of many interventions in low- and middle-income countries. In addition to this, there is an ongoing need for dissemination of current research findings in these settings in order to target future research appropriately.
This review presents the current evidence from studies done in LMICS over the last five years. We included studies or systematic reviews which reported on interventions targeted at (i) the prevention of preterm birth and stillbirth and (ii) decreasing mortality in preterm and low birth weight infant and ill newborns. This review aims to update existing guidance from the Every Newborn Action Plan [6–8] and report on the latest evidence from recent studies.
A systematic literature review was done for studies published between (January 2013-May 2018) which reported on interventions in both mothers and infants from LMICs with the outcomes of 1) reducing preterm birth incidence 2) reducing stillbirth incidence 3) reducing neonatal mortality in preterm, low-birth weight, or unwell infants. The review adhered to MOOSE and PRISMA guidelines for reporting of systematic reviews and was pre-registered with Prospero (CRD42018099173). Key outcomes were defined as following: stillbirth was defined as baby born without signs of life after 28 weeks gestation; preterm birth was defined as Baby born before 36 weeks gestation; neonatal mortality was defined as death of baby born live, before 28 days of age .
MEDLINE, EMBASE, Web of Science, LILACS, CINAHL, Global health, MIDIRS, and the Cochrane Library were searched using keywords related to the above outcomes and limited to LMICs as defined by World Bank groupings . The search was time-limited in order to build on previous reviews published in 2014 as part of the Lancet Every Newborn Series . The keyword searches were done separately for maternal and neonatal interventions. Selected studies were screened by two reviewers (EW, DW). Inclusion criteria were: intervention studies published 2013-2018 in pregnant women or preterm newborns reporting on outcomes of preterm birth, stillbirth or neonatal mortality conducted in LMIC (or systematic reviews reporting >50% LMIC data or providing LMIC subgroup analysis. Accepted study designs were RCTs, cohort studies, case/control studies or before/after studies based in either community or hospital settings on both singleton and multiple pregnancies. Studies were excluded if they did not adequately report sample size calculation, or if they reported on wider public health interventions not exclusively aimed at pregnant women. Editorials, commentaries, reviews, conference abstracts and trial protocols were excluded, as were studies not published in the English language.
Reference lists from relevant articles were also searched. Details of the search strategy can be found in Table S1 in the Online Supplementary Document. The PRISMA flow diagrams of search strategies for maternal interventions and neonatal interventions can be seen in Figure 1 and Figure 2.
Figure 1. PRISMA flow diagram for maternal intervention study selection.
Figure 2. PRISMA flow diagram for neonatal intervention study selection.
In light of the results from the ACT trial showing potential harm from antenatal corticosteroids in LMICs, and the pending publication of the WHO ACTION trials , we excluded studies related to antenatal corticosteroids and preterm birth prevention .
All studies which met these criteria were entered into an evidence gap map which is a visual representation of the studies in each area, demonstrating the volume of evidence in each area . antenatal and delivery care (Table 1), lifestyle interventions (Table 2), health systems, training and guidelines (Table 3), pharmacological interventions (Table 4), nutritional supplements (Table 5), and community groups (Table 6). Neonatal interventional studies were grouped as: infection prevention and treatment (Table 7), respiratory support (Table 8), cardiovascular support (Table 9), health systems, training and guidelines (Table 10), feeding and nutrition (Table 11) and community mobilization (Table 12). Complex interventions were disaggregated where possible and the relevant data for each of the above groups was assessed independently.
Table 1. Maternal interventions – characteristics of individual studies (antenatal and delivery care)
Table 2. Maternal interventions- characteristics of individual studies (lifestyle)
Table 3. Maternal interventions- characteristics of individual studies (health systems, training and guidelines)
Table 4. Maternal interventions- characteristics of individual studies (Pharmacological interventions)
Table 5. Maternal interventions- characteristics of individual studies (nutrition supplements)
Table 6. Maternal interventions- characteristics of individual studies (community groups)
Table 7. Neonatal interventions- characteristics of individual studies grouped by intervention type (infection prevention and treatment)
Table 8. Neonatal interventions- characteristics of individual studies grouped by intervention type (respiratory support)
Table 9. Neonatal interventions- characteristics of individual studies grouped by intervention type (cardiovascular support)
Table 10. Neonatal interventions- characteristics of individual studies grouped by intervention type (Health systems, training and guidelines)
Table 11. Neonatal interventions- characteristics of individual studies grouped by intervention type (feeding and nutrition)
Table 12. Neonatal interventions- characteristics of individual studies grouped by intervention type (community mobilisation)
A total of 81 studies of maternal interventions (68 implementation studies and 13 systematic reviews) were included in analysis. Fifty-five studies reported on data from Sub-Saharan Africa, 25 from South Asia, 20 from East Asia and Pacific, 15 from the Middle East and North Africa, 12 from Latin America and the Caribbean and 2 from Europe and Central Asia.
