Global research priorities for interpersonal violence prevention: a modified Delphi study

Abstract Objective To establish global research priorities for interpersonal violence prevention using a systematic approach. Methods Research priorities were identified in a three-round process involving two surveys. In round 1, 95 global experts in violence prevention proposed research questions to be ranked in round 2. Questions were collated and organized according to the four-step public health approach to violence prevention. In round 2, 280 international experts ranked the importance of research in the four steps, and the various substeps, of the public health approach. In round 3, 131 international experts ranked the importance of detailed research questions on the public health step awarded the highest priority in round 2. Findings In round 2, “developing, implementing and evaluating interventions” was the step of the public health approach awarded the highest priority for four of the six types of violence considered (i.e. child maltreatment, intimate partner violence, armed violence and sexual violence) but not for youth violence or elder abuse. In contrast, “scaling up interventions and evaluating their cost–effectiveness” was ranked lowest for all types of violence. In round 3, research into “developing, implementing and evaluating interventions” that addressed parenting or laws to regulate the use of firearms was awarded the highest priority. The key limitations of the study were response and attrition rates among survey respondents. However, these rates were in line with similar priority-setting exercises. Conclusion These findings suggest it is premature to scale up violence prevention interventions. Developing and evaluating smaller-scale interventions should be the funding priority.


Introduction
Almost half a million people are victims of homicide every year 1 and one in three women has experienced violence from an intimate partner at some point during her life. 2 Furthermore, a quarter of adults report having been physically abused in childhood and one in five women and one in 13 men report having been sexually abused in childhood. 3,4 Interpersonal violence during childhood can have serious, lifelong consequences that affect mental and physical health, academic and job performance and social functioning. 5,6 In addition, interpersonal violence, which includes child maltreatment, intimate partner violence, youth violence, armed violence, sexual violence and elder abuse (Box 1), create an economic burden on society. 7 Over the last two decades, the prevention of interpersonal violence has risen on the international public health agenda. 8 In May 2016, the World Health Organization (WHO) adopted a global plan of action to strengthen the role of health systems in addressing interpersonal violence, particularly against women and girls and against children. The 17 sustainable development goals (SDGs) recently adopted by the United Nations include four targets on interpersonal violence: (i) to eliminate all forms of violence against women and girls (target 5.2); (ii) to eliminate all harmful practices against women and girls (target 5.3); (iii) to reduce significantly all forms of violence and related deaths everywhere (target 16.1); and (iv) to end abuse, exploitation, trafficking and all forms of violence against children (target 16.2). 9 In spite of progress in the past 20 years, major gaps in violence prevention remain. The Global status report on violence prevention 1 reveals that civil and vital registration data on homicide are lacking in 40% of countries. Moreover, fewer than half of all countries have reported conducting population-based surveys on most forms of nonfatal violence, such as child maltreatment, youth violence and elder abuse. 1 Only 9.3% of all outcome evaluation studies in violence prevention have been conducted in low-and middleincome countries and there is no indication that this is increasing, despite over 85% of violent deaths occurring in these countries. 10 Research has a major role to play in reducing the global burden of interpersonal violence, by: (i) clearly defining the magnitude and distribution of violence; (ii) identifying risk and protective factors; (iii) developing effective interventions that target these factors to prevent and respond to violence; and (iv) increasing understanding of the legislative and policy environment and the human, institutional and financial resources required to scale up effective interventions. However, current research remains under-resourced relative to the burden of the problem, it is fragmented and disproportionately focused on high-income countries.
Objective To establish global research priorities for interpersonal violence prevention using a systematic approach. Methods Research priorities were identified in a three-round process involving two surveys. In round 1, 95 global experts in violence prevention proposed research questions to be ranked in round 2. Questions were collated and organized according to the four-step public health approach to violence prevention. In round 2, 280 international experts ranked the importance of research in the four steps, and the various substeps, of the public health approach. In round 3, 131 international experts ranked the importance of detailed research questions on the public health step awarded the highest priority in round 2. Findings In round 2, "developing, implementing and evaluating interventions" was the step of the public health approach awarded the highest priority for four of the six types of violence considered (i.e. child maltreatment, intimate partner violence, armed violence and sexual violence) but not for youth violence or elder abuse. In contrast, "scaling up interventions and evaluating their cost-effectiveness" was ranked lowest for all types of violence. In round 3, research into "developing, implementing and evaluating interventions" that addressed parenting or laws to regulate the use of firearms was awarded the highest priority. The key limitations of the study were response and attrition rates among survey respondents. However, these rates were in line with similar priority-setting exercises. Conclusion These findings suggest it is premature to scale up violence prevention interventions. Developing and evaluating smaller-scale interventions should be the funding priority.

