Impact of neonatal resuscitation trainings on neonatal and perinatal mortality: a systematic review and meta-analysis

Background Training of birth attendants in neonatal resuscitation is likely to reduce birth asphyxia and neonatal mortality. We performed a systematic review and meta-analysis to assess the impact of neonatal resuscitation training (NRT) programme in reducing stillbirths, neonatal mortality, and perinatal mortality Methods We considered studies where any NRT was provided to healthcare personnel involved in delivery process and handling of newborns. We searched MEDLINE, CENTRAL, ERIC and other electronic databases. We also searched ongoing trials and bibliographies of the retrieved articles, and contacted experts for unpublished work. We undertook screening of studies and assessment of risk of bias in duplicates. We performed review according to Cochrane Handbook. We assessed the quality of evidence using the GRADE approach. Results We included 20 trials with 1 653 805 births in this meta-analysis. The meta-analysis of NRT versus control shows that NRT decreases the risk of all stillbirths by 21% (RR 0.79, 95% CI 0.44 to 1.41), 7-day neonatal mortality by 47% (RR 0.53, 95% CI 0.38 to 0.73), 28-day neonatal mortality by 50% (RR 0.50, 95% CI 0.37 to 0.68) and perinatal mortality by 37% (RR 0.63, 95% CI 0.42 to 0.94). The meta-analysis of pre-NRT versus post-NRT showed that post-NRT decreased the risk of all stillbirths by 12% (RR 0.88, 95% CI 0.83 to 0.94), fresh stillbirths by 26% (RR 0.74, 95% CI 0.61 to 0.90), 1-day neonatal mortality by 42% (RR 0.58, 95% CI 0.42 to 0.82), 7-day neonatal mortality by 18% (RR 0.82, 95% CI 0.73 to 0.93), 28-day neonatal mortality by 14% (RR 0.86, 95% CI 0.65 to 1.13) and perinatal mortality by 18% (RR 0.82, 95% CI 0.74 to 0.91). Conclusions Findings of this review show that implementation of NRT improves neonatal and perinatal mortality. Further good quality randomised controlled trials addressing the role of NRT for improving neonatal and perinatal outcomes may be warranted. Trial registration number PROSPERO 2016:CRD42016043668

conclusions Findings of this review show that implementation of NRT improves neonatal and perinatal mortality. Further good quality randomised controlled trials addressing the role of NRT for improving neonatal and perinatal outcomes may be warranted.

IntroductIon
Approximately a quarter of f million neonatal deaths worldwide are as a result of birth asphyxia. 1 A large majority of these deaths occur in low-resource settings and are preventable. Approximately 5%-10% of newborns require some support to adapt to the extrauterine environment and to establish regular respiration. 1 2 Simple resuscitative measures are often enough to resuscitate newborns that may even appear to be lifeless at birth. Studies have shown that essential newborn care has been effective in reducing stillbirths (SB). 3 In developing countries, measures to improve resuscitative efforts through training of basic steps of neonatal resuscitation are expected to reduce birth asphyxia and neonatal mortality. Numerous studies

What this study adds?
► This meta-analysis assessed the impact of NRT on stillbirths, 1-day neonatal mortality, 7-day neonatal mortality, 28-day neonatal mortality and perinatal mortality. ► NRT resulted in significant reduction in stillbirths and early neonatal mortality. However, continuum of care is needed for mortality reduction from day 7 to 28. ► Future studies also need to establish the best combination of settings, trainee characteristics and training frequency to sustain the existing effect on perinatal mortality reduction.

Impact of neonatal resuscitation trainings on neonatal and perinatal mortality: a systematic review and meta-analysis
Open Access have suggested that imparting neonatal resuscitation training (NRT) to healthcare providers involved in delivery process and handling of newborns has the potential to save newborn lives in low-income and middle-income settings [4][5][6][7][8][9][10] Improvements in knowledge and skills of trainees following training programme in resource-limited settings have been reviewed. However, the impact on perinatal mortality outcomes has not been updated in last 5 years. 9 The effect estimates of mortality reduction as a result of training of healthcare providers involved in delivery process and handling of newborns needs to be updated to inform hospital administrators and policy-makers the importance of investing in NRT to sustain and improve neonatal survival. A previous systematic review and meta-analysis 11 assessed knowledge, skills, neonatal morbidity, neonatal mortality in first 7 days after birth and from day 8 to 28. However, it did not include outcomes of stlillbirth, 1-day neonatal mortality or perinatal mortality which has been included in our review.
The objective of this review is to assess the impact of NRT programme in reducing stillbirths, 1-day neonatal mortality, 7-day neonatal mortality, 28-day neonatal mortality and perinatal mortality.