Ninety-eight studies of neonatal interventions (79 implementation studies, 19 systematic reviews) were included in analyses. 36 reported data from Sub-Saharan Africa, 35 from South Asia, 12 from East Asia and Pacific, 10 from the Middle East and North Africa, 14 from Latin America and the Caribbean, and 12 from Europe and Central Asia.
Thirteen studies were identified reporting on the impact of antenatal care [17–28]. Five of these showed antenatal care was associated with a reduction in stillbirth [17,20,22,25,26], two of which also showed reduction in neonatal mortality [25,26], two studies found antenatal care reduced a composite outcome of stillbirth, preterm birth and low birthweight [18,19], and four studies showed no significant difference in our outcomes [21,23,24,27]. A meta-analysis found antenatal care reduced both stillbirth (relative risk (RR) = 0.82, 95% Confidence Interval (CI) = 0.73-0.93) and neonatal mortality (RR = 0.80, 95% CI = 0.72-0.8) .
Two studies evaluated the effect of introducing routine ultrasound scanning during antenatal care [27,30]. Neither found any improvement in stillbirth or preterm birth reduction, although both led to increased identification of fetal problems. This included a large cluster randomised controlled trial encompassing 51 423 births, with stillbirth reduction (RR = 1.09, 95% CI = 0.97-1.23) and neonatal mortality reduction (RR = 0.99, 95% CI = 0.86-1.14) .
Genito-urinary infection management
Three randomised-controlled trials reported on genito-urinary infection management, focussing on testing and treating for bacterial vaginosis during the second trimester [31–33]. Two of these trials found no significant difference in any outcomes[31,32] however one trial comparing bacterial vaginosis treatment vs placebo in women with abnormal vaginal flora found a significant reduction in preterm birth (RR = 1.65, 95% CI = 1.04-2.63) . A prospective cohort study in China evaluating the introduction of routine syphilis testing at first antenatal appointment, estimated a 39.4% reduction in stillbirth and an 8.8% reduction in preterm birth .
Twenty three studies reported on antenatal nutritional supplements [35–57]. Folic acid, zinc, calcium and multiple-micronutrient supplementation were all found to reduce preterm birth, whereas vitamin D, vitamin C and iron had no effect when given alone.
Three studies reported folic acid supplementation before conception and during pregnancy [37,50,52]. All found a significant reduction in preterm birth and this effect size was increased if taken from 3 months before the last menstrual period . There were two studies with zinc supplementation, a Cochrane review on zinc alone showed a significant reduction in preterm birth rate (RR = 0.86, 95% CI = 0.76-0.97)  and a RCT using zinc in a combination tablet with other multivitamins was associated with a reduction in stillbirth, preterm birth and early neonatal mortality . There were two RCTs with vitamin D supplementation one of which found no significant effect and the other found a reduction in preterm birth . Vitamin C alone was found to have no difference in outcomes  however another study combining vitamin E and C observed preterm birth reduction . Three studies used iron supplement regimes, pre-natal iron had no effect , neither did testing and treating anaemia . Although low dietary iron was significantly associated with stillbirth and preterm birth (RR = 0.12, 95% CI = 0.036-0.377) and early neonatal mortality (RR = 0.23 95% CI = 0.15-0.35) , a systematic literature review comparing daily with intermittent iron found no significant difference in preterm birth (odds ratio (OR) = 1.82, 95% CI = 0.75-4.4) . Low dietary calcium was associated with increased preterm birth (RR = 0.76, 95% CI = 0.65-0.88)  and accordingly calcium supplementation was associated with a reduction in preterm birth (RR = 0.76, 95% CI = 0.60-0.97) .
Seven papers compared maternal micronutrient supplementation (MMN) with combined iron and folic acid supplementation [35,38,45,47–49,56]. All of these studies, including a Cochrane review, found significant reduction in both preterm birth and stillbirth with MMN compared with iron and folic acid alone. This effect was even more pronounced in women with anaemia or low body mass index (BMI) [47,48].
