Newborn Resuscitation Training Programmes Reduce Early Neonatal Mortality

Transcription

1 Systematic Review and Meta-Analysis Received: vember 12, 2015 Accepted after revision: January 8, 2016 Published online: May 25, 2016 Newborn Resuscitation Training Programmes Reduce Early Neonatal Mortality Mohan Pammi a Eugene M. Dempsey b, c C. Anthony Ryan b, c Keith J. Barrington d a Section of Neonatology, Department of Pediatrics, Baylor College of Medicine, Houston, Tex., USA; b Irish Centre for Fetal and Neonatal Translation Research, and c Department of Paediatrics and Child Health, University College Cork, Cork, Ireland; d Department of Pediatrics, CHU Sainte-Justine, Montreal, Qué., Canada Key Words Morbidity Mortality Neonates Resuscitation programme Training Abstract Background: Substantial health care resources are expended on standardised formal neonatal resuscitation training (SFNRT) programmes, but their effectiveness has not been proven. Objectives: To determine whether SFNRT programmes reduce neonatal mortality and morbidity, improve acquisition and retention of knowledge and skills, or change teamwork and resuscitation behaviour. Methods: We searched CENTRAL, MEDLINE, PREMEDLINE, EMBASE, CI- NAHL, Web of Science and the Oxford Database of Perinatal Trials, ongoing trials and conference proceedings in April 2015, and included randomised or quasi-randomised trials that reported at least one of our specified outcomes. Results: SFNRT in low- and middle-income countries decreased early neonatal mortality [risk ratio () 0.85 (95% CI )]; the number needed to treat for benefit [227 (95% CI 122 1,667; 3 studies, 66,162 participants, moderate-quality evidence)], and 28-day mortality [ 0.55 (95% CI ); 1 study, 3,355 participants, low-quality evidence]. Decreasing trends were noted for late neonatal mortality [ 0.47 (95% CI )] and perinatal mortality [ 0.94 (95% CI )], but there were no differences in fresh stillbirths [ 1.05 (95% CI )]. Teamwork training with simulation increased the frequency of teamwork behaviour [mean difference (MD) 2.41 (95% CI )] and decreased resuscitation duration [MD (95% CI to 84.34); low-quality evidence, 2 studies, 130 participants]. Conclusions: SFNRT in low- and middle-income countries reduces early neonatal mortality, but its effects on birth asphyxia and neurodevelopmental outcomes remain uncertain. Followup studies suggest normal neurodevelopment in resuscitation survivors S. Karger AG, Basel Introduction The fourth Millennium Development Goals (MDG-4) commits to reducing mortality in children aged younger than 5 years of age [1]. Neonatal deaths account for a significant proportion of deaths in children less than 5 years. It is estimated that 2.76 million neonates died worldwide This paper is based on a Cochrane review first published in The Cochrane Library 2015, Issue 9. DOI: / CD pub2. karger@karger.com S. Karger AG, Basel /16/ $39.50/0 Mohan Pammi, Associate Professor Section of Neonatology, Department of Pediatrics Baylor College of Medicine 6621, Fannin, MC.WT 6-104, Houston, TX (USA) bcm.edu

2 in 2013, and the most common causes were preterm birth complications, birth asphyxia and neonatal sepsis [2]. The vast majority of these neonatal deaths (98%) occurred in the middle- and low-income countries where asphyxia accounted for approximately one quarter of all the deaths [3]. It is estimated that approximately 10% of all newborns require some assistance at birth and less than 1% require extensive resuscitation [4]. However, although early neonatal depression is common, it is difficult to predict prior to birth, and the need for resuscitation may be anticipated in only 50% of cases [5]. Therefore, the universal presence of personnel adequately prepared to perform resuscitation is an important first step in newborn resuscitation. There are significant variations in the presence of skilled personnel at newborn resuscitation. In South Asia and sub-saharan Africa, only about one third of women deliver in the presence of a skilled birth attendant [6]. In the developed world, debate continues around the area of planned home births. A recent meta-analysis has found that there is a tripling of neonatal mortality in planned home deliveries [7]. Some of this increase has been attributed to inadequate training in newborn resuscitation [7]. In the past, there has been great variation in neonatal resuscitation practices, but this has been addressed recently with the introduction of standardised formal neonatal resuscitation training (SFNRT) programmes. Numerous neonatal resuscitation programmes exist, including the NRP (Neonatal Resuscitation Programme), NLS (Neonatal Life Support) and ENLS (European Neonatal Life Support). The NRP was introduced in 1987 in the US, and there are now over 24,000 NRP instructors in the US and over two million people trained in NRP. It is now taught in over 140 countries worldwide. In the middle- and low-income countries, resuscitation programmes often form a part of an overall newborn care intervention package [8], and an SFNRT has been specifically developed as part of this package, Helping Babies Breathe (HBB). Since its launch in 2010, the HBB has been introduced in 77 countries and well over 160,000 birth attendants have been trained and equipped [9]. ILCOR (the International Liaison Committee on Resuscitation) [4] presents scientific statements, addresses consensus on cardiopulmonary resuscitation science statements and prepares treatment recommendations for resuscitation, including the newborn infant. While neonatal resuscitation training programmes may differ both in their content and format, they generally include a theoretical knowledge-based component and a practical, skill-based component. Standardised programmes have a fixed content and course set for the programme, and formal standardised programmes are defined by the intended and focused teaching session as opposed to teaching at the bedside or during resuscitation. The purpose of neonatal resuscitation education is to translate the science of resuscitation into a training programme, allowing transfer of