A working group that managed the agenda–setting process consisted of staff responsible for newborn health in WHO and Saving Newborn Lives. The group defined the scope of the priority setting exercise ( Box 1 ). Methodology developed by the Child Health and Nutrition Research Initiative (CHNRI) was adapted and used for this priority setting exercise, to enable systematic listing and transparent scoring of many competing research questions [24–26]. This methodology had been used in the previous priority setting exercises by the WHO on five major causes of child deaths: pneumonia, diarrhea, preterm birth and low birth weight, neonatal infections, and birth asphyxia [8–12]. The previous exercise coordinated by the WHO was sharply focused on short–term gains, ie, within the MDG4 target of the year 2015. In addition, the CHNRI methodology has been used by many other subject groups and multiple organizations [27–33]. Box 2 shows the steps we followed during this priority setting process.

A large group of researchers and program experts were identified and asked to submit three ideas for improving newborn health outcomes by 2025 ( Box 2 ). Two hundred of the most productive researchers, representing a broad range of technical expertise and regional diversity, identified through Web of Science® ranking tools, were invited by email to propose research questions on newborn health and birth outcomes. A further 400 program experts in newborn health programmes were also invited to propose research questions.

The proposed research questions and scoring criteria were refined by a small group of 14 experts who were invited by the WHO to participate in a two–day workshop. Each question was assigned to a domain and a technical area. The first of the three domains was “discovery”, which included research aimed at finding new solutions such as new medicines, vaccines or other preventive interventions, or new diagnostics. The second domain was “development”, which included research questions aimed at improving existing interventions, reducing their costs or making them simpler to deliver. The third domain was “delivery”, which included research questions that would help deliver existing interventions to more mothers and newborns with high quality. The five separate technical areas included: (i) preterm birth; (ii) intrapartum–related events including intrapartum stillbirths; (iii) newborn infections; (iv) congenital malformations and other specific conditions; and (v) integrated care including the care for mothers and neonates;

The final list of research questions and scoring criteria were sent to the original group of 600 experts with an invitation to score them. Each research question was assessed by the expert and received a score of 1.0, 0.5 or 0 for five preset criteria, with the option of not assigning any score in case the expert did not feel confident to decide on that criterion. Scoring took place over eight weeks and was conducted and returned to the coordinators at the WHO by 91 experts.

Intermediate scores for each research question against the 5 criteria were computed as the sum of the scores for that particular criterion divided by the total number of scorers. This resulted in a number between 0–100% that captured the “collective optimism” of the group of 91 scorers that a given research question would fulfill each given criterion. The overall research priority score (RPS) for each research question was then computed as the mean of the intermediate scores calculated for each of the five criteria: RPS = [(Criterion 1 score %)+(Criterion 2 score %)+(Criterion 3 score %)+(Criterion 4 score %)+(Criterion 5 score %)]/5. The confidence interval was calculated using the bootstrapping methods in STATA version 11.2.

In total, 132 of the 600 invited experts proposed a total of 396 research questions, which were then checked for similarity and consolidated in a final list of 205 questions to be scored. The characteristics of respondents are summarized in Figure 1 . The 205 research questions were then scored by 91 experts. About 40% of the scorers were based in low and middle income countries (LMICs) in Africa, Asia, and South America. About two–thirds (65%) worked in academic or research institutions and the remainder was divided between program managers (16%), clinicians (7%), donor representatives (7%) and policy makers (5%) ( Figure 1 ).

The overall research priority scores given to the 205 proposed questions ranged from 90% (high) to 47% (low; full list of scored questions is presented in the Online Supplementary Document(Online Supplementary Document) ). The level of agreement between the 91 experts ranged from 77% (high) to 34% (low), suggesting that on average, for each research question of interest, between three–quarters and one–third of the scorers were in agreement in their responses to each criterion.