We identified nine papers including two systematic literature reviews and two Cochrane reviews evaluating the impact of community groups, all of which demonstrated significant reductions on stillbirth and neonatal mortality [26,29,58–63]. Three of these used community based intervention packages to deliver antenatal care- for example upscaling home visits by community health workers [26,29,64]. Seven papers reported on setting up women’s groups as a means of providing peer counselling, community support, and increased prioritisation of women’s health issues [58,59,61,64–66]. Improvements were seen in clean delivery practices, early breastfeeding, improved nutrition during pregnancy and improved health care seeking for neonates [63,64]. Having high population coverage and high proportions of pregnant women participating were both significant predictors of effect .
Two RCTs assessed the effect of exercise during pregnancy in women who were overweight or had gestational diabetes mellitus (GDM) [68,69]. One found no reduction in preterm or stillbirth, but did find a significant reduction in GDM (22.0% vs 40.6%; P < 0.001) . The other involved both an exercise regime and dietary changes and this was associated with reduction in preterm birth (OR = 1.64, 95% CI = 1.14-2.36) .
Indoor air pollution
One RCT trialled a low-emission ethanol cook stove (as opposed to traditional kerosene stoves) to determine the differential effects of ethanol vs kerosene cook stoves on pregnancy outcomes . There were no significant difference in rates of preterm birth or stillbirth, but there was a significant increase in extended perinatal mortality (7.9% vs 3.9% P = 0.045) (stillbirth or death within first 28 days of life) associated with kerosene cook stoves.
Five trials involved malaria prevention during pregnancy [71–75]. Interventions included bed net provision plus indoor residual spraying (IRS), community scheduled screening and treatment plus intermittent preventative treatment (IPT) with Sulphadoxine Pyramethanine (SP) and all showed reductions in preterm or stillbirth. IRS led to decreases in neonatal mortality (17.2 vs 1.5% P = 0.006), stillbirth (7.5% vs 0% P = 0.03) and placental parasitaemia . These reductions were seen despite poor compliance with the full course of SP – even with Community Health Worker (CHW) home delivery and Directly Observed Treatment Short-Course (DOTS), only 43% of women received the recommended 3 doses .
One systematic literature review reported on anti-helminthic treatment, focusing specifically on community-based eradication programmes . This found no significant difference in preterm birth or stillbirth reduction (RR = , 1.54 95% CI = : 0.93-2.58).
Four studies evaluated the effect of guideline implementation for pregnancy care [18,77–79], and all had some beneficial effects on either preterm birth, stillbirth or neonatal mortality. The first study evaluated the impact of a guideline for the first ANC visit consisting of a checklist for health care workers. This led to a reduction in a composite outcome including stillbirth, preterm birth and low birthweight (RR = 0.72, 95% CI = 0.65-0.93) . The second implemented a labour management guideline in the form of a pocket book ad wall posters, and observed a significant reduction in stillbirth (RR = 0.66, 95% CI = 0.53-0.82) . Mgaya et al. found the introduction of a guideline for diagnosis and management of obstructed labour was followed by a reduction in perinatal mortality (stillbirth and neonatal mortality within the first 72 hours of life) from 16% to 8.8% (P = 0.01) . Finally, Okonofua et al. found guidelines for diagnosis and management of eclampsia reduced the case fatality rate from 15.1 to 3.2% (P < 0.001) .
Health systems strengthening
Five studies evaluated efforts to strengthen health systems [78,80–83]. A variety of methods were used though mainly included audit or quality improvement cycles where baseline data was collected and used to identify key areas of deficiency before prioritising and implementing intervention packages to address these [78,82]. In general, improvements were seen in outcome measures and this resulted from improvements in practice as well as more efficient and effective use of resources [82,83]. For example, a Ghanaian study trialling a series of QI bundles developed following a prolonged needs assessment, and observed a 36% reduction in stillbirth . A Tanzanian study ran a series of audit cycles and achieved significant reductions in stillbirth and perinatal death , and a Zimbabwean study found a series of changes in leadership and accountability led to a reduction in intrapartum stillbirth to almost zero .
Nine implementation studies examined the impact of staff training on maternal and neonatal outcomes [18,61,78–81,83,85,86]. All studies reported some level of improvement in stillbirth, neonatal mortality or maternal mortality to varying degrees. Programmes involving repeated sessions reinforcing knowledge were more effective than one-off sessions [79,86]. Programmes were only effective if administrative authorities were involved from the outset [79,86]. Involving pregnant women in training and education was also found to improve outcomes in one study .