the knowledge and skills of resuscitation into improved clinical practice with the ultimate goal being to reduce neonatal morbidity and mortality. Therefore, every resuscitation education programme should be rigorously evaluated to verify that it is both valid and effective. In addition, substantial health care resources are expended on standardised, formal resuscitation training programmes. Our primary aim was to determine the effectiveness of SFNRT programmes in reducing mortality and morbidity in the newborn infant. Our secondary aim was to determine the effect of SFNRT on changing health care provider behaviour, acquisition of knowledge and skills, and retention of knowledge and skills. Methods Procedures The guidelines from the Cochrane Neonatal Review Group (CNRG; were applied. We followed the PRISMA guidelines for reporting systematic reviews and meta-analyses [10] and the GRADE methodology to evaluate evidence as low-, moderate- and high-quality evidence [11]. We included randomised, cluster-randomised or quasi-randomised trials of SFNRT programmes to train health care professionals caring for newborns at delivery. Definitions We defined SFNRT broadly as programmes which include the essential elements of teaching and testing resuscitation skills by instructors who are certified by a national or international body that oversees resuscitation. Resuscitation programmes may include content in addition to the essential elements (such as behavioural training or boosters), use any instructional format (didactic vs. simulation) and may include single or multiple sessions. Resuscitation programmes may use lecture- or video-based didactics, teaching skills in interactive sessions using mannequins and simulation based on real-life scenarios using low- or high-fidelity mannequins, or be virtual or computer based. Our primary outcome was neonatal mortality, which was defined as death in the first 28 days of life. Subgroup categories for neonatal mortality were early neonatal mortality (death in the first 7 days of life) and late neonatal mortality (death between 8 and 28 days of life). We also derived data for the outcomes of stillbirths (fresh stillbirths) and perinatal deaths (stillbirths + deaths in the first 7 days of life). Stillbirths will be defined as babies born after 6 months of gestation without any movement, spontaneous breathing or heartbeat during or after the delivery [12] and fresh stillbirths as stillbirths with absence of maceration [13]. Secondary outcomes were measures of neonatal morbidity, changes in health care professional and resuscitation team behav- Neonatal Resuscitation Training Programmes and Outcomes 211

3 MEDLINE, EMBASE, CINAHL, Oxfort Database of Perinatal Trials 1960 to April ,526 citations CENTRAL Up to April citations Ongoing trials Up to April citations Conference proceedings 1960 to April citations 792 non-duplicate citations screened Inclusion/exclusion criteria applied 749 articles excluded after title/abstract screen 43 articles retrieved Inclusion/exclusion criteria applied 27 articles excluded after full-text screen 2 articles excluded during data extraction 14 articles included Fig. 1. PRISMA flow diagram detailing study inclusion. Fourteen studies met the inclusion criteria and 8 studies were included in data synthesis by meta-analysis. iour, and acquisition and retention of knowledge and skills. Neonatal morbidity was defined as follows: (i) Hypoxic ischaemic encephalopathy moderate to severe encephalopathy defined using a combination of clinical and biochemical parameters. This is defined as follows: in infants >36 weeks of gestation, either cord or arterial ph <7 or base deficit >16 within the 1st h of life OR if there is no blood gas, or cord/arterial ph and base deficit between 10 and 16, then additional criteria of a history of an acute perinatal event and Apgar scores <5 at 10 min OR need for mechanical ventilation at 10 min of age [14] ; (ii) In view of unavailability of biochemical measures of defining hypoxic ischaemic encephalopathy in low- and middle-income countries, we also used the WHO definition of birth asphyxia, namely failure to initiate or sustain normal breathing at birth, as determined by the birth attendant [15] ; (iii) Low Apgar scores defined as a 10-min Apgar score <5; (iv) Seizure: electroencephalographically confirmed seizure within 72 h of birth; (v) Hypothermia (defined as a temperature <36 C) on admission to the neonatal unit [16] ; (vi) Admission to the neonatal unit in randomised infants including those requiring resuscitation; (vii) Meconium aspiration syndrome defined as respiratory distress in an infant born through meconium-stained amniotic fluid whose symptoms cannot be otherwise explained [17] ; (viii) Neurodevelopmental outcome at months as by a validated tool. We also evaluated change in health care professional and resuscitation team behaviour, and acquisition and retention of knowledge and skills. Our search strategy and databases searched are attached in the online supplementary file (available at www. karger.com/doi/ / ). Data Collection and Analysis Review authors (M.P. and E.M.D.) independently the titles and the abstracts of studies identified by the search strategy for eligibility for inclusion. Two authors separately extracted, and coded all data for each included study. We used the standardised review methods of the CNRG to assess the methodological quality of the studies. We performed statistical analysis according to the recommendations of the CNRG and used the statistical package RevMan 5.3 [18] as recommended by the Cochrane handbook [19]. We used the fixed-effect model for metaanalyses. In cluster-randomised trials (CRTs) that were analysed appropriately at the cluster level using the intra-cluster coefficient (ICC), the summary estimate was used to generate the natural logarithm of the risk ratio () and the standard error of the logarithm of, entered in RevMan and meta-analysed using the generic inverse-variance method. In CRTs that were not analysed at the cluster level, where the ICC was available, we calculated the design effect using the ICC and adjusted the sample size for analyses. If the ICC was not available, we used an assumed ICC from similar trials [ICC from 12 or 20 ] or performed approximate analysis as recommended [19]. When the ICC could not be assumed due to variability in the study design or outcome, we summarised the results without meta-analysis. We heterogeneity of treatment effects between trials using the I 2 statistics to check the appropriateness of pooling data and performing meta-analysis. Meta-analysis was deferred if heterogeneity was high (>75%). 212 Pammi/Dempsey/Ryan/Barrington