MAterIAls And Methods Inclusion criteria
Types of studies We included relevant randomised, quasi-randomised controlled trials, interrupted time series studies and before-after studies regardless of language or publication status.

Types of participants (population) trained
We considered studies where NRT was provided to healthcare providers (including neonatologists, physicians, nurses, interns, midwives, traditional/community birth attendants, auxillary nurse midwives, village health workers, paramedics) involved in delivery process and handling of newborns in a community (home-based, rural and village clusters) or a hospital (including district hospitals, health centres, dispensaries, teaching/university hospitals, regional hospital, delivery/health centres, local hospitals and tertiary care hospital) setting.

Types of interventions and comparison
Studies in which any NRT was compared with a control group (that received no NRT) or compared with data before the study (pre-NRT vs post-NRT) were included. For this purpose, we considered any NRT programme of healthcare professionals, including the American Academy of Pediatrics' (AAP) Neonatal Resuscitation Program (NRP), Helping Babies Breathe (HBB) or any other training programme that had NRP or HBB as a clearly mentioned component of training methodology.

Types of outcomes measures
We included following outcomes in the review: 1. Stillbirths: defined as number of deaths prior to complete expulsion or extraction of products of conception from its mother.

Data extraction and management
For all studies that fulfilled the inclusion criteria, two reviewers (KK, SB) extracted data (table 1 and 2). Third review author (AP) cross-checked the data and resolved discrepancies. For studies where required data was lacking or could not be calculated, we requested the corresponding author for details.

Assessment of heterogeneity
We assessed heterogeneity amongst studies by inspecting forest plots for the overlap of confidence intervals, analysed statistical heterogeneity through Χ 2 test (P value >0.10) and quantified through I 2 statistics(Chapter 9.5 of Cochrane Handbook for Systematic Reviews). 13 We regarded heterogeneity as substantial if in the Χ 2 test for heterogeneity there was either I 2 >50%, or P value <0.10. We interpreted I 2 values between 0% and 40% as possibly unimportant, 30% and 60% as possibly significant, 50% and 90% as possibly substantial and 75% and 100% as possibly considerable.

Assessment of reporting bias
We used funnel plots for assessment of publication bias if ten or more studies were included in a meta-analysis.

Data synthesis and analysis
We analysed the data using Review Manager V.5.3 software. 14 We conducted meta-analyses for individual studies and reported pooled statistics as relative risk (RR) between experimental and control groups with 95% CI. We explored possible clinical and methodological reasons for heterogeneity, and in the presence of significant heterogeneity, we carried out sensitivity analysis and employed inverse-variance method with Random-effects model. We did not pool randomised and non-randomised (pre-post NRT) studies in the same meta-analysis.

Summary of findings table
We created 'summary of findings' (SoF) table using five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence. We used methods and recommendations described in Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions 13 using GRADEpro software. 15 GRADE working Group grades of evidence were used in the SoF. 16

results search results
We identified 148 records through database searching and 11 records through other sources. After initial screening on the basis of title and abstract, we assessed 47 full-text articles for eligibility and finally included 20 articles in the meta-analysis. The screening details are presented in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram (figure 1).

Included studies
Amongst included studies, two randomised trials addressed the efficacy of NRT in improving neonatal and perinatal outcomes, whereas 18 were pre-post studies. A full description of each study is included in table 1 and 2. All studies were from low-income and middle-income countries. Four studies were done in community setting, whereas 16 studies were carried in hospital setting.
Carlo et al 17 18 assessed baseline perinatal outcomes, then imparted Essential Newborn Care (ENC) training to all which also included basic steps of NRT. They then randomised all clusters that had received ENC training into two groups. One group received an in-depth NRT while the other group did not (control group). For this study we evaluated the pre-ENC outcome of all clusters Open Access Figure 2 Forest plot comparing all SB between the NRT and the control groups. NRT, neonatal resuscitation training; SB, stillbirths.

Figure 3
Forest plot comparing 7-day neonatal mortality between the NRT and the control groups. NRT, neonatal resuscitation training.