There were eight intervention studies about delivery care including one systematic literature review [22,87–93]. Four studies compared facility with home delivery or delivery with traditional birth attendant [87,90,91,93]. All studies, including a systematic literature review  found that facility delivery conferred worse outcomes both for the neonate and the mother with significantly higher rates of stillbirth and maternal mortality. The systematic literature review evaluated data from 9 studies from sub-Saharan Africa including 47475 women, and found no significant change in perinatal mortality between home and facility delivery (OR 1.21, 95% CI = : 0.79-1.84) but did show a significant increase in maternal mortality . Khanam et al. conducted a prospective cohort study in Bangladesh covering 24271 births and found a significant increase in perinatal mortality (OR = 2.4, 95% CI = 2.08-2.76) .
There were ten trials of pharmacological interventions for the reduction of preterm and stillbirth. Three studies investigated the impact of antiretroviral therapy (ART) [94–96]. One found no significant difference in rates of preterm birth in mothers receiving ART  and the other two found ART significantly reduced preterm birth [94,96]. This effect was increased if ART was started before conception . One study compared single dose antibiotic prophylaxis to a multi-day course for routine C-section and found there was no significant difference in stillbirth or neonatal mortality between the two . Vaginal progesterone for tocolysis in twin pregnancies was not found to increase gestational age at delivery , and a comparison of IM progesterone and nifedipine for threatened preterm labour found no significant change in either time to delivery or rates of Neonatal Intensive Care Unit (NICU) admission .
Periodontal disease management
One small randomised controlled study of antiseptic mouthwash provision along with dental education found no significant differences in preterm (OR = 1.59, 95% CI = 0.51-4.92) or stillbirth (OR = 1.01, 95% CI = 1.06-12.22) .
Feeding & nutrition
There were twenty-two studies of nutritional interventions in small or sick newborns to reduce neonatal mortality [101–123]. Three large-scale randomised controlled trials on preterm neonates found no beneficial impact on neonatal mortality from vitamin A supplementation, a result echoed by a subsequent meta-analysis including these studies .
A meta-analysis of three community studies in Ghana, Nepal, and India reported a significant reduction in all-cause neonatal mortality associated with initiation of breastfeeding within 24 hours compared with commencement later than 24 hours (OR = 0.56, 95% CI = 0.40-0.79) .
Thermal regulation & homeostasis
The burden of neonatal hypothermia in babies born in hospital in LMICs has been estimated at 32%-85% and is an important contributor to preventable neonatal mortality, with one study reporting an 80% increase in adjust mortality for every degree Celsius drop in first observed body temperature [125,126]. Seven studies reported on thermal regulation [110,127–132]. A recent Cochrane review and meta-analysis of Kangaroo Mother Care (KMC) found in an LMIC subgroup analysis a reduction in all-cause neonatal mortality by 43% (OR = 0.57, 95% CI = 0.37-0.89), highlighting the significant potential impact of this low-cost intervention. One study reported a multi-faceted quality improvement project targeted at reducing transitional hypothermia in very low birthweight infants including implementing radiant warmers in the delivery room and specialised transport equipment along with comprehensive staff training. The initial results of this small study of 192 neonates showed a sustained improvement in normothermia (56% normothermic on arrival to NICU compared with 19% pre-intervention) and a significant decrease in mortality .
Infection prevention & management
Thirty studies reported on this important aspect of neonatal care [102–106,109,111–115,117,122,123,130,133–147]. Topical emollient therapy was suggested in the data analysis for the Every Newborn series to have a significant impact on neonatal mortality , however, this has been challenged by subsequent research. A recent Cochrane review including 5 studies published in the last 5 years showed no significant reduction in neonatal mortality (RR = 0.94, 95% CI = 0.81-1.08) or incidence of invasive bacterial infection .
Probiotics and synbiotics have been studied for their role in preventing necrotizing enterocolitis (NEC) and associated mortality. A randomized controlled trial of 4000 newborns in India showed a 40% reduction in a combined endpoint of sepsis/death (RR = 0.60, 95% CI = 0.48-0.74) associated with once daily administration of the probiotic Lactobacillus Reuteri . Importantly this study excluded neonates born before 35 weeks gestation or <2000g in birthweight however a meta-analysis of probiotics in preterm infants in LMICs also showed a significant reduction in all-cause mortality (RR = 0.73, 95% CI = 0.59-0.90), incidence of NEC (RR = 0.46, 95% CI = 0.34-0.61)and incidence of late-onset neonatal sepsis (RR = 0.80, 95% CI = 0.71-0.91) .