4 Table 1. Characteristics of the studies included Study Type of study Participants Interventions Outcomes tes Community-based CRTs (5 studies) Carlo CRT [20], 2007 Carlo [24], 2010 Gill [12], 2011 Xu [26], 2014 Opiyo [25], 2008 CRT CRT ICC was not available and hence ICC of from the studies by Carlo et al. [20, 24] was used in the metaanalyses Communitybased CRT Randomised controlled trial Traditional birth attendants, nurses, midwives and physicians; birth attendants from 6 countries (Argentina, Democratic Republic of Congo, Guatemala, India, Pakistan, Zambia) Traditional birth attendants, nurses, midwives, and physicians; birth attendants from 6 countries (Argentina, Democratic Republic of Congo, Guatemala, India, Pakistan, Zambia) Traditional birth attendants in the Lufwanyama district of Zambia Intervention group 11 intervention hospitals (97 health care providers) Control group 11 control hospitals (87 health care providers) Health care workers randomised to early neonatal resuscitation training (n = 28) and late training (control, n = 55) A modified version of the NRP was provided in a cluster-randomised design (after Early Newborn Care) A modified version of NRP was provided in a cluster-randomised design (after Early Newborn Care) Modified version of NRP; all were trained in immediately drying the infant, suctioning the mouth and nose, optimally positioning the airways; the intervention arm was trained to provide positive pressure ventilation with a reusable resuscitator mask and to recognise infection, administer oral amoxicillin and transfer to a hospital setting Neonatal programme based on ILCOR guidelines including the initial steps of resuscitation, physiol ogy, positive pressure ventilation, intubation, cardiac compression and medication administration Neonatal training programme adapted from the programme recommended by the UK Resuscitation Council Primary outcome Early neonatal mortality Secondary outcomes Death attributable to birth asphyxia; stillbirth rates; perinatal death rate; 24-hour mortality; Apgar scores; resuscitation techniques; neurological outcome at 7 days Primary outcome Early neonatal mortality Secondary outcomes Death attributable to birth asphyxia; stillbirth rates; perinatal death rate; 24-hour mortality; Apgar scores; resuscitation techniques; neurological outcome at 7 days Primary outcome Neonatal mortality at 28 days Secondary outcomes Stillbirths; early neonatal mortality; late neonatal mortality from day 8 to 28 Incidence of birth asphyxia (defined by Apgar scores 7), asphyxia-related deaths in the delivery room and health care providers knowledge and self-confidence by questionnaires Primary outcome Proportion of appropriate resuscitation steps Secondary outcomes Frequency of inappropriate practices and delivery room mortality Large NICHD-sponsored study by the Global Network for Women s and Children s research (First Breath Study Group) in neonates born 1,500 g ICC (0.002) was requested and obtained from the author Report of infants with very low birth weight ICC (0.002) was requested and obtained from the author The definition of asphyxia is different from our a priori definition and there are no data on neonatal mortality in the first 28 days of life Setting Public hospital in Kenya; data were collected on 97 resuscitation episodes in the intervention group and 115 resuscitation episodes in the control group n-community randomised studies addressing knowledge, skills, teamwork, boosters and decision support tools (9 studies) Dunn Neonatal nurses [29], 1992 Kaczorowski [33], 1998 Thomas [27], 2007 Bould [34], 2009 CRT, where hospitals were randomised to NRP or no intervention Randomised trial of 2 booster strategies designed to improve retention of skills and knowledge in neonatal resuscitation Randomised controlled trial Randomised controlled trial Family practice residents from a single institution Interns (paediatric, combined paediatric and internal medicine, family medicine and obstetrics and gynaecology) Anaesthesia trainees Neonatal resuscitation training with didactics and demonstration of resuscitation skills Boosters of NRP video, hands on and no boosters 3 5 months following original NRP training; the hands on group received a 2-hour supervised hands on mannequin practice Standard NRP compared to NRP with team training; the team training consisted of a 2.5-hour session comprised of lectures, lowfidelity simulation, short video clips and a question/answer period; they also received prompting during the NRP skill station to engage in team behaviours Trainees were randomised to the use of a cognitive aid or no aid during a resuscitation scenario Assessment of knowledge, resuscitation performance and measurement of subjects perception of their competence NRP written examination and observer assessment by checklists 6 8 months following the original NRP course Teamwork behaviour consisting of inquiry, information sharing, assertion, evaluation of plans, workload management and vigilance Primary outcome Technical score of resuscitation using a validated checklist Secondary outcome Anaesthetist s n-technical Skills of behavioural scores First randomised controlled trial on NRP intervention; although cluster randomised, analyses were done at the individual level; not meta-analysed in this review as ICC was not known Data acquired from the author for meta-analyses Conducted in St. Michael s Hospital, University of Toronto Neonatal Resuscitation Training Programmes and Outcomes 213