Research
Research priorities for violence prevention Christopher R Mikton et al.
A systematic and transparent process of establishing global research priorities can provide useful guidance on allocating scarce resources more equitably and on developing a coherent research agenda. 11,12 Priority-setting exercises on research have long been carried out in other health fields. [12][13][14][15] However, in the field of interpersonal violence, such an exercise has only been conducted on child maltreatment and intimate partner violence in highincome countries. 16 The aim of this study was to identify global research priorities for the prevention of the main forms of interpersonal violence. The specific objectives were: (i) to rank the priority of the four steps (presented in Fig. 1) of the public health approach to violence prevention for each type of violence (i.e. child maltreatment, intimate partner violence, youth violence, armed violence, sexual violence and elder abuse) and the priority of broad subtypes of research questions within each step; 17 and (ii) to identify more detailed research priorities for the most highly ranked step. We chose the public health approach because it has been adopted by WHO and other national and global public health agencies to address a broad range of health issues, including violence and unintentional injury, and because it has been gaining prominence outside public health as a way of addressing violence.

Methods
We carried out the study from October 2010 until September 2013, in consultation with the 20-member Research Agenda Project Group of the WHO-led Violence Prevention Alliance. Our approach combined elements of the Delphi method and the Child Health and Nutrition Research Initiative priority-setting method. The Delphi method is a formal way of developing a consensus that is used when evidence in an area is limited or contradictory. Its aim is to determine, by means of an iterative process, the extent of agreement in that area. 14 This method often uses a large group of experts to generate research questions. The Child Health and Nutrition Research Initiative priority-setting method is a structured and transparent method that uses predetermined criteria to generate and score research questions systematically. This method assigns a quantitative research priority score to each item on a list of systematically generated research options based on scores given by experts using several criteria. Both methods have been extensively used to establish priorities in health research. [12][13][14][15][16]18

Study process
The study involved three rounds of expert consultations, which were conducted electronically (Fig. 2). We asked

Child maltreatment
The abuse or neglect of a child younger than 18 years. It includes all types of physical and emotional ill treatment, sexual abuse, neglect, negligence and commercial or other exploitation that result in actual or potential harm to the child's health, survival, development or dignity in the context of a relationship of responsibility, trust or power. Exposure to intimate partner violence is sometimes included as a form of child maltreatment.

Intimate partner violence
Behaviour by an intimate partner or ex-partner that causes physical, sexual or psychological harm. It includes physical aggression, sexual coercion, psychological abuse and controlling behaviours.

Youth violence
Violence occurring between people aged 10 to 29 years of age. It includes all types of physical and emotional ill treatment and generally takes place outside of the home. It also includes harmful behaviours that may start early and continue into adulthood. Some violent acts, such as assault, can lead to serious injury or death; others, such as bullying, slapping or hitting, may result more in emotional than physical harm.

Armed violence
The intentional use of physical force, threatened or actual, with arms against another person or group that results in loss, injury, death or psychosocial harm to an individual or individuals and that can undermine a community's development, achievements and prospects.

Sexual violence
Any sexual act or attempt to obtain a sexual act -including unwanted sexual comments or advances or acts to traffic a person for sexual exploitation -directed against a person using coercion by any person regardless of their relationship to the victim, in any setting, including but not limited to the home and work. It also includes rape, which is defined as physically forced or otherwise coerced penetration, however slight, of the vulva or anus using a penis, another body part or an object.