We found 25 studies exploring respiratory support in LMICs [150–173]. The implementation of bubble continuous positive airways pressure devices (bCPAP) has been shown to be feasible in multiple LMIC settings, in secondary and tertiary level facilities [151,153–156,160,163,166,170,171]. 2 recent systematic reviews have examined the impact of this on neonatal outcomes, specifically in LMICs [152,161]. Although there were no RCT data on mortality available, a pooled analysis of 4 observational studies showed bCPAP to have a major impact on mortality in preterm neonates (OR = 0.34, 95% CI = 0.14-0.82) . Importantly, a study from a neonatal unit in Malawi showed that this beneficial effect is strongly mediated by neonatal temperature, with hypothermic neonates failing to respond to bCPAP , emphasising the importance of basic neonatal care as a requirement prior the addition of more complex interventions such as bCPAP.
Nine middle-income country studies were identified investigating the feasibility and impact of surfactant administration to premature neonates with neonatal respiratory distress syndrome (RDS) [150,151,158,159,162,167,169,172].Predominantly these were small-scale studies of either differing methods of surfactant administration or comparisons of early vs late administration, both of which showing inconsistent results. It appears that surfactant is a feasible intervention to implement in LMICS, and a meta-analysis showed reductions in mortality in eligible infants comparable with those observed in high-income settings (RR = 0.67, 95% CI = 0.57-0.79)[159,175]. Unfortunately, however, the high cost of the intervention and the infrastructure required in terms of delivery and ventilatory support is likely to leave surfactant beyond the reach of many low-income facilities for some time to come .
We identified 11 studies investigating aspects of community mobilization with regards to neonatal mortality reduction [25,59,63,107,131,176–181]. These largely focused on predominantly maternal interventions (eg, participatory women’s groups) and so are discussed above in the maternal interventions section.
There were two studies of interventions for cardiovascular therapy [182,183]. A Brazilian study of very low birthweight neonates with patent ductus arteriosus (PDA) showed a significant mortality reduction with pharmacological or surgical treatment of PDA compared with conservative management . A Cochrane review reported low-quality evidence suggesting paracetamol to be as effective as indomethacin in PDA closure however highlighted insufficient follow-up data to establish evidence on neurodevelopmental outcomes of this treatment .
Training programmes & resuscitation
Nineteen studies described implementation of training programmes [63,86,128,130,139,141,157,181,184–195]. Thirteen of these described training programmes in neonatal resuscitation, specifically the Helping Babies Breathe programme (HBB). Neonatal resuscitation was identified as a potentially very impactful intervention in the Every Newborn series. Studies investigating this subsequently have highlighted additional complexity in low-resource settings. A large study of HBB implementation in Nepal involving over 40 000 deliveries showed a significant reduction in intrapartum stillbirth and neonatal mortality in the 1st day of life however, no reduction in overall perinatal mortality (stillbirth or neonatal death within 7 days of life) [186,192].Other studies in India, Kenya and Tanzania, and 2 recent meta-analyses showed similar results [184,185,194–196]. These data suggest that although HBB training was effective in prolonging life of babies with intrapartum-related complications or those previously categorised as fresh stillbirths, the quality and availability of care in the postnatal period was not sufficient to sustain these gains . In addition, although babies with severe asphyxia may have been successfully resuscitated, in reality the preferred intervention would have been earlier delivery by timely Caesarean-section. This highlights the importance of a systems-based approach to improving the continuum of care.
Health systems strengthening & guideline implementation
Twenty studies were identified that described aspects of guideline implementation or health systems strengthening to improve outcomes of preterm babies in LMICs [59,63,104,128,130,135,142,144,145,157,177,178,180,181,188,191,197–200].
One initiative which targeted several different health system issues and showed impressive results was the development of a health partnership between 4 hospitals (2 university, 2 district) in Rwanda and a team of neonatologists and neonatal nurses from the United Kingdom . They conducted a detailed needs assessment in all facilities and identified 6 major areas of need including nutrition, respiratory support, thermoregulation, and enhanced record keeping. Targeted intervention packages were developed for each of these needs, with examples being KMC utilisation for thermoregulation and training staff in breast feeding support and developing systems for storing expressed breast milk to improve nutrition. The total budget of the project was only US$45 000 and yet overall in-hospital neonatal mortality decreased from 23.6% to 21.7% in university hospitals and from 10% to 8.1% in district hospitals (P = 0.056).
On a smaller scale Harris et al reported results from a pre- post-implementation trial aiming to improving neonatal outcomes in a single Ugandan neonatal unit . Protocols for management of unwell newborns were developed using literature and local consensus and disseminated through ward meetings and regular ward-based teaching. Audits at 3 and 6 years showed a reduction of neonatal mortality from 0.58 deaths per admission to 0.26 and 0.17 deaths respectively. The sustained nature of this improvement without major investment in infrastructure is particularly striking.