The overall scores for the highest priority questions ranged from 79% to 90% ( Table 1 ). Agreement scores indicated that more than two thirds of the experts had a common view towards the list of research priorities. Nine of the ten top priorities were in the domain of “delivery”, with simplified neonatal resuscitation programs and clinical algorithms and improved skills of community health workers leading the list. Among the 11 priorities shown in this table, three addressed preterm birth, four addressed intrapartum–related events and four addressed newborn infections.

In the domain of “development”, the top 10 priorities ( Table 2 ) were ranked between 8^th and 50^th on the list of all research questions (displayed in full in Online Supplementary Document(Online Supplementary Document) ). They were led by ideas on improved Kangaroo Mother Care, improve accuracy of diagnosis by community health workers, and perinatal audits. Two priorities among the leading ten in this domain were identified in each of the areas of preterm birth, intrapartum related events and newborn infections, while the remaining 4 priorities related to integrated care.

The 10 leading priorities for discovery research ( Table 3 ) ranked between 58^th and 129^th on the list of all research questions (see Online Supplementary Document(Online Supplementary Document) ) and they focused on stable surfactant with novel modes of administration, ability to diagnose fetal distress and novel tocolytic agents. Agreement scores for the ten leading questions ranged from 42% to 49%. Three priorities were identified in each of the areas of preterm birth and newborn infections, two on preventing intrauterine growth restriction and one each on intrapartum–related events and antepartum stillbirths.

There was a remarkable similarity in the scoring pattern between experts from a research background and those from a program background for the top 10 ranked priorities ( Table 4 ). The programme experts had a tendency to assign somewhat higher overall scores to “delivery” questions, which was mediated through their higher scoring of maximum potential impact and equity criteria. Among “development” questions, the scorers with a background in research gave higher scores for efficacy and deliverability, while programme experts gave higher scores for impact and equity criteria. Surprisingly, the scoring pattern of both groups of experts for “discovery” questions was very similar, both for overall score and for each of the 5 criteria.

In this paper, we present global research priorities that have the potential to impact mortality, morbidity, child development, and long–term health outcomes among neonates in the period between 2013–2025. Despite the broad focus on these outcomes and a 12–year timeline, “delivery” questions received highest scores, followed by “development” and “discovery” questions, as was the case in previous exercises with shorter time lines focusing only on reducing mortality [8–12].

The major emerging themes in the domain of “delivery” included simplifying intervention delivery to implementation at lower levels of the health system, evaluating delivery of interventions by community health workers, developing strategies to improve quality of care during labour and childbirth, and addressing barriers in the scaling up of high impact interventions. It is interesting to note that 5 of the questions were related to neonatal resuscitation. This could be related to neonatal resuscitation being the most dramatic intervention in newborn care. The major themes in the domain of “development” were adapting known interventions to make them deliverable at the community level, adapting effective interventions to increase deliverability in health facilities in low and middle income countries, and approaches such as perinatal audits to improve quality of care to mothers and newborns. The themes in the domain of “discovery” included new, more effective and less expensive medicines for preventing preterm birth and treating sepsis, point of care diagnostics for infections, maternal vaccines to prevent newborn infections, and basic science work on causal pathways for identifying intervention targets and biomarkers for preterm birth, IUGR, and antepartum stillbirths. It is noteworthy that preterm prevention was not ranked highly, even though it may have the largest impact. This appears to be the result of these questions being scored low in answerability.

The relatively lower scores for the “development” and “discovery” groups of research questions may have several possible explanations. First, more than 95% of the neonatal deaths occur in low and middle–income countries (LMICs). Therefore, research addressing neonatal health issues that are relatively more important in wealthy countries may be perceived to contribute less to global reduction in mortality and morbidity, explaining some of the lower scores received by potentially promising research on novel interventions based on high technologies. Second, “discovery” research often takes longer to be translated into measurable benefits in terms of mortality burden reduction, and by definition the link to reduction in mortality and inequity is less direct. One specific example is research on prevention of preterm birth – while it was likely to have high impact, it was ranked only 129^th among the 205 questions. Thereby, respondents sent a message that this research question would likely be difficult to answer given the current stage of knowledge. Third, the process of delivery of novel interventions usually requires specific funding mechanisms, such as PEPFAR or Advance Market Commitment (AMC), which require time for a political agreement [34,35].

