A “Rapid Best‐Fit” model for framework synthesis: Using research objectives to structure analysis within a rapid review of qualitative evidence

The short time frame associated with rapid reviews can be challenging for researchers conducting qualitative evidence synthesis. In these circumstances a Best‐Fit Framework Synthesis, drawing upon existing theory and/or research, may be conducted to rapidly make sense of qualitative evidence. This article discusses a “Rapid Best‐Fit” approach to conducting Framework Synthesis within an 6‐week rapid systematic review of qualitative evidence. In the absence of a suitable theoretical model to inform the synthesis, we drew upon our research objectives to structure a framework, and to identify the studies which contained the most relevant data. Themes from these studies were used to revise the initial framework before inductive thematic synthesis finalized theme content. This “Rapid Best‐Fit” approach yielded results aligned with the needs of the commissioners of the review and is a useful addition to qualitative synthesis methods, allowing for the incorporation of diverse views and experiences into fast‐paced decision making scenarios.


| INTRODUCTION
The importance of qualitative research to answer pressing policy and practice questions is increasingly being recognized, 1 with reviews commissioned to meet policy or practice specific needs often needing to be conducted within a rapid timeframe. [2][3][4] This can present a challenge for most established methods of qualitative evidence synthesis, such as thematic synthesis or meta-ethnography, which are usually conducted over a longer time period, and there is currently very little methodological guidance on how to undertake rapid qualitative evidence synthesis. 5 To address the need for qualitative synthesis methods that can be undertaken within a rapid timeframe, a method used to support the analysis of qualitative data within primary research, Framework Analysis, has recently