Elder abuse
A single or repeated act or lack of appropriate action that occurs within any relationship where there is an expectation of trust and which causes harm or distress to an older person. Elder abuse includes: (i) physical, sexual, psychological, emotional, financial and material abuse; (ii) abandonment; (iii) neglect; and (iv) serious loss of dignity and respect.
Note: We provided these definitions to survey respondents. participants to specify their areas of expertise on different forms of violence and, in each round, to give responses related to high-income countries and low-and middle-income countries, respectively. Respondents could give the same ranking to more than one research item if they judged them of equal priority. Given the broad scope of this exercise, which was the prevention of all the main forms of interpersonal violence globally, we regarded the Delphi method as an ideal way of asking a large group with extensive expertise to generate initial research questions (round 1) and to rank these questions on a scale of 1 to 7, with 1 being the highest priority, as appropriate (round 2). In round 3, the Child Health and Nutrition Research Initiative method was used to identify detailed research priorities for the step of the public health approach that ranked highest in round 2. This method produces a finely graded ranking by scoring each research question on several criteria. 12 For round 3, survey respondents were asked to grade 34 more detailed intervention research questions by rating them from 1 ("strongly dis-agree") to 5 "strongly agree") along five criteria (Box 2), which we developed from previous priority-setting exercises that used this method, 12,15 adapting them slightly. For each of the 34 questions, we calculated the mean rating and expressed it as a percentage (rather than out of 5). For example, if the mean rating across the five criteria was 3.8, we reported 76% (3.8/5).
To assess agreement across participants, we calculated intra-class correlations for each item measured on a 5-point scale. Correlations ranged from 0 to 1, with a value of 0.75 or above considered excellent. Respondents were also asked whether it was possible to rank the priority of research according to the type of violence and, if yes, to do so.
Potential respondents were first identified through the extensive global network of collaborators in WHO's Prevention of Violence Unit and the Violence Prevention Alliance, which includes some 65 organizations internationally. In addition, we asked potential respondents to suggest other experts in their region or country. In the two surveys used in rounds 2 and 3, we provided the definitions of key terms to survey respondents. These surveys were created using the web-based free open source

Round 1. Generation of initial research questions (Delphi method)
Aim: to generate research questions for the next 5 to 7 years that will be prioritized in subsequent study rounds. • 95 international experts were contacted and asked to propose research questions. • 158 research questions were proposed.

Round 2. Initial survey (Delphi method)
Aim: to rank the importance of research in the four steps, and various substeps, of the public health approach and the importance of cross-cutting questions for the six types of violence.
• 700 potential participants from 107 countries were contacted. • 280 participants completed the survey (response rate: 40%); respondents were based in 27 high-income and 38 low-and middle-income countries.
• 16% of participants gave responses that were specific to high-income countries, 45% gave responses specific to low-and middle-income countries and 39% gave responses applicable to both.

Round 3. Second survey (Child Health and Nutrition Research Initiative priority-setting method)
Aim: to rank research questions on interventions relevant to Step 3 of the public health approach. • 276 participants from Round 2 were contacted. • 131 completed the survey (response rate: 47%); respondents were based in 24 high-income and 24 low-and middle-income countries.
• Most participants chose to give responses for both high-income and low-and middle-income countries.
• Analysis of initial survey results and development of a second survey that focused on Step 3 of the public health approach (i.e. developing, implementing and evaluating interventions), which was ranked highest in Round 2. • Research questions relevant to Step 3 were selected from questions generated in Round 1, with the addition of research questions about violence prevention interventions that were known to be effective (i.e. supported by several high-quality evaluations) or promising (i.e. supported by at least one high-quality evaluation), as indicated by a literature review.
• Collation, refinement and grouping of research questions according to the four steps (and various substeps) of the public health approach to violence prevention (Box 2) for each of the six types of violence considered (Box 1); in addition, some questions were grouped into three topics that cut across these steps (i.e. knowledge translation, laws and methodological questions). • Identification of potential survey participants.

