Taking knowledge users’ knowledge needs into account in health: an evidence synthesis framework

The increased demand for evidence-based practice in health policy in recent years has provoked a parallel increase in diverse evidence-based outputs designed to translate knowledge from researchers to policy makers and practitioners. Such knowledge translation ideally creates user-friendly outputs, tailored to meet information needs in a particular context for a particular audience. Yet matching users’ knowledge needs to the most suitable output can be challenging. We have developed an evidence synthesis framework to help knowledge users, brokers, commissioners and producers decide which type of output offers the best ‘fit’ between ‘need’ and ‘response’. We conducted a four-strand literature search for characteristics and methods of evidence synthesis outputs using databases of peer reviewed literature, specific journals, grey literature and references in relevant documents. Eight experts in synthesis designed to get research into policy and practice were also consulted to hone issues for consideration and ascertain key studies. In all, 24 documents were included in the literature review. From these we identified essential characteristics to consider when planning an output—Readability, Relevance, Rigour and Resources—which we then used to develop a process for matching users’ knowledge needs with an appropriate evidence synthesis output. We also identified 10 distinct evidence synthesis outputs, classifying them in the evidence synthesis framework under four domains: key features, utility, technical characteristics and resources, and in relation to six primary audience groups—professionals, practitioners, researchers, academics, advocates and policy makers. Users’ knowledge needs vary and meeting them successfully requires collaborative planning. The Framework should facilitate a more systematic assessment of the balance of essential characteristics required to select the best output for the purpose.


Introduction
Increasing demands for the use of knowledge to assist evidencebased practice have led to a bourgeoning of different responses from funders and academics to evidence synthesis designed to support knowledge translation (Hansen and Rieper 2009). Each synthesis method and the type of output produced has its own merits and fulfils a particular knowledge need, for a particular primary audience, in a particular context. There are a number of factors that need to be considered when planning an evidence synthesis output including timeliness, length and format and the type of information to be included-whether solely research-based information, or the views of experts in the field, or a hybrid of both (Ogilvie et al. 2009;Abrami et al. 2010).
A diverse range of evidence synthesis outputs has been developed to meet users' knowledge needs, including evidence articles, evidence briefs, knowledge summaries and systematic reviews. Yet identifying the most suitable evidence synthesis method and type of output for a particular need may be far from straightforward. One reason for this is that the labels given to different forms of output are not standardized, leaving scope for misunderstanding when commissioning and designing such reports (Arksey and O'Malley 2005;Grant and Booth 2009).
Each potential audience has different knowledge needs and the evidence may need to be presented in different ways to enhance its utility. Based on the opinions of an expert panel, we focus on six primary audiences: researchers, academics (who may also be researchers), advocates (largely those working for non-governmental organizations, NGOs), policy makers, administrative and managerial professionals, and practitioners. (The latter two groups are concerned with policy implementation, through delivering services and may also include NGO workers). Each of these groups requires knowledge for different purposes (Table 1). Evidence syntheses may have mutliple users and be used at mutliple levels of the health system. The audience groups that we have not addressed are considered in the discussion section, as one of the study's limitations.
This study aims to contribute to an understanding of different users' knowledge needs and how they can be met through matching them with relevant evidence synthesis outputs. The objectives are to identify: different evidence synthesis outputs and their distinguishing features; as well as issues to consider when planning the development of an evidence synthesis to match users' knowledge needs.
We have created an evidence synthesis framework describing the features, benefits and limitations of outputs, based on a literature search, and consultations and interviews with experts in the field of synthesizing research for policy and practice. This framework should benefit both commissioners and producers of synthesis outputs-including knowledge brokers, who are responsible for deciding which type of output will best meet the needs of the evidence users they support.
The scope of this study is the wide range of diverse evidence synthesis outputs, which encompasses, but is not exclusive to systematic reviews. Much of the existing literature focuses on methodologies to analyse quantitative and/or qualitative studies that are variants of systematic reviews, e.g. Gough and Elbourne 2002;Mays et al. 2005;Tricco et al. 2011;Hansen and Rieper 2009. These are welldefined, distinct approaches (e.g. meta-analysis, or realist, diagnostic test or complex reviews etc.). However in this study, the nature of systematic reviews is acknowledged as a generic type of evidence synthesis output.