5 Table 1 (continued) Study Type of study Participants Interventions Outcomes tes Thomas [28], 2010 Lee [30], 2012 Bender [32], 2014 Rubio-Gurung [31], 2014 Fuerch [35], 2015 Randomised trial Randomized trial Randomised controlled study CRT Randomised controlled study Interns admitted during the years 2007 and 2008 and not NRP certified 2nd-, 3rd- and 4th-year emergency medicine residents Residents from the neonatal unit and non-neonatal programmes 12 maternity hospitals (cluster); in each hospital a random sample of 10 professionals were recruited Participants included 18 residents, 1 fellow, 7 attending physicians, 2 respiratory therapists and 37 nurses Intervention group n = 35 Control group n = 30 2-hour teamwork training intervention with low-/high-fidelity NRP; control subjects practiced low-fidelity NRP without team training Intervention group 4-hour simulation-based; education al intervention beginning with a 45-min lecture on NRP algorithm and high-fidelity simulation scenarios Control group Standard emergency medicine curriculum with once a month conference and at least paediatric scenario (including a neonatal scenario) per session Simulation booster 7 10 months after NRP (half-day session) compared to no booster 4-hour simulation training session delivered in situ Participants were randomised to an electronic decision support tool that gives visual and auditory prompts during a resuscitation scenario Teamwork behaviour Teamwork event rate, workload management and vigilance NRP performance Performance scores and duration of resuscitation Primary outcome Initial and after intervention differences in resuscitation scores and confidence between groups Secondary outcome Number of critical resuscitation actions performed and time required to complete critical actions Outcomes by a validated neonatal resuscitation scoring tool Outcomes months after taking an NRP course Knowledge (% of correct answers), procedural skills (sum of 107 elements) and teamwork behaviour (sum of 25 elements) were Outcomes evaluated by 2 standardised resuscitation scenarios immediately after the intervention and 3 months later Outcomes Technical score of resuscitation performance, frequency of achieving a heart rate of >90 b.p.m. and number of hazardous events Outcomes Percentage of time a participant appropriately decided on performing positive pressure ventilation or cardiac compressions, and frequency of FiO2 adjustment Data acquired from the author for meta-analyses 75% of eligible residents participated in the study; none were NRP certified; intra-class correlation 0.64; single-institution study at Rhode Island Hospital Single-centre study at Rhode Island Hospital, Providence, USA In situ simulation study in France in the hospitals of the AURORE perinatal network using mannequins Study performed at CAPE (Center for Advanced Pediatric and Perinatal Education) at the Lucille Packard Children s Hospital at Stanford University, USA Results The inclusion process from the search strategy to inclusion of articles into our review is outlined in the PRIS- MA flow diagram ( fig. 1 ). Details on the 14 included studies are reported in table 1 and the 29 excluded studies with reasons for exclusion in table 2. Two studies presented as abstracts (Campbell and Finan [21] and Yamada et al. [22] ) and 1 ongoing study (Bang et al. [23] ) are awaiting classification as no data were available and thus not included. Methodological assessment of the studies included was performed along established guidelines [19]. Risk of bias in included studies is reported in table 3. Effects of Neonatal Resuscitation Training on Neonatal Mortality, Stillbirth and Neonatal Complications In community-based CRTs, where birth attendants were randomised to SFNRT in addition to basic resuscitation training, the estimated showed a reduction in early neonatal mortality: 0.88 (95% CI ). Using the approximate analysis method [19], the estimated was 0.85 (95% CI ) and the risk difference was (95% CI to ), which implies a number needed to treat for benefit of 227 (95% CI 122 1,667; 3 CRTs, 66,162 participants, moderate-quality evidence; fig. 2 a). For every 227 deliveries occurring in a setting where health care workers have been trained in SFNRT, there is 1 fewer neonatal death. We downgraded the quality of evidence from high to moderate quality as the participants were not blinded to the intervention and also due to inconsistency in the direction of the effects across studies (heterogeneity, I 2 = 71%). Studies by Carlo et al. [20, 24] did not report a decrease in neonatal mortality, whereas Gill et al. [12], who included only traditional birth attendants, showed a significant decrease in early neonatal mortality in the intervention group. In subgroup analy- 214 Pammi/Dempsey/Ryan/Barrington

6 Table 2. Studies excluded First author, year Reasons for exclusion n-randomised studies evaluating neonatal outcomes (10 studies) 1 Duran [50], 2008 n-randomised study which the impact of NRP in Turkey before and after NRP training 2 Patel [51], 2001 n-randomised study and evaluation before and after introduction of NRP 3 Patel [52], 2002 n-randomised study but a retrospective evaluation of Apgar scores before and after the introduction of NRP 4 Deorari [53], 2001 n-randomised study and impact of NRP before and after intervention in 14 teaching hospitals in India 5 Chomba [54], 2008 n-randomised study and impact of NRP before and after intervention 6 Boo [55], 2009 n-randomised study but impact of NRP before and after NRP in Malaysia 7 Ashish [56], 2012 n-randomised study and impact of HBB before and after intervention in Nepal 8 Carlo [13], 2010 n-randomised study and has a pre-post intervention study design 9 Msemo [57], 2013 n-randomised trial and impact of HBB before and after intervention 10 Goudar [36], 2013 n-randomised trial and impact of HBB before and after intervention n-randomised studies evaluating knowledge and skills (13 studies) 11 Duran [58], 2008 n-randomised study and retention of knowledge and skills before and after intervention 12 Durojaiye [59], 2002 n-randomised study and study of knowledge