Box 2. Criteria for rating research questions on violence prevention, survey of global research priorities, 2010-2013
1. Significance: this research is important and needs to be carried out in the coming 5 years. 2. Feasibility: it is feasible to design and implement a study that addresses this research question in the coming 5 years. 3. Applicability, including effectiveness: conducting research into this question will influence practice and policy in the coming 5 years. 4. Equity: conducting research into this question will help under-resourced populations in the coming 5 years. 5. Ethics: research into this question can be carried out in an ethical manner in the coming 5 years.

Research
Research priorities for violence prevention Christopher R Mikton et al.
software LimeSurvey 19 and statistical analyses were carried out using SPSS version 22 (SPSS Inc., Chicago, United States of America) and SAS version 9.3 (SAS Institute, Cary, USA). We obtained ethical approval from McMaster University, Hamilton, Canada.

Generation of research questions
The demographic characteristics of the 95 study participants are shown in Table 1 (more detailed information on their region of residence is available from the corresponding author). The participants put forward 158 research items in the expert consultation. The suggested research questions were similar across country income levels. Most experts from low-and middle-income countries suggested research topics specific to these countries, whereas experts from high-income countries proposed topics for both high-income and low-and middle-income countries, either separately or without specifying the country income level. After collation and refinement, we grouped 26 research items according to the four steps ( Fig. 1) -and the substeps -of the public health approach to violence prevention. In addition, we grouped 20 items into three topics that cut across these steps (i.e. knowledge translation, laws and methodological questions).

Initial survey
The results of the research question ranking in round 2 are shown in Table 2 for all country income levels combined. There were clear trends in the ranking of the four steps of the public health approach.
Step 3 (i.e. developing, implementing and evaluating interventions) was ranked highest for child maltreatment, intimate partner violence, armed violence and sexual violence, whereas step 2 (i.e. identifying risk and protective factors and the causes and correlates of violence) was ranked highest for youth violence and step 1 (i.e. describing the nature, magnitude, distribution and consequences of violence) was highest for elder abuse.
Step 4 (i.e. scaling up interventions and evaluating the impact and cost-effectiveness of scaling up) was consistently awarded the lowest priority across all types of violence.
The ranking of broad subtypes of research questions within each step also showed marked trends across types of violence, particularly for steps 1 and 3. For step 1, research on the magnitude and distribution of violence was ranked highest for all types of violence except armed violence. For step 2, research on protective factors was ranked highest for four of the six types of violence. For step 3, research on evaluating the effectiveness of programmes that target actual violence was ranked highest for all types other than sexual violence. For step 4, participants ranked research on adapting effective programmes to new contexts highest for four of the six types of violence. Results for the cross-cutting questions are available from the corresponding author.

Second survey
In round 2, step 3 of the public health approach (i.e. developing, implementing and evaluating interventions) was awarded the highest priority for most types of violence for all country income levels combined (Table 2) and the second highest priority for low-and middle-income countries (results for low-and middle-income countries are available from the corresponding author). We decided to focus on step 3 in round 3 because the aim of the study was to establish global research priorities for interpersonal violence prevention rather than priorities for low-and middleincome countries specifically.
The second survey involved 131 experts scoring 34 interventions and seven cross-cutting questions applicable to step 3 using five criteria (Box 2) for both high-income and low-and middle-income countries. In Table 3 (available at: http://www.who.int/bulletin/volumes/95/1/16-172965), the 34 intervention research questions are listed by their overall research priority score, which was the mean score across all five criteria expressed as a percentage. Overall scores ranged from 83.4% to 70.0%. Across all items, scores for high-income and low-and middleincome countries were similar: the mean difference was 1% (standard deviation: 1%) and the maximum difference was 4.8%, which was for "increasing access to prenatal and postnatal services in health-care settings".
We examined three characteristics of the 34 interventions: (i) whether they were universal (i.e. directed at the whole population regardless of risk), selective (i.e. targeted at higherrisk subpopulations) or indicated (i.e. targeted populations that had already been exposed to violence); (ii) the type of violence they primarily addressed; and (iii) the risk factor they principally aimed to reduce. Universal and selective interventions had similar mean research priority scores (77.6% and 76.2%, respectively); the score for indicated interventions was 74.7%. Interventions that addressed child maltreatment had the highest mean score (79.7%), followed by those that addressed sexual violence (77.6%), intimate partner violence (77.3%), armed violence (76.5%), youth violence (75.4%) and all types of violence (75.1%). The single intervention that addressed the use of firearms as a risk factor had the highest mean score (83.3%), followed by those that addressed parenting (80.5%), social norms or laws (77.0%), alcohol (72.0%) and poverty or inequality (71.8%).
In response to a question about prioritizing research according to the type of violence, out of the 131 respondents, 58% (76) regarded it as possible, 28% (37) regarded it as not possible and 14% (18) expressed no view. There was no association between the respondent's area of expertise and their response. The mean priority ranking for the different types of violence, from 1 for highest to 6 for lowest, was child maltreatment (2.05), intimate partner violence (3.22), youth violence (3.46), armed violence (3.96), sexual violence (4.07) and elder abuse (4.43). There was no association between the respondent's area of expertise and the type of violence assigned the highest priority, except for child maltreatment, where Spearman's rank correlation coefficient was 0.54 (P < 0.0001).