Figure 4
Forest plot comparing 28-day neonatal mortality between the NRT and the control groups. NRT, neonatal resuscitation training. and compared them to outcomes of those clusters that received ENC +post ENC in-depth NRT. We therefore did not include this study in the NRT versus control analysis because the control group had also received NRT as a part of ENC training.
The study from Kenya had a complex design of randomisation of health workers to two groups-early training (phase I) or late training (phase II) and did not include a control group without training. 19 Therefore, we analysed this study as before-after study where the rate of stillbirths prior to any training were compared with the rate of stillbirths after all phases of training.
We also included two unpublished trials after permission from authors (tables 1 and 2).

excluded studies
Studies that included interventions that did not qualify as NRT were excluded from the review. These included trainings in safe birthing techniques, 36 Emergency Obstetric and Neonatal Care (EmONC), 37 38 ENC, 39-41 promotion of antenatal care and maternal health education, 42 and newborn care intervention package. 43 Other interventions that did not qualify as NRT [44][45][46][47][48][49][50] or included interventions like neonatal intensive care unit/ special neonatal care unit training 51 52 were also excluded.
Some studies that were subgroups of larger studies like Ersdal et al. 74

⨁⨁⨁◯
Moderate § *I 2 is 67% and the two trials were inconsistent in the direction of effect. Quality of evidence downgraded by two for inconsistency and imprecision (figure 2). †The 95% CI of the pooled estimate includes null effect. Quality of evidence downgraded by one for imprecision ( figure 2). ‡No evidence to support or refute. §Though I 2 is 68%, the 95% CI of the pooled estimate does not include the null effect. Quality of evidence downgraded by one for inconsistency ( figure 5). NRT, neonatal resuscitation training; RCTs, randomised controlled trial; RR, risk ratio.
in the analysis for outcomes where data from 22 Msemo et al 22 were not available.
Risk of bias in included studies has been depicted in table 3.

effects of interventions
Neonatal and perinatal outcomes were reported in majority of included studies. The overall analysis showed a trend towards reduction in neonatal deaths, early neonatal deaths, perinatal deaths and stillbirths with NRT; most of which are statistically significant.
The grade of quality of evidence for the meta-analysis of the trials was moderate to high (table 4).
The quality of evidence for NRT verses control was very low for SB and 1-day neonatal mortality, high for 7-day and 28-day neonatal mortality and moderate for perinatal mortality (table 4). The quality of evidence for post-NRT verses pre-NRT was very low for all our outcomes (table 5).