Methods
A four-strand literature search, described below, was conducted to ascertain what research exists that contributes to answering the study objectives. Using the methodology for a systematic review was

Key Messages
• The increased demand for evidence-based health policy in recent years has provoked a parallel increase in diverse evidence-based outputs designed to translate knowledge from researchers to policy makers and practitioners, yet matching users' specific knowledge needs to the most suitable output, while essential, can be challenging. • We have developed an evidence synthesis framework classifying 10 distinct evidence synthesis outputs under four domains: key features, utility, technical characteristics and resources, in relation to six primary groups of usersprofessionals, practitioners, researchers, academics, advocates and policy makers. • We propose a process for matching users' knowledge needs with an appropriate evidence synthesis output, using essential characteristics to consider when planning an output-Readability, Relevance, Rigour and Resources. • When used in combination, the framework and process should facilitate a more systematic assessment of the balance of essential characteristics required to select the best output for the purpose and help knowledge users, brokers, commissioners and producers decide the best 'fit' between 'need' and 'response'. not feasible because of the nature of the documents on which the literature search was based. Such documents, for policy makers and a general audience are not generally found in databases of academic peer-reviewed articles. Nevertheless, the methodology we used followed parameters which were intended to make it systematic. The first strand of the literature search was a search of five bibliographic databases of peer-reviewed journal articles: Embase, Global Health, Medline, Social Policy & Practice and Web of Science. Based on the number of relevant articles from particular journals identified in the database search, the second strand was a hand search of three peer-reviewed journals that were considered particularly relevant: Systematic Reviews Journal; Journal of Health Services Research & Policy; and BMC Medical Research Methodology. The third strand was a search for relevant grey literature using Google. This was not exhaustive, but was as comprehensive as possible, representing five leading organizations involved in producing evidence syntheses: the UK Department for International Development (DFID), the Overseas Development Institute, INASP (an international development charity working with a global network of partners to improve access, production and use of research information and knowledge), the UK Economic and Social Research Council (ESRC) and the International Initiative for Impact Evaluation (3iE). DFID's Research for Development (R4D) database was also searched.
The inclusion criteria for the literature search were that articles were written in English and were either review or discussion articles. The initial search terms used for the first two strands of the literature review were: evidence synthesis (singular and plural) AND methodology.
A second search of the bibliographic databases was then undertaken using the search terms (expert opinion OR consensus statement) AND policy making.
Once the search results were compiled using Endnote, the titles and abstracts (or executive summaries) of all records were appraised and 49 were considered to be relevant. Given the small number of documents, one researcher read all 49 in full and made a decision as to whether or not they met the study objectives of identifying different types of evidence synthesis, or highlighted issues to consider when planning the development of an evidence synthesis to match users' knowledge needs (Supplementary data 1 are available at HEAPOL online). Twelve documents were considered to meet these objectives. The fourth strand of the literature search was to use a snowball technique to identify further documents from the references cited in these 12 documents, as well as two key documents identified by the expert panel we consulted, bringing the total number of relevant documents to 24. Figure 1 shows a flow diagram of the literature search. One researcher conducted the literature search and the decisions made were reviewed with a second researcher at regular intervals.
Experts in synthesis designed to get research into policy and practice were also consulted to hone issues for consideration and ascertain key studies. We consulted with a panel of eight experts, from diverse backgrounds, with experience of producing evidence syntheses. They were selected purposively because they represented the various types of expertize needed to produce such outputs and included a leading research scientist involved in knowledge translation, health system researchers, advocacy and communications specialists and representatives from large organisations that regularly produce evidence synthesis outputs, and advisers to policy makers (Supplementary data 2 are available at HEAPOL online). Prior to the literature search, a discussion guide was devised to focus phone and face-to-face meetings with four of these experts. It included identifying the need for evidence syntheses, the value of a questionbased evidence synthesis, the value of synthesized evidence versus expert opinion, sound examples of typologies of evidence syntheses and different types of evidence synthesis outputs and their relative validity.
These meetings developed into free-flowing discussion, providing insights and suggestions that helped to determine some of the essential characteristics of different types of evidence synthesis outputs. These discussions informed a manual synthesis of the literature search findings, from which a framework and report were developed with the participation of all eight experts, who gave useful feedback, particularly in fine-tuning the framework and recommendations.