This review has highlighted a set of measures with a strong evidence base and the potential to prevent a large proportion of stillbirth and preterm birth, and reduce neonatal mortality in low- and middle- income settings. Maternal interventions found to be successful were antenatal care, multiple micronutrient supplementation and malaria prevention. Successful neonatal interventions included vitamin A supplementation, probiotics and early breastfeeding initiation, KMC and bCPAP. Staff training and guideline implementation and community mobilisation were beneficial for both mothers and neonates. Many of these interventions are relatively simple and it is clear that new innovations are not a priority- we need innovative ways to successfully implement the interventions we know can work.
Strengths and limitations
Our study has several important strengths. Our explicit focus on LMIC evidence delivered an important set of conclusions with direct applicability to LMIC policy-makers, something that is increasingly crucial in the context of the ACT trial and other recent findings. By intentionally limiting our review to the last 5 years we were able to focus on changes to the evidence base since the Every Newborn series and highlight important interventions with new recommendations or that merit further investigation such as optimising neonatal thermal care, and minimising indoor air pollution. The production of evidence gap maps is a relatively novel interactive way of presenting this visually and can play an important role going forward in understanding the current state of evidence for maternal and newborn health interventions. By investigating maternal and newborn interventions in conjunction we were able to draw important conclusions regarding the continuum of care involved and the role of whole-system approaches to improving outcomes.
There are also several limitations to consider. First, although the review was intentionally limited to the last 5 years to focus on changes to the evidence base since the Every Newborn series, this could have resulted in missing important evidence generated prior to 2013 but not included in the Every Newborn publications however we feel this is unlikely. We only included maternal interventions delivered to pregnant women, thereby excluding other population-level interventions delivered to all women of childbearing age which may have an impact on preterm birth and stillbirth incidence (eg, access to birth control) [201,202]. LMICs are a broad and heterogeneous group and some of our conclusions may not be transferable to LMIC contexts other than those where the studies took place.
Strategies for stillbirth reduction
Many stillbirths can be prevented by relatively simple measures. Facilitating condition recognition and diagnosis of high-risk pregnancy is the cornerstone and this can in large part be achieved by enabling early antenatal care attendance, providing appropriate staff training and guidelines, and ensuring women are able to access the health care they require. Additionally, the ability to perform timely delivery is crucial- this means the ability to induce labour or perform emergency Caesarean-section without undue delay caused by lack of facilities, limited staffing, or infrastructure (eg, transport) problems.
Access to quality antenatal care is a key priority and has been clearly demonstrated to reduce perinatal mortality [17,20,22,26,203]. Data from Ghana showed that receiving high-quality antenatal care halved risk of stillbirth as it enabled provision of malarial treatment and prophylaxis, screening for anaemia, helminth management and blood pressure monitoring. In addition it facilitated Prevention of Mother to Child Transmission of HIV (PMTCT) and syphilis detection and treatment. It also allowed provision of nutritional supplements and an opportunity for educating women about ways to improve their health and recognition of danger signs in pregnancy . Modelling has predicted that the provision of 10 basic services in antenatal care could avert 45% of all stillbirths . As part of strategies to reduce both perinatal and maternal mortality, there has been increasing efforts in many LMICs to encourage women to give birth in health facilities in order to allow early complication recognition and life-saving interventions. However, in many settings facility based delivery has been associated with paradoxically worse outcomes compared with home delivery [22,87,90,93]. This is multifactorial, and in part can be attributed to increased likelihood of high-risk cases being selected for facility delivery. In Ethiopia, a cohort study of 4442 women found that women who experienced intrapartum complications were twice as likely to deliver in a health facility than those who did not, explaining in part the fact there was no stillbirth reduction in facility births . However, this is not the only modifying factor and quality of obstetric care available is likely to be a key determinant. One study using Demographic Health Survey (DHS) data from Malawi for example found that, even when controlling for the risk profile of patients, facility delivery conferred poorer outcomes for both the mother and the baby . The additional risk burden is likely to be attributable to the fact that increasing uptake of facility delivery does not translate into increasing availability of high quality CEmONC care. This has been demonstrated in multiple low- and middle-income settings where, even when women are able to reach health facilities, access to the care they need is limited by lack of medical supplies such as blood, or lack of staff with sufficient expertise and training . This highlights the need for coordinated systems and infrastructure development in conjunction with service provision in order to provide the necessary care.