dIscussIon
This meta-analysis assessed the impact of any NRT programme either by itself or as a part of newborn care package on rates of stillbirths, perinatal mortality, all-cause neonatal mortality on day-1, up till day-7 and till 28th day after birth. We did not evaluate intrapartum-related neonatal deaths or asphyxia/cause-specific neonatal mortality. Mortality in neonates <7 days of life is a proxy measure for intrapartum-related deaths. 43 78 Meta-analysis of before-after studies showed a significant reduction in all stillbirths by 12% (12 studies) and of FSB by 26% (8 studies). The reduction in fresh stillbirths can be attributed to NRT that helps in resuscitating neonates that appear lifeless at birth. 17 18 Of 12 studies, seven studies reported a significant and one study reported a non-significant reduction in fresh stillbirths. However, a non-significant increase in risk of stillbirths was reported in three African studies which blunted the impact of NRT on reduction of stillbirths.
There was reduction in 1-day mortality of 42% (6 studies) and that of 7-day mortality was 18%. All studies included in the analysis (figures 8 and 9) showed a   reduction with an exception of one study. 27 Failure to observe reduction in mortality in Bellad et al could be due to two reasons. First, NRT was provided in diverse health systems within a short period of time. Second, mortality was not assessed in facilities where training was imparted but was measured in the population.
The meta-analysis showed a non-significant reduction of 14% in 28-day mortality. Of the seven included studies only two studies reported a significant reduction in mortality. Resuscitation at delivery helps to reduce neonatal mortality in the first hour of birth when the neonate is at the highest risk of intrapartum-related deaths 3 and the impact diminishes subsequently. For reduction of 28-day neonatal mortality, post-resuscitation specialised care for survivors is required and only NRT is unlikely to have the desired impact on 28-day neonatal mortality. 79 80   Trials that randomise facilities to NRT versus controls (where NRT is not a standard practice) would be ideal to assess the reduction in neonatal mortality. Trials are also likely to result in higher impact as compared with before-after studies as other changes at health facilities or in communities during the time period of before-after studies can confound the results. Because NRT is a standard practice and randomising individuals or clusters to no resuscitation training is unethical, there were only two trials available for the meta-analysis. 20 21 They showed a reduction of 7-day neonatal mortality and 28-day mortality by 47% ( figure 3) and 50% (figure 4), respectively. The perinatal mortality reduced by 37% (figure 5) with no significant reduction in SB rates.
Previously, an expert panel published a systematic review for community-based studies and conducted a meta-analysis that evaluated whether NRT reduced all-cause neonatal mortality in th first 7 days of life. They reported a 38% reduction in mortality which is larger than the 18% (7 studies) reduction observed in the current meta-analysis. Our meta-analysis included community-based studies that resulted in a smaller effect size. Community-based studies (trials or before-after) report a smaller reduction effect on any day neonatal mortality. 8 17 18 47 The reduction in effect size of neonatal mortality in these studies can arise due to several reasons. All births in the intervention community may not be attended by birth attendants trained in neonatal resuscitation, especially if it is a home delivery. 81 82 Second, women may decide to deliver at facilities or homes outside communities where NRT has been imparted. Finally, assessing mortality outcomes in the community can be challenging. Another meta-analysis 11 was published in Cochrane which evaluated outcomes such as knowledge, skills, neonatal morbidity, neonatal mortality in first 7 days after birth and from day 8 to 28. This analysis did not include stillbirths, 1-day neonatal mortality or perinatal mortality that was included in the current meta-analysis.
The current meta-analysis consists largely of beforeafter studies with lack of concurrent control group that limits isolation of effect of resuscitation training alone from other changes at health facilities or in communities during the time period. Other limitation is lack of consistency of settings, duration of training, varying study designs and lack of consistent outcomes which contributed to substantial heterogeneity. Lack of subgroup analysis of type of health facilities may be perceived as a limitation. An improvement in mortality would be maximised in low-resource settings with poor quality of care. However, it is presumed that there is regular training of health workers in basic resuscitation skills in higher levels of care that would translate to higher quality of care. Our recent study 83 84 that evaluated the knowledge and skills of trainees trained in HBB included 384 tertiary-level facilities in India. Only 3% of physicians and 5% of nurses were able to pass the pre-training bag and mask resuscitation skill assessment. 84 Therefore, in the absence of reporting of pre-training skills of health workers in low-resource or high-resource settings or any indicator of quality of care, it would be erroneous to conduct a subgroup analysis based merely on resource settings and mostly will not change the results or the main message of this meta-analysis. We emphasise that despite the heterogeneity in settings, type Open Access  ‡ ‡The effect estimate crosses the line of no effect. Quality of evidence downgraded by one for imprecision ( figure 10). § § Although I 2 is 90%, the effect estimates of all the included studies do not differ in the direction of effect. Quality of effect downgraded by one for inconsistency (figure 11). ¶ ¶ Studies differ in setting, type of NRP and trainees. Quality of evidence downgraded by one for indirectness (table 1 and 2). NRP, Neonatal Resuscitation Program; NRT, neonatal resuscitation trainings; RR, risk ratio; SB, stillbirths.
of training, type of trainees, type of trainers and the duration of training, this study showed an improvement in mortality at and soon after birth.
To conclude, NRT resulted in reduction in stillbirths and improved survival of newborns. The impact on survival of newborns can be further improved by providing a continuum of care beyond 7 days which is not addressed by NRT alone.
The meta-analysis performed showed beneficial effect of NRT in improving neonatal and perinatal outcomes. The models of training were not consistent across studies, with variations in training, trainee and setting. Generalisation of results of the pooled analysis to many currently available programme may not be appropriate. There was evidence of heterogeneity across studies in our meta-analyses; however, overall there is consistency in the direction of effect.
This review identified several important limitations of the current evidence from included studies. Due to inadequate information about the methodology followed and variety of resuscitation programmes in included studies, the quality of the evidence was downgraded for risk of bias and indirectness resulting in inability to adequately assess the effects of this intervention.
conclusIons Implications for practice This review shows that the implementation of NRT improves neonatal and perinatal outcomes.

Implications for research
Further good quality, multicentric randomised controlled trials addressing the role of NRT for improving neonatal and perinatal outcomes may be warranted. Impact of NRT Open Access on improving neonatal and perinatal outcomes as well as the best combination of settings and type of trainee should be established in future trials. More studies need to be done to assess the frequency with which NRT needs to be conducted to sustain the existing effect on perinatal mortality reduction.