of paediatric trainees before and after intervention 13 Ergenekon [60], 2000 n-randomised study and study of trainees before and after intervention 14 Ersdal [61], 2013 n-randomised trial and a 7-month follow-up of practical skill retention following a HBB course 15 Lopez-Herce [62], 1999 n-randomised study and a study of participants evaluation of the course after intervention 16 Nadler [63], 2011 n-randomised study but video debriefings on teamwork in newborn resuscitation 17 Singhal [64], 2001 n-randomised study and an assessment of a telephone questionnaire on newborn resuscitation 18 Skidmore [65], 2001 n-randomised study and study of knowledge and skills before and after intervention 19 Trevisanuto [66], 2007 n-randomised study and evaluation of NRP before and after intervention and comparing cities in Italy and Sudan 20 Trevisanuto [67], 2005 n-randomised study but evaluation of NRP before and after intervention in paediatric residents 21 Tan [68], 2014 n-randomised study evaluating quality of chest compression 22 Walker [69], 2014 n-randomised study of a simulation-based obstetric and neonatal emergency training programme 23 Mathai [70], 2015 n-randomised mannequin study on training in the self-inflating bag and the T-piece resuscitator NRP evaluated in trainees who are not health care providers (4 studies) 24 Cavaleiro [71], 2009 Participants are medical students not health care providers 25 Curran [72], 2004 Participants are medical students not health care providers 26 Hubballi [73], 2014 Participants are auxiliary nursing midwife students 27 Deindl [74], 2015 Participants are medical students not health care providers Other excluded studies (2 studies) 28 Senarath [75], 2007 Community-based trial but did not report any of our pre-specified outcomes 29 Finan [40], 2012 A comparison of high- vs. low-fidelity SFNRT and no specified outcomes are reported ses, training of traditional birth attendants in neonatal resuscitation decreased early neonatal mortality: 0.79 (95% CI ; 3 CRTs, 27,673 participants, moderate-quality evidence; fig. 2 b). One CRT [12] reported no differences in late neonatal mortality in 3,274 neonates [ 0.47 (95% CI ); 1 study, low-quality evidence] but a decreased neonatal mortality at 28 days in 3,355 neonates [ 0.55 (95% CI ); 1 study, lowquality evidence]. Studies by Opiyo et al. [25] and Xu et al. [26] reported on mortality or morbidity but not as defined in our protocol, and hence they were not included in the meta-analyses. Resuscitation behaviour related to the use of bag mask ventilation by health care providers was reported in 3 CRTs that enrolled 29,664 neonates, but heterogeneity was high (I 2 = 95%) and meta-analysis was deferred. The quality of evidence was downgraded to low as the participants were not blinded to the intervention and due to the inconsistent effects across studies. A significant increase in the use of bag mask ventilation was reported by Carlo et al. [20] in 2010 [ 1.18 (95% CI )] and Gill et al. [12] in 2011 [ (95% CI )], but was not significant in another study by Carlo et al. [24] in 2010 [ 0.85 (95% CI )]. We also performed a meta-analysis on stillbirths and perinatal deaths by the approximate analysis method. Carlo et al. [13] did not include a definition of stillbirths in their study although they define fresh stillbirth as absence of maceration. Gill et al. [12] defined stillbirth, but data on fresh stillbirths were not separately available. There was no significant difference in the rate of fresh stillbirths when SFNRT was compared with early newborn care, 1.05 (95% CI ; fig. 3 ) or in the rate of all stillbirths [ 1.04 (95% CI ); 3 CRTs, Neonatal Resuscitation Training Programmes and Outcomes 215

7 Table 3. Risk of bias Risk of bias for CRTs Carlo et al. [20], 2010 Carlo et al. [24], 2010 Gill et al. [12], 2011 Dunn et al. [29], 1992 Xu et al. [26], 2014 Rubio-Gurung et al. [31], 2014 Recruitment bias (bias due to knowledge of the type of cluster before participant recruitment to the cluster) Baseline imbalance Loss of clusters or individuals from the cluster Whether analysed taking clustering into account (unit of analysis error) Contamination between clusters Yes Yes Yes ne none Risk of bias for randomised controlled trials Bias Kaczorowski et al. [33], 1998 Thomas et al. [27], 2007 Thomas et al. [28], 2010 Lee et al. [30], 2012 Bender et al. [32], 2014 Opiyo et al. [25], 2008 Bould et al. [34], 2009 Fuerch et al. [35], 2015 Random sequence generation (selection bias) Allocation concealment (selection bias) Blinding of participants and personnel (performance bias) Blinding of outcome assessment (detection bias) Incomplete outcome data (attrition bias) Selective reporting (reporting bias) Other bias t reported t clearly stated Blinding not possible Random number generator t clear Control and intervention groups on separate floors 5 NRP instructors Video recordings who were blinded to reviewed by independent assessors participants group membership (low risk) 75% completed the study t noted t noted 7 recordings were excluded due to technical issues; 80% followup rate t noted t noted Random number generator t clear Blinding is not possible Video recordings reviewed by assessors unaware of status 90.7% follow up at 6 months t noted t noted Computer generated t clear Blinding is not possible Computer generated t clear (unclear risk) Blinding is not possible Video recordings reviewed (unclear t clear by assessors risk) unaware of status All enrolled were analysed t noted t noted 13.8% attrition (moderate risk) t noted t noted t reported t clear Blinding is not possible Yes Only 83 of 90 (92%) enrolled participants were observed (moderate risk) t clear t noted t reported t reported Transparency t reported of sealed envelope not clear Blinding is not possible Yes ne noted t noted t noted Blinding is not possible t clear ne noted ne noted ne noted 62,366 births, moderate-quality evidence]. Perinatal mortality showed a decreasing trend with borderline statistical significance [ 0.94 (95% CI ); 3 CRTs, 62,152 live births, moderate-quality evidence; fig. 4 ]. ne of the