The CHNRI process we followed for setting priorities has several strengths. The methodology is transparent, replicable, and feasible to apply via e–mail [8–12, 27–33]. The output is intuitive and easily understood, and it has been refined and improved through many exercises over the past several years [36]. In this particular exercise, further improvements have been introduced to the process. We chose a large number of experts based on their productivity in the previous five years using Web of Science®, thus transparently identifying the group that was most likely to understand the field and its present research challenges and gaps. A very wide global network of programme experts in the Saving Newborn Lives’ Network was also invited. Moreover, we used online data collection tools, such as Survey Monkey® and Google Analytics®, which allowed monitoring of the progress of the exercise in real time, ensured adequate representation of experts by their background and region, and increased the efficiency of data management. Finally, 132 experts proposed research questions and 91 scored all the questions in this exercise; this is considerably more then in previous priority setting exercises using CHNRI methodology, where we typically involved fewer scorers, research ideas, and criteria scored by each expert.

There may be concern that the results derived from the CHNRI approach might represent only the collective opinion of the limited group of people who were included in the process. However, we were able to obtain questions and scores from a large number of experts worldwide, who were selected in a transparent and replicable manner, based on their research productivity in the field. The large number of participants and the protection against potential bias provided by the CHNRI approach make our results more credible, although it remains apparent that the highest scored questions may still be biased towards those that researchers are most familiar with and so may bias reflect research already in progress. This issue may be particularly relevant in view that only about a quarter of originally invited researchers, policy makers and programme experts eventually contributed to generating research questions, and only about one in six completed the scoring process, making response bias an important potential concern. Second, even though the list of proposed questions was reviewed and refined before sending for scoring, there were still overlaps in some research questions, possibly creating confusion in scoring such questions. Those and other possible strengths and limitations of CHNRI methodology are described and discussed in greater detail in Online Supplementary Document(Online Supplementary Document) .

A recent analysis of funding committed globally to improving neonatal health and birth outcomes has shown that donor mention of the “newborn” has increased quite sharply since 2005. However, given a total of only 10% of all donor aid to RMNCH mentioning the word “newborn”, and only 0.01% referring to interventions expected to reduce newborn deaths, it still seems unlikely that donor aid is commensurate with the large burden of 3.0 million newborn deaths each year, or with the burden of morbidity, developmental and long–term health outcomes [37]. The word “stillbirth” occurred only twice in the OECD database between 2002 and 2010, suggesting even lower attention for the world’s 2.7 million stillbirths.

Large inequities in current research funding support exist not only in the amounts invested in newborn health in comparison to other diseases globally, but also between different neonatal conditions themselves. Conditions that affect newborns in high–income countries receive more funding and attention than conditions that largely affect newborns in low–income countries. For instance, the research on care of preterm babies in neonatal intensive care units has received considerably more funding over the past several years in comparison to intrapartum–related birth outcomes or newborn sepsis [38].

The results presented in this paper will assist both the donors and the researchers in setting evidence based priorities to address the key gaps in knowledge, that could make the most difference in saving newborn lives and preventing stillbirth. In addition, attention to many of these questions could also improve maternal and child health outcomes. Likewise, research priorities to address other related areas such as maternal, child and adolescent health and health system issues may have substantial effect on newborn health. Complementary exercises are under way to identify research priorities in these areas. Using the identified research priorities, WHO, SNL and other partners, that are linked to the Every Newborn action plan launched in 2014 [39], will work to generate research interests among key national stakeholders, governments, NGOs, and research institutes, while encouraging research funders to support these priorities. We will track research funding, relevant request for proposals and trial registers to monitor if the priorities identified by this exercise are being addressed, and highlight those that are not being addressed.