Discussion
Globally, our priority-setting exercise found that research on the development, implementation and evaluation of interventions -step 3 of the public health approach to violence prevention -was ranked as having the highest priority. Research on identifying risk and protective factors and the causes and correlates of violence (step 2) was ranked second highest, though somewhat less consistently across different types of violence. Research describing the nature, magnitude, distribution and consequences of violence (step 1) was ranked third highest for most types of violence, with the notable exception of elder abuse, Around 85% of respondents provided this information in rounds 2 and 3. c All respondents were from high-income countries. d Around 84% of respondents were from low-and middle-income countries and 16% were from high-income countries. e Around 67% of respondents were from low-and middle-income countries and 33% were from high-income countries. f All respondents were from high-income countries. g Around 18% of respondents were from low-and middle-income countries and 82% were from high-income countries. h Around 32% of respondents were from low-and middle-income countries and 68% were from high-income countries. i The proportion of respondents who gave survey responses on a specific type of violence and also worked with that type of violence was 81% for child maltreatment, 80% for intimate partner violence, 80% for youth violence, 70% for armed violence, 76% for sexual violence and 67% for elder abuse.

Research
Research priorities for violence prevention Christopher R Mikton et al. The four steps of the public health approach 1. Describing the nature, magnitude, distribution and consequences of violence Step NA: not applicable. a Rank awarded by survey respondents in round 2 of the study to the importance of the research question for high-income and low-and middle-income countries combined. The rank was based on the mean ranking score awarded by respondents and ranges from 1 for highest to 7 for lowest, as appropriate. The number of respondents for each type of violence is given.
Research priorities for violence prevention Christopher R Mikton et al. for which it ranked highest. The most consistent finding, however, was that scaling up interventions and evaluating their cost-effectiveness (step 4) ranked lowest across all types of violence. One explanation for our main findings is that respondents considered it premature to scale up interventions (step 4) before there is sufficient evidence of an intervention's effectiveness (step 3). Such an interpretation is consistent with the findings of recent systematic reviews of interventions to prevent and respond to different forms of violence, which suggest that the evidence base remains thin and substantial investment in research is required. [20][21][22][23] Another possible interpretation is that respondents thought countries may lack the political will to scale up violence prevention interventions they view as being too costly or may lack the capacity to scale them up. 24,25 Although our findings converge with those of similar priority-setting exercises that focused on child maltreatment and intimate partner violence in high-income countries 16 and on adolescent sexual and reproductive health, including gender-based violence, in low-and middle-income countries, 18 they stand in stark contrast to recent calls to scale up violence prevention interventions. 26 Our finding that research on identifying risk and protective factors and the causes and correlates of violence (step 2) had the second highest priority overall, and the highest priority in low-and middle-income countries, concurs with recent reviews that concluded that the evidence base in this area is still limited, particularly on the causal status of risk factors and their relative importance. 27,28 Given the gaps in knowledge about the prevalence of fatal and nonfatal violence existing in many countries, 1 it is surprising that step 1, which includes describing the magnitude and distribution of violence, was ranked third for most types of violence. Perhaps respondents considered the gaps in research on other steps as comparatively greater and of more pressing concern. Also, it is possible that respondents were based in countries for which adequate knowledge of the magnitude of violence was available and they lacked a more global perspective.
Two noteworthy findings emerged in round 3 on ranking the 34 more detailed, intervention research questions. First, highest ranked questions were interventions that addressed violence against children and violence against women, both sexual and intimate partner violence. This may reflect the prominence of these types of violence on international agendas and acknowledges the importance of violence against children as a risk factor for involvement in other forms of violence, such as youth violence and intimate partner violence, throughout those children's lives. 29,30 Second, among interventions that targeted risk factors, those that addressed firearms or parenting were ranked highest, whereas those that addressed alcohol or poverty and social inequality were ranked lowest. However, in the absence of detailed, well-supported evidence on the relative importance of different risk factors for most types of violence and given the lack of consensus on other risk factors, such as the relative importance of poverty and social inequality as a risk factor for homicide, 31 these rankings may primarily reflect respondents' perceptions.
Our priority-setting exercise has several strengths. First, the number of experts who participated in the surveys and the number of countries, sectors and organizations they represented (Table 1) are as large or larger than most similar global research priority-setting exercises. 13,15,16 Second, the hybrid Delphi-Child Health and Nutrition Research Initiative method allowed us to identify priorities among and within the steps of the public health approach in the context of a complex field.
The study has several limitations. First, the response rate in round 2 was only 40% after follow-up reminders and there was an attrition rate of 53% between rounds 2 and 3. However, these response and attrition rates are in line with those of similar priority-setting

