The results of this consensus development process provide specific guidance to researchers and other stakeholders regarding priority knowledge gaps in the areas of child maltreatment, intimate partner violence and resilience. In all three areas, the top-ranked priority was to examine key elements of promising or successful programmes to build intervention pilot work. This emphasis on intervention development and testing in part reflects the PreVAiL mandate, but is based on the recognized gap in knowledge regarding proven-effective interventions in both CM  and IPV [9, 10], and the lack of even preliminary intervention work in resilience specific to IPV and CM [14, 15]. While promising interventions exist in some areas, these are often based on studies in specific groups and in better-resourced settings. Developing pilot work to take elements from promising existing programmes and services and adapt and test them in new contexts was viewed as an evidence-based, resource-effective and feasible approach to moving these fields forward. Similarly, in the IPV area, evaluating, using rigourous methods, existing services was a top-three priority.
There was a relatively wide range in the number of priorities identified, in large part reflecting the areas’ various stages of development with respect to research. For example, resilience research in the context of violence exposures is in its beginning stages [14, 15] and was deemed to require basic definitional and epidemiological work before moving to other kinds of research – this was a primary reason for keeping it as a separate thematic area, rather than trying to integrate it as a cross-cutting theme highly relevant to both CM and IPV. At a subsequent face-to-face meeting, the resilience theme group discussed at length the conceptual, definitional and methodological challenges in resilience research. They agreed that they viewed resilience as a dynamic life course process that was influenced by interactive individual, biological, social and environmental factors which may assist in the development, maintenance or regaining of mental health despite adversity. The group relied on a broad conceptualization of “mental health”, such as that endorsed by the World Health Organization, that went beyond psychopathology to include wellbeing. This led the resilience group to develop and approve the following definition, which will now be used by PreVAiL in its future work:
Resilience is a dynamic process in which psychological, social, environmental and biological factors interact to enable an individual at any stage of life to develop, maintain, or regain their mental health despite exposure to adversity.
While more research is available in CM and IPV, proven-effective interventions exist for only a fraction of possible settings and populations.
The highest degree of consensus was in the methodological category, where 65% agreed that investigating methods for collecting and collating datasets and conducting pooled, meta and sub-group analyses was the top priority (with appropriate caveats regarding inclusion of only higher-quality studies in these aggregated analyses, and the relative lack of such studies in some areas), along with better technological tools for tracking, surveillance and data linkage. For example, the existence of many high-quality national and international datasets is a potentially rich and efficient starting place to use evolving data linkage techniques  to answer questions regarding interrelationships among types of violence, risk and resilience factors, and mental health and addictions, using gender and sex-based analysis methods. Related to this, there was strong support for the integration of violence-specific questions in existing and new national and international surveys, as well as in administrative data sets. Though challenging, moving towards common definitions and questions across content areas would allow meaningful cross-sectional and longitudinal comparisons. There was strong consensus regarding the need to develop innovative ways to think about “evidence” broadly, and to employ rigorous designs of various kinds – quantitative, qualitative and mixed – to problems in these fields, finding appropriate ways to integrate different kinds of data in answering priority questions.
Across priorities, however, there was a recognition that both team-specific and external constraints must be taken into account when considering the feasibility of planning, conducting and implementing research, and this was true both from the researcher perspective, as well as from the policy and practice decision-maker partners.
The Delphi method was useful for our purposes for several reasons. First, it is a technique designed specifically to generate consensus from a panel of knowledgeable people. Second, it is a relatively quick and efficient technique, which utilized various communication tools to gather data from our globally-dispersed Network. Potential limitations of the Delphi approach have been noted , and Sackman , points out that the reliability of measurement and validity of findings using this approach are unknown. Nevertheless, recent critiques  have concluded that Delphi is a valuable research method when care is taken with its use; our identification of initial priorities using syntheses of best-available evidence, and known evidence gaps, lends credibility to our process. More quantitative approaches to assessing research priorities are emerging [25, 26], which include scoring priorities along specific dimensions, such as significance, answerability, applicability, equity and ethics , however, for the purposes of developing priorities within a relatively well-defined scope and among an established research group, the Delphi method yielded results that are specific and relevant, with consideration given to the kinds of dimensions listed above. In addition, beginning the process by building in part on pre-identified research gaps from the PreVAiL Research Briefs (Additional file 2), meant that evidence and systematic reviews based on English-language literature were privileged. However, the priorities we identified through this process complement the broader set of high-profile priorities and “grand challenges” highlighted for global mental health [26, 28, 29]. A potential follow-up to this process would include soliciting feedback from a broader group of identified stakeholders regarding these priorities, both to better align them with those in the broader context, but also to begin building opportunities for ongoing knowledge translation and exchange with those stakeholders.
In terms of lessons learned, the varying types and scope of PreVAiL’s expertise meant that some members felt able to provide input on some, but not all, topics, which is a reasonable approach given the scope of PreVAiL’s mandate. That said, a group comprised of more tightly-focused expertise in one of these content areas might provide a different set or ordering of priorities. In fact, comments related to feasibility pointed out that PreVAiL’s mandate and timeline are potentially limited, and thus, while broadly applicable, some priorities would have to be taken on by others. As one member said:
Addressing violence in low and middle-income countries (LMICs) is an important issue and one that deserves more attention. Whether PreVAiL should take this on or not depends on whether we have investigators willing to build on the inventory to advance intervention work in these countries. If there is no champion within the group, it may not be the best use of PreVAiL time and resources.
Therefore, based in part on discussions that arose during our Delphi process, as well as PreVAiL’s membership in the World Health Organization (WHO) Violence Prevention Alliance (VPA), PreVAiL has taken the lead for VPA in conducting a research priority-setting process, using a similar modified Delphi approach, on the topic of interpersonal violence prevention. The goal is to determine a violence prevention research agenda in LMICs, on behalf of WHO VPA, for the next five years. One of the limitations of the PreVAiL process – reliance on a majority of respondents from high-income countries – will be addressed with the VPA survey, which emphasizes participation of representatives from middle and low-income countries. Both have their strengths – the PreVAiL-based survey provided an in-depth focus on CM and IPV, while the VPA survey will address additional types of interpersonal violence, including youth violence and elder abuse, but will focus in less depth on the specific types of violence.
An additional challenge when beginning to address these priorities will be how to explicitly build-in a gender and sex-based analysis (GSBA) approach. In cases where the initial priorities suggest evidence synthesis or extraction of best/promising practices from existing interventions, the ability to incorporate GSBA will depend on how those initial data were collected and what is available in various datasets. However, as priorities are implemented that involve developing and testing new interventions, or adding new questions to surveys, etc., explicit incorporation of sex and gender variables and analyses should be prioritized.