trials reported the following relevant outcomes related to neonatal morbidity: neonatal outcomes of hypoxic ischaemic encephalopathy, low Apgar scores (<5) at 10 min, seizure, hypothermia, admission to the neonatal unit, meconium aspiration syndrome, longterm neurodevelopmental outcome at 2 years or mortality during initial hospitalisation. Teamwork Behaviour and Resuscitation Duration after Teamwork Training Two randomised trials by Thomas et al. [27, 28] reported teamwork behaviour in 130 participants after supplementing SFNRT with teamwork training. Teamwork behaviours, information sharing, inquiry, assertion, teaching and advising, managing workload and vigilance were by observations of the frequency or duration. Inquiry, information sharing, assertion, teaching and advising were measured as rates (behaviours/min), whereas workload management and vigilance were measured as a percentage of simulation time. 216 Pammi/Dempsey/Ryan/Barrington

8 First author of the study SFNRT Control Weight, even ts total events total % M-H, fixed (95% CI) M-H, fixed, 95% CI Carlo [20], , , ( ) Carlo [24], ( ) Gill [12], , , ( ) Total (95% CI) 14,052 14, ( ) Total events Heterogeneity: χ 2 = 3.48, d.f. = 2 (p = 0.18), I 2 = 43% Test for overall effect: Z = 2.55 (p = 0.01) a Favours SFNRT Favours control First author of the study log () SE SFNRT total Control total Weight, % IV, fixed (95% CI) IV, fixed, 95% CI b Carlo [20], ,770 13, ( ) Carlo [24], ( ) Gill [12], ,923 1, ( ) Total (95% CI) 12,752 14, ( ) Heterogeneity: χ 2 = 2.49, d.f. = 2 (p = 0.29), I 2 = 20% Test for overall effect: Z = 2.44 (p = 0.01) Favours SFNRT Favours control Fig. 2. a Forest plot of studies reporting s with 95% CIs for early neonatal mortality. Derived from approximate analyses method where the sample size is reduced by the design effect (dependent on ICC and mean size of the cluster) and using the Mantel- Haenszel (M-H) method and a fixed-effect model for meta-analyses. RevMan 5.3 was employed for the analyses and generation of forest plots. b Forest plots of subgroup analysis of traditional birth attendants who were trained and attended deliveries, reporting with 95% CIs for early neonatal mortality. The generic inversevariance (IV) method was used with a fixed-effect model. RevMan 5.3 was used for the analyses and generation of forest plots. First author of the study SFNRT Control Weight, even ts total events total % M-H, fixed (95% CI) M-H, fixed, 95% CI Carlo [20], , , ( ) Carlo [24], ( ) Gill [12], , , ( ) Total (95% CI) 14,417 15, ( ) Total events Heterogeneity: χ 2 = 0.19, d.f. = 2 (p = 0.91), I 2 = 0% Test for overall effect: Z = 0.78 (p = 0.44) Favours SFNRT Favours control Fig. 3. Forest plots of studies reporting s with 95% CIs for stillbirths. Derived from approximate analyses method where the sample size is reduced by the design effect (dependent on ICC and mean size of the cluster) and using the Mantel-Haenszel (M-H) method and a fixed-effects model for meta-analyses. RevMan 5.3 was used for the analyses and generation of forest plots. Neonatal Resuscitation Training Programmes and Outcomes 217

9 First author of the study SFNRT Control Weight, even ts total events total % M-H, fixed (95% CI) M-H, fixed, 95% CI Carlo [20], ,011 12,235 1,176 13, ( ) Carlo [24], ( ) Gill [12], , , ( ) Total (95% CI) 14,335 15, ( ) Total events 1,282 1,475 Heterogeneity: χ 2 = 1.53, d.f. = 2 (p = 0.46), I 2 = 0% Test for overall effect: Z = 1.94 (p = 0.05) Favours SFNRT Favours control Fig. 4. Forest plots of studies reporting s with 95% CIs for perinatal deaths. Derived from approximate analyses method where the sample size is reduced by the design effect (dependent on ICC and mean size of the cluster) and using the Mantel-Haenszel (M-H) method and a fixed-effect model for meta-analyses. RevMan 5.3 was used for the analyses and generation of forest plots. Teamwork training increased the frequency of any teamwork behaviour (reported as behaviours/min): mean difference (MD) 2.41 (95% CI ; fig. 5 ). Heterogeneity was low (I 2 = 0%). Components of teamwork behaviour that increased after teamwork training were (MD) information sharing [0.84 behaviours/min (95% CI )], inquiry [0.29 behaviours/min (95% CI )] and managing workload [9.93% of simulation time (95% CI )]. Teamwork behaviour related to teaching or advising [0.08 behaviours/min (95% CI 0.01 to 0.16)], assertion [0.68 behaviours/min (95% CI )] and vigilance [0.20% of simulation time (95% CI 0.13 to 0.53)] did not significantly increase after teamwork training. We downgraded the quality of evidence to low as evidence was from only 2 studies from a single institution, unclear allocation concealment and imprecision of the estimate with wide CIs. Two randomised trials [27, 28] reported on the duration to complete resuscitation in 130 participants after teamwork training with SFNRT, and the estimated MD showed a decrease in resuscitation duration: MD s (95% CI to 84.34; fig. 6 ). We downgraded the quality of evidence to low as evidence was from only 1 study from a single institution, unclear allocation concealment and imprecision of the estimate with wide CIs. One randomised trial [28] reported on NRP performance scores in 98 participants after teamwork training with SFNRT. The estimated MD in the NRP scores between the intervention and the control groups was 1.40 (95% CI 2.02 to 4.82). We downgraded the quality of evidence to low as evidence was from only 1 study from a single institution, unclear allocation concealment and imprecision of the estimate with wide CIs. Resuscitation Knowledge, Skills and Resuscitation Performance Scores In the CRT by Dunn et al. [29], which included 166 participants, the proportion of participants who scored more than 80% on an evaluation of their knowledge of neonatal resuscitation by means of a multiple-choice examination increased after training from 36 to 91%, compared to 15% in the controls. ne of the subjects scored 100% on the skills before the test (using structured evaluation of a mock code involving a mannequin), but 100% of the intervention group did so after the test. Eighty-five