Rank Intervention to be researched
Mean research priority score a (%) . For each of the 34 questions, we calculated the mean rating and expressed it as a percentage (rather than out of 5). For example, if the mean rating across the five criteria was 3.8, we reported 76% (3.8/5). b The intra-class correlation indicates the level of agreement across survey respondents. A correlation of 0.75 or above was considered excellent. c Universal interventions are directed at the whole population, selective interventions target high-risk subpopulations and indicated interventions target populations that have already been exposed to violence.

Type of violence
(. . .continued) exercises. 12,15 A comparison of the available demographic characteristics of respondents and nonrespondents indicated they were similar but it is possible they differed on variables we were unable to assess. Second, the extent to which respondents were representative of the global community of violence prevention experts is unknown.
Nevertheless, the WHO and Violence Prevention Alliance networks we used to identify potential respondents are probably among the most extensive in the world. Third, use of the public health approach to organize research priorities may have dissuaded those unfamiliar with this approach from completing the surveys. However, the interventions respondents were asked to prioritize in round 3 were not specific to the public health approach and included interventions with which most experts were likely to have been familiar. Fourth, the length of the surveys and the interval between rounds 2 and 3 of almost 1 year may have discouraged some potential respondents. Fifth, it is possible that the decision taken in round 3 to focus on more detailed research priorities related to the step of the public health approach ranked highest in round 2, namely step 3, may have precluded the emergence of more detailed research priorities related to another step of the public health approach. Finally, this paper focused on the global results of this research priority-setting exercise; more finely grained analyses by region, country-income level and individual country will be published in the future.
This priority-setting exercise on global research into violence prevention showed that scaling up violence prevention interventions was consistently awarded the lowest priority, whereas de-veloping, implementing and evaluating interventions was awarded the highest. It appears that a massive investment in outcome evaluations, which matches the global burden of violence, is required before the field is ready to scale up preventive measures. The hope is that, within a decade, enough evidence will have accumulated to start scaling up interventions that will help achieve the ambitious SDG targets of altogether eliminating some forms of violence from the world and substantially reducing others by 2030. ■

Resumen
Prioridades de investigación globales para la prevención de la violencia interpersonal: un estudio de Delphi modificado Objetivo Establecer prioridades de investigación globales para la prevención de la violencia interpersonal mediante el uso de un enfoque sistemático.