Results
We identified 10 different forms of evidence synthesis outputs and have classified them in an evidence synthesis framework. The Framework arranges the characteristics of these outputs under four domains: there is a brief description of each output's key features; its utility for the primary audience we suggest it is best suited to; technical characteristics, including limitations; (Tables 2-4) and the production resources that should be considered, in order to meet knowledge users' needs, such as a timeframe ( Figure 2).

Different forms of evidence synthesis outputs and their distinguishing features
These outputs synthesize different types of evidence; some include evidence outside that produced by scientific research. Hansen and Rieper (2009) observe the rise of evidence-based policy making and delivery in Europe since the 1990s and differentiate between the forms of evidence used, based on Eraut's (2004) work on the credibility of evidence used for decision making. Eraut distinguishes between research-based evidence in peer-reviewed published research; other scientific evidence (generated using scientific procedures with a track record of producing valid results); and practice-based evidence (derived from recognized professional practices that have been undertaken using criteria expected by experts within the profession). Any, or all of these could make a valid and useful contribution, but may not in themselves be sufficient to meet policy makers' needs. (Mays et al. 2005) The Partnership for Maternal, Newborn and Child Health (PMNCH) strategy briefs (2014b) are an example of practice-based evidence syntheses combined with tools to develop and implement strategies to inform advocacy, policy and practices.
We found a number of studies that describe some of the different evidence synthesis outputs in similar terms, and these have contributed to the development of the evidence synthesis framework, yet none covers all four domains. For example, to help commissioners identify which evidence synthesis output would best suit a particular need, the UK Civil Service (2010) and DFID (2013) suggest when an output might be useful and its limitations, but neither includes many technical characteristics. Other frameworks are based on synthesis methods, but do not take users' perspectives or the resources required into account. Grant and Booth (2009) present a comparison framework based on the four main processes used to review evidence-Search, AppraisaL, Synthesis and Analysis (SALSA)-to distinguish between different syntheses and define their characteristics. Classification differences mean that some of the outputs they identify share a definition in the Framework we have developed. Kastner et al. (2012) also map the characteristics of existing evidence synthesis methods, and Tricco et al. (2011) use the qualitative or quantitative nature of sources of evidence to tabulate the characteristics of different synthesis methods, which they refer to as ' . . . types of systematic reviews'.
Other studies focus on evidence synthesis outputs guided by a clear question and primarily synthesizing research evidence, and present methodological frameworks based on the type of research question to which an answer is sought (Petticrew and Roberts 2003;Mays et al. 2005). The need for an evidence synthesis to have a research or learning question came up repeatedly in the literature search and was discussed with the expert panel. A carefully structured research or learning question can help to clarify and target the literature search and places the synthesis within a context, including a theoretical context (Gough and Elbourne 2002) and some consider that it guides the whole production process (Gough and Elbourne 2002;Petticrew and Roberts 2003;Mays et al. 2005; DFID Evidence Brokers 2013).
For researchers and practitioners, who are generally concerned with impact and effectiveness issues, well-established outputs that are primarily based on research studies-such as systematic reviews-are designed to answer specific impact questions, e.g. What evidence is there that misoprostol can prevent postpartum haemorrhage? Although the knowledge to action (KTA) evidence summaries prepared as part of a collaborative project between the Champlain Local Health Integration Network and the University of Ottawa, funded by Canadian Institutes of Health Research, were not initially developed with a predetermined primary research question, user feedback suggested it would provide 'clarity and direction' (Khangura et al. 2012). An iterative process was built into future summaries, so that the research team worked with users to agree a research question. Similarly, Chambers and Wilson (2012) propose a checklist by which researchers and users' representatives, or commissioners can clarify the research question.
UK Civil Service guidelines (2010) group evidence that can be synthesized around non-impact questions e.g. needs, process, implementation, correlation, attitude and economic questions, such as; How much does it cost to deliver misoprostol to pregnant women in community settings? Yet a research question may not be a key requirement for all knowledge users; for some, a more general focus might be appropriate. Advocates, policy makers and implementers may have a variety of issues to consider and require a range of evidence beyond scientific research, to guide them (Sheldon 2005;Lavis et al. 2009;Abrami et al. 2010). Davies (2006) notes that policy makers often want answers to broad questions, which may not always be sufficiently focussed to guide a tight search for evidence beyond that available from research; 'such as administrative data and evidence used by lobbyists, pressure groups and think tanks (which may or may not be research based)'. While there are a limited number of databases available to help guide such searches, e.g. Open Grey, these are not exhaustive and often have a basic search function. A clear statement of the issue might be a more suitable starting point (Gough and Elbourne 2002;Petticrew et al. 2004;Mays et al. 2005; UK Civil Service 2010; Chambers and Wilson 2012; Khangura et al. 2012) as in the PMNCH knowledge summaries (2014a). Our evidence synthesis framework distinguishes between those evidence synthesis outputs which address a specific research question and those which provide a broad thematic overview of the evidence relating to issues in a policy area, such as significance, as in the PMNCH knowledge summary Maternal mental health: Why it matters and what countries with limited resources can do (Hashmi 2014).   Variations in the names and characteristics of some types of evidence synthesis outputs meant that categorizing them in the Framework was not always straightforward. For example, the Alliance for Health Policy and Systems Research (2011) splits synthesis outputs into short syntheses and systematic reviews, noting that the names of short synthesis documents-policy brief, research summary and briefing note, ' . . . are typically used indiscriminately, and could refer to similar or highly dissimilar ideas'. It reclassifies short synthesis outputs, by the type and extent of the information they summarize.
While standardizing the names and methods would help clarify and distinguish between outputs with partially or fully overlapping characteristics, some researchers consider this unnecessary or even restrictive, suggesting that a preferable solution would be to include a transparent statement of methods in each output (Gough and Elbourne 2002;Watt et al. 2008;Ganann et al. 2010). The Effective Health Care bulletins, commissioned by the English Department of Health, are one example where methodological information is included (Centre for Reviews and Dissemination 2004). Another is the evolution of evidence summaries produced under the KTA