Strategies for preterm birth reduction
Preterm birth reduction is complex and involves national level commitment to improving maternal health and well-being. Specific interventions with a strong evidence base in LMICs include optimising nutritional intake, which is important for both micro- and macro-nutrients, and there is a wealth of evidence for different nutritional supplements which reduce preterm birth rates [5,36,42,43,46,47,55,56,206]. Appropriate HIV management is also important ,as is adequate malarial prophylaxis [71,73,75].
Strategies to improve women’s health at a population level however must be in tandem with advocacy for women’s rights. Intimate partner violence is a threat to women’s well-being worldwide and lack of female empowerment means many women are unable to make safe decisions about their reproductive health. Improving access to contraception, family planning, and abortion is crucial in forwarding female- and therefore maternal health.
Strategies for neonatal mortality reduction
Our findings have shown a number of interventions which can reduce neonatal mortality. Early initiation of breastfeeding, KMC and probiotics are all simple, low-cost interventions which could be easily implemented at low cost. There are also promising results from use of bCPAP and surfactant. Introduction of HBB training showed promise in mortality reduction however also served to highlight the need for systems strengthening alongside HCW training.
Kanagroo Mother Care (KMC) to aid thermal regulation in babies <2000g birthweight has been rolled out across LMICS as part of the Every Newborn Action Plan . Currently in the majority of settings KMC is initiated only on stable neonates once they have completed any required treatment and so unstable babies on treatment remain at high risk of hypothermia. The results are awaited of an ongoing WHO multi-site LMIC trial of immediate KMC (i-KMC) where KMC is initiated immediately after birth regardless of other ongoing treatments  and may provide an important recommendation of the benefit of KMC even in unstable babies in the first days of life.
This review has also shown important changes to the evidence base subsequent to the Every Newborn Action Plan series of reviews and emphasised the importance for LMIC health policy development of considering evidence derived in LMICs. We have derived important conclusions regarding general and intervention-specific barriers and enabling factors to implementation, which are essential to consider when developing new maternal and newborn health policy initiatives.
Our review has identified that while much of the burden of stillbirth and preterm birth can be prevented by interventions already available, the real challenge is in the successful implementation of these strategies. A number of common themes arose in our review highlighting some of the key areas which must be addressed in order to create successful strategies.
Health systems strengthening
Many of the deficits in health care in LMICs are due to weak health care systems and a lack of robust protocols and support systems. These are often deeply ingrained, multi-faceted and challenging to address, however without such attempts, only limited improvement in outcomes can be achieved.
Quality improvement projects are a burgeoning area of research in LMICs as they require relatively low resource input and can effectively target areas of need. Success is however dependant on investment of facility staff in the cycle and in the ongoing re-evaluation of outcome measures . Some successful attempts to improve this have included use of “local champions” and also leadership boards allowing inter-site comparisons . Several studies delivered and evaluated HBB through QI methodological approaches. The value of these were apparent, one example being the utilisation of intermediate data in the form of ‘run charts’ to highlight the impact of staff turnover on outcomes and supplement advocacy at hospital administration level . In an era of renewed global focus on quality of care, pragmatic trials and learning health systems, these approaches are likely to prove highly important [210–212].
One initiative which targeted several different health system issues and showed impressive results was the development of a health partnership between 4 hospitals (2 university, 2 district) in Rwanda and a team of neonatologists and neonatal nurses from the United Kingdom . They conducted a detailed needs assessment in all facilities and identified 6 major areas of need. Targeted intervention packages were developed for each of these needs, with examples being KMC utilisation for thermoregulation and training staff in breast feeding support and developing systems for storing expressed breast milk to improve nutrition. The total budget of the project was only US$45 000 and yet overall in-hospital neonatal mortality decreased from 23.6% to 21.7% in university hospitals and from 10% to 8.1% in district hospitals (P = 0.056). The project emphasized the potential for locally-driven health systems strengthening through partnership alongside government support and also highlighted the important barrier to training of high staff turnover which hampered the development of institutional memory with regards to new working practices.
In many LMICs, the majority of health care is not performed by doctors but by clinical officers and other lower skilled health care workers. There has been much success in introducing algorithmic approaches to health care, allowing lower skilled workers with less expertise to assess and manage sick patients, however for guidelines to be successful, they must be implementable and accessible to the staff who use them. Relevant care providers should be identified and involved in establishing standards and developing guidelines and protocols which are locally relevant. This ensures guidelines are applicable, attainable, and gives clinicians ownership over them, meaning they are more likely to be adhered to .
The most successful models were those combining guidelines with training programs and education, and much greater levels of adherence were achieved [18,79,84]. This was augmented further with the use of audit cycles to evaluate the guidelines and improve on them and their usage over time . Ultimately Guidelines will only lead to improved quality of care if they are used in daily clinical practice therefore staff need to be supported in order for this to happen with reminders, monitoring and feedback .