percent of the intervention group and 23% of controls passed the knowledge multiple-choice test with a score of 80% or more. ne of the participants in either group passed the skill test (mock code) 6 months after the SFNRT. In the other CRT by Xu et al. [26], the mean knowledge acquisition scores (SD) were 9.2 (1.2) in the intervention group and 8.4 (1.5) in the control group (p < ). Both CRTs were incorrectly analysed at the level of the individual and hence not included in the meta-analysis. The quality of evidence was rated very low because of unit of analysis error and evidence available from only 2 studies. In randomised studies, Lee et al. [30] and Rubio-Gurung et al. [31] reported on resuscitation scores and team scores. We are awaiting more data to be considered for inclusion in the meta-analyses. Two trials [32, 33] reported the effect of NRP boosters on retention of knowledge and skills. One trial [17] evaluated hands on and video boosters on knowledge and skill retention in 187 participants and found no differences in knowledge retention after hands on booster [MD 7.00 (95% CI 2.87 to 16.87)], video booster [MD 4.00 (95% 218 Pammi/Dempsey/Ryan/Barrington

10 First author of the study SFNRT with team SFNRT training mean SD total mea n SD total Weight % MD IV, fixed (95% CI) MD IV, fixed, 95% CI Any teamwork behaviour Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 0.10, d.f. = 1 (p = 0.75), I 2 = 0% Test for overall effect: Z = 6.79 (p < ) Information sharing Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 5.13, d.f. = 1 (p = 0.02), I 2 = 80% Test for overall effect: Z = 5.64 (p < ) Inquiry Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 4.65, d.f. = 1 (p = 0.03), I 2 = 78% Test for overall effect: Z = 3.99 (p < ) Assertion Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 5.18, d.f. = 1 (p = 0.02), I 2 = 81% Test for overall effect: Z = 3.82 (p = ) Teaching or advising Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 1.25, d.f. = 1 (p = 0.26), I 2 = 20% Test for overall effect: Z = 1.76 (p = 0.08) Managing workload Thomas [27], ( ) Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: χ 2 = 0.29, d.f. = 1 (p = 0.59), I 2 = 0% Test for overall effect: Z = 5.13 (p < ) Vigilance Thomas [27], t estimable Thomas [28], ( ) Subtotal (95% CI) ( ) Heterogeneity: t applicable Favours SFNRT Favours SFNRT + training Fig. 5. Forest plot of studies reporting MD with 95% CIs for teamwork behaviour. Subgroups of teamwork behaviour with versus without teamwork training were compared. The inverse-variance (IV) method was used with a fixed-effect model. RevMan 5.3 was used for the analyses and generation of forest plots. Neonatal Resuscitation Training Programmes and Outcomes 219

11 First author of the study SFNRT with team training SFNRT Weight, % MD IV, fixed (95% CI) MD IV, fixed, 95% CI mean SD total mea n SD total Thomas [27], ( to 51.48) Thomas [28], ( to 63.12) Total (95% CI) ( to 84.34) Heterogeneity: χ 2 = 0.03, d.f. = 1 (p = 0.87), I 2 = 0% Test for overall effect: Z = 4.50 (p < ) Favours SFNRT Favours SFNRT + training Fig. 6. Forest plot of studies reporting MD with 95% CIs for resuscitation duration (in s) with or without teamwork training. The inverse-variance (IV) method was used with a fixed-effect model. RevMan 5.3 was used for the analyses and generation of forest plots. CI 9.72 to 17.72)] or any booster [MD 5.50 (95% CI 4.37 to 15.37)]. There were no significant differences in skill retention after hands on booster [MD 5.00 (95% CI 3.18 to 13.18)], video booster [MD 6.00 (95% CI )] or any booster [MD 5.48 (95% CI 1.07 to 12.03)]. Another trial [19] reported on the knowledge and skill retention after a simulation booster 7 10 months after NRP and months after NRP. The study found increases in procedural skills (scores of 18.8 vs. 16.2, p = 0.02) and behaviour (scores of 71.6 vs. 68.1, p = 0.02) but not in knowledge scores (scores of 71.6 vs. 68.1, p = 0.57). We downgraded the quality of evidence for boosters to very low as evidence was from only 2 studies and also the high risk of selection, performance and attrition biases. Two trials reported on decision support tools during resuscitation [34, 35]. In a randomised study of a cognitive aid (poster) during a resuscitation scenario, Bould et al. [34] reported no differences in resuscitation technical scores [median 20.3, interquartile range (IQR) , range with intervention vs. median 18.2, IQR , range with control] or scores of Anaesthetists n-technical Skills (median 10.2, IQR , range with intervention vs. median 9.3, IQR , range with control). Fuerch et al. [35] randomised 65 participants to an electronic decision support tool with prompts and reported that positive pressure ventilation was performed correctly more frequently (94 95% with intervention vs % with control, p < ) as well as cardiac compressions (82 93% with intervention vs % with control, p < ); FiO 2 was adjusted three times more frequently in the intervention group compared to the control group. Discussion Meta-analyses of data from 3 community-based CRTs suggest that SFNRT results in a reduction in early neonatal mortality in the low- and middle-income countries, where birth attendants had previously received training in basic newborn care. These well-designed, large, communitybased studies provide moderate-quality evidence supporting the efficacy of newborn resuscitation training programmes in improving neonatal outcomes. The greatest reduction in mortality was seen in the community where births were attended by SFNRT-trained traditional birth attendants. All three studies are from low- and middleincome countries, where baseline neonatal mortality rates are high, and so these findings cannot be directly extrapolated to high-income countries, where neonatal mortality is already much lower, and where resuscitation training programmes are considered the standard of care and required for accreditation of training in newborn care. The classification of a birth as a live birth or a stillbirth can be affected by training in neonatal care programmes, which can influence the denominator in the calculation of mortality rates in the perinatal period. In 2010, Carlo et al. [20] showed a marked reduction in stillbirth rates, as babies previously thought to be stillborn were then recognised to be depressed but to have a potential for survival after the introduction of the Early Newborn Care programme. Recently, the HBB programme has also reported a similar occurrence [36]. Our meta-analysis of the 3 CRTs of SFNRT in this review did not show differences in fresh stillbirths or all stillbirths (both fresh and macerated) after SFNRT compared to the Early Newborn 220 Pammi/Dempsey/Ryan/Barrington