Factors to consider when planning an evidence synthesis output
Planning an evidence synthesis ideally involves collaboration between those commissioning and those producing an output. The challenge is to ensure that it meets the users' specific information needs, is user-friendly, timely and credible (Sheldon 2005). Consideration of some essential characteristics should help. When offering guidance to researchers writing for a diverse audience, Largay (2001) identifies Three Rs-Readability, Relevance and Rigour as essential characteristics. Rigour relates to the systematic and transparent application and recording of the method used. Relevance refers to planning the scope of the evidence synthesis to fit the knowledge requirements of potential users, ensuring timely production and identifying the primary audience-why the research topic is important to them and what the context is. Readability includes using plain, non-technical language, clarity of thought and a brief summary or visual display of the conclusions reached.
Considering the three Rs should help secure a credible, timely and appropriate output that meets users' needs. Grant and Booth (2009) and Thomson (2013) highlight a tension between rigour and relevance, given that the opportunities for using an output, for example within a defined policy window, may not allow sufficient time to undertake a systematic review. To help address this, Thomson (2013) considers the Three Rs as 'interrelated principles' that can be applied to planning evidence syntheses, particularly complex reviews to support policy making, and suggests they are considered in relation to a fourth R-Resources available for production (including time, funding and personnel). This helps determine a feasible and relevant scope for the synthesis output within the time available. Building on Thomson's concept, Figure 3 shows how the Four Rs fit into a process for matching information needs with appropriate evidence synthesis outputs: once the need for synthesized evidence has been established, an acceptable balance between the Four Rs is agreed and used to make an objective assessment of the types of evidence synthesis outputs, to help identify the most appropriate output.
Relevance often relates to the particular context in which evidence synthesis outputs are to be used (Petticrew et al. 2004;Sheldon 2005;Ogilvie et al. 2009;Chambers and Wilson 2012;Saul et al. 2013). Researchers and producers of evidence syntheses need to develop some understanding of the knowledge needs of the primary audience and the environment in which they are working so as to analyse and present the information in a way that is relevant and helpful to users (Sheldon 2005). Such factors may relate to context, cost effectiveness and expert-or even public-opinion (Ogilvie et al. 2009), e.g. PMNCH strategy briefs (2014b) are often produced in more than one language and use regional case studies, to support international or regional meetings.
A study eliciting the views of UK policy makers on how research evidence influences public health policy found that the attributes of evidence synthesis they considered to be important were broadly in line with three of the four Rs: clarity, timeliness and relevance to current policy debates, with the addition of attending to evidence of cost-effectiveness (Petticrew et al. 2004). In some instances, the inclusion of different types of evidence drawing on a wide range of information sources may be best suited to the production of a hybrid output that offers a peer-reviewed synthesis of recent scientific evidence with practical information for policy makers and practitioners (Abrami et al. 2010), such as the PMNCH knowledge summaries (2014a).
The relationship between the relevance of a synthesis output and the resources available to ensure its timeliness is an important planning consideration (Saul et al. 2013;Thomson 2013). Figure 2 gives indicative average production times for each of the evidence synthesis outputs in the evidence synthesis framework. Consideration of this and other resource issues by both commissioners and producers will likely affect various aspects of an output, including its rigour, depth, quality appraisal and scope. For example, resources generally influence the number of reviewers who can be employed to work on an output in the time available. Abrami et al. (2010) make this distinction clear by using brief review to describe a synthesis limited in both timeframe and scope, and comprehensive review, for one which is time bound, but not limited in scope because a number of researchers can work on it.