Staff training is imperative in achieving optimal maternal and neonatal care. A systematic literature review evaluating perinatal mortality found that across 9 centres in Sub-Saharan Africa, perinatal mortality was 21% in facilities than with home deliveries . Although some of this may be because more complex pregnancies are more likely to attend facilities for delivery, much of this is due to the inability of staff to appropriately manage complications. In Ethiopia for example, only 1.3% of facilities were able to provide Basic Emergency Obstetric and Neonatal Care (BEmONC) .
One-off staff training programmes rarely produced successful outcomes and there is a need for repetition of training allowing reinforcement of new skills and knowledge over time [79,86]. Training should take place in the context of regular audits of outcomes allowing identification of successes and also of failures to feedback to staff allowing continued development of knowledge and skills .
It is crucial to involve authorities such as Ministry of Health or hospital administrative staff from an early stage to ensure the staff are supported and encouraged to attend training sessions, and also to adopt new practices. Administrative bodies must also be involved to enable the supply of the necessary equipment to implement new skills and practice [79,86].
Necessary staff must be identified and enabled to attend the training session. In health facilities, a critical mass of training staff is needed in order to effect systems change  · This is particularly challenging to achieve in settings where staff retention is poor and turnover is high, as is frequently the case in low resource health facilities .Strategies need to be put in place to incentivise staff retention or facilitate frequent training provision and again this will mandate involving administrative staff and relevant officials from the outset .
As well as improving the skillset of staff, it is important to involve women in the outcomes of their pregnancy and to empower them with increased skills and advocacy to effect change both on their own outcomes, but also the wider health services in their community . When trying to improve maternal and child health, this must take place in parallel with increasing the voice and the rights of women, and education is the cornerstone of this approach.
Community groups and community mobilisation are examples of relatively low-cost, low-resource, low-intensity innovations with the potential for significant effect. There are various models however they are mainly based around participatory learning and action cycles run within small groups of women led by a trained (and usually salaried) facilitator . This model enables identification and prioritisation of problems, planning and implementation of strategies which are locally feasible, and re-assessment and evaluation .
There are a number of reasons why groups such as these can be so effective. They require minimal resources and empower communities to address locally relevant health determinants. They facilitate capacity building at a community level but also enable increased advocacy for local issues . It is worth mentioning however, a large RCT performed as part of The Global Network study which implemented a package of interventions including community mobilisation and also health centre quality improvement and staff training. Despite being rolled out across five countries with considerable funding and ongoing support, they saw no improvement in their outcome measures over a two year period . There are a number of possible explanations for this however it highlights the fact that, as with many of the interventions discussed in this review, community mobilisation can only have a significant impact on outcomes if it occurs alongside improvements in the abilities of the health service to manage obstetric emergencies effectively and in a timely manner.
Research as part of routine care
Much of the emerging LMIC data highlighted in this review emphasizes the need for whole-system approaches to delivering individual interventions, and the concept of creating ‘learning health systems’ (LHS) has much to offer in delivering this and driving health system improvement in LMICs. In an LHS, data capture processes are optimised and embedded into routine care, clinical data entered once can then be repurposed many times for administration, quality improvement and research . Although the technological and human resource infrastructure required to implement LHS are often lacking in LMIC contexts , there are important examples of where this has already been shown to be possible, such as the Clinical Information Network of paediatric departments in Kenya . In addition the current lack of infrastructure offers the opportunity to implement LHS now, rather than retrofit them as has been necessary and very costly in many high-income contexts  · The potential for using LHS to conduct rapid pragmatic trials at low cost is particularly appealing in LMICs given the lack of evidence base for many intervention in these settings, and can support ongoing developments in LMIC quality improvement methodology as described above. LMICs stand potentially to benefit the most from learning health systems as a means of unifying fragmented approaches and producing systems capable of continuous improvement.
Using intervention evidence generated in LMICs, this review has highlighted several areas regarding preterm birth and stillbirth reduction, and the management of small and ill newborns, where the evidence base has changed significantly since the Every Newborn Action Plan. The key point to emphasise is the potential to deliver marked reductions in preterm birth and stillbirth with relatively simple interventions, however these must be delivered as part of an approach of a whole-system strengthening to be effective. Learning health systems can offer an opportunity to bring the current fragmented context in many LMIC health systems together and provide important means of understanding implementation challenges and running rapid pragmatic trials, delivering locally relevant data at a low cost and driving continual improvement in quality of care.
Ethics: Ethics approval not required as only published aggregated data used.