12 Care programme. Absence of differences in stillbirths may be because the universal introduction of the Early Newborn Care programme had already achieved appropriate designation of stillbirths. We report in this review that for every 227 births attended by personnel trained in NRP there was 1 neonatal death less an important finding. If 1 trained attendant attends 1,000 births per year, then there will be 4 deaths prevented, resulting in a significant reduction in neonatal mortality. This figure translates into 80 fewer deaths over 20 years: 1 training course, 8 refresher courses resulting in 80 fewer deaths (32 h spent on mandatory courses, which translates into almost 3 deaths prevented per hour spent on the course). This is an astonishing figure, and in those parts of the world with the highest perinatal mortality, such as sub-saharan Africa [37], major impacts on neonatal mortality are possible. We estimate that if every birth was attended by personnel trained in SFNRT, then 140,000 lives would be saved annually. As there were no data analysable on neonatal morbidities, we cannot be certain from this review that there was no increase in morbidity associated with the increase in survival. If a substantial number of survivors had morbidities, especially permanent morbidities, then the burden imposed on low-income countries by introducing SFNRT could be large. However, recently reported follow-up cohort studies suggest normal neurodevelopment in resuscitation survivors. Carlo et al. [38] have reported that resuscitated children from a low-income country did not have an excess of neurologic or developmental morbidity to 12 months of age and a second observational study has come to the same conclusion; the large majority of infants who required resuscitation at birth had normal development up to 36 months of age [39]. In view of the clear benefit, lack of harm and good long-term outcomes of resuscitated infants, it may not be reasonable to continue to perform other studies of SFNRT compared to no training. The efficacy and cost benefits of SFNRT may vary by setting, health care provider and by whether the health care providers are already trained in basic newborn care, and therefore future studies may be needed to address these issues. In contrast, different methods of training providers, and different interventions and equipment are certainly worth comparing in prospective randomised trials. Assessment of methods to reduce cost and increase efficacy of training programmes will be important to allow wider dissemination of SFNRT in low- and middle-income countries. Despite the importance of SFNRT, there is a paucity of high-quality evidence regarding methods of teaching the programmes. We could find no evidence from randomised trials that simulation-based training was preferable to non-simulation-based SFNRT, and the evidence from randomised trials suggests that high-fidelity simulation was no more effective than low-fidelity simulation [40]. This is important considering the financial costs of highfidelity training. Retention of knowledge and skills is poor, and there are little reliable data about how to improve retention. Further studies of effective ways of teaching and maintaining skills in neonatal resuscitation are needed in order to further increase the impact of these programmes. The precise composition of the programmes has also not been well studied. New programmes, such as HBB ( with a reduced requirement for very advanced skills, seem effective from before/after studies and are tailored to the environment in which many deliveries occur. However, the individual components of such programmes warrant investigation, since training different ways to administer positive pressure breaths, as one example, could have a big influence on the effectiveness of a programme. Resuscitation devices are central to neonatal resuscitation in low- and middle-income countries, and have been named as 1 of 4 commodities for saving newborn lives in a recent UN report. The design, durability, availability and costs of current resuscitation devices pose significant barriers to the dissemination of newborn resuscitation. The bag mask devices, currently available, consist of many parts, making them difficult to disassemble and clean. Consideration should be given to alternative bag masks, such as the Laerdal Global Health upright bag mask, which was designed to be easier to use, easier to clean due to fewer parts and cheaper than standard devices [41]. The upright stance of the device, in addition to a new mask with a thicker and broader top surface, was designed to make it easier to hold the device correctly and enable a better mask seal. In a recent mannequin-based randomised controlled trial, expiratory volumes were higher, mask leakage lower and mean airway pressure slightly higher with upright bag mask versus a standard resuscitator [42]. An alternative airway device, the uncuffed, gel laryngeal mask airway, is currently being investigated in comparison to the bag mask in low-resource settings in a randomised controlled trial [43]. The training mannequins also need further enhancement and innovation, but still remain low-cost simulators with adequate fidelity. Our experience in Sudan led us to examine the NeoNatalie mannequin in 3 modalities, showing that the mannequin filled with water or half water/air were the best configurations as by user preference ( most realistic ) and Neonatal Resuscitation Training Programmes and Outcomes 221