Discussion
The Framework identifies 10 different forms of evidence synthesis outputs drawn from the literature search and consultation with experts. It shows the range of outputs that have been developed in recent years to accommodate different evidence needs, beyond clinical decision making. Given the confusion produced by the many different terms used in the literature to describe these various forms of evidence synthesis outputs, the Framework, used in conjunction with the process for matching users' information needs with an appropriate evidence synthesis output, is intended to offer greater clarity to users, commissioners and producers of outputs.
Using the process outlined in Figure 3, in conjunction with the evidence synthesis framework, offers a more systematic approach than was previously available to planning an appropriate evidence synthesis output by ensuring that all the essential features and characteristics, including resources, are considered. If planning is an iterative and participatory collaboration between users and/or commissioners and the production team, it will be a significant contributing factor towards producing an output tailored to meet users' knowledge needs (Watt et al. 2008;Khangura et al. 2012;Saul et al. 2013) and increase the prospect of research being used in policy development (Corluka et al. 2014). Once the need for an evidence synthesis has been identified, those commissioning it should consider what sorts of evidence would be relevant and the level of rigour with which the evidence needs to be analysed for the particular context in which the synthesis will be used. In addition, the level of knowledge and understanding of the end-users needs to be appraised, to guide the level of technical language and detail that is required. Alongside these considerations, the resources available for production should also be taken into account. Taking the decisions made on relevance, rigour, readability and resources a match can then be made using the outputs listed in the Framework and the indicative average production times, in order to identify the most suitable output.
The strength of our approach was that we consulted with specialists in this field to guide the focus of the evidence synthesis framework and the process for matching users' information needs with appropriate evidence synthesis outputs, but we acknowledge that in this field other perspectives on the issues considered may exist. Our approach had inevitable limitations. We were only able to search peer-reviewed studies and grey literature in English, and documents that were not widely available on the Internet, such as NGO reports, were not included. The specific needs of audience groups such as industry, the private sector, the media and the general public (who other than when involved in advocacy, have no defined role) were beyond the scope of this study. Nevertheless, this study addresses the needs of a wide range of users. An assessment of the in-depth knowledge needs of other audiences may require some adaptation of the framework. Furthermore, while it was beyond the scope of this study, the use of the framework in conjunction with the process for identifying knowledge users' information needs with an evidence synthesis output, would benefit from being pre-tested and pilot tested with different groups of knowledge users. Although the process currently suggests equal weighting is given to considerations of rigour, relevance, readability and resources, we would expect that different groups of policy and decision makers might emphasize different components in different contexts. For example, the primary concern for academic stakeholders might be rigour, while policy makers might consider readability and relevance to be of primary importance, and practitioners might prioritize relevance. The emphasis given to each component might lead to the adaptation and development of the framework, in order to increase its utility to different user groups.

Conclusion
Users' knowledge needs vary and meeting them successfully requires collaborative planning. The Framework describes the various evidence synthesis outputs identified and the process for matching users' information needs with an appropriate output. It is intended to offer a more systematic way for users, commissioners and producers to establish a common understanding of users' knowledge needs, and the essential characteristics to be considered when matching those needs with the most suitable output, given the resources available.
Further work would help to address the limitations of this study, such as taking the knowledge needs of other audiences into account.

Supplementary Data
Supplementary data are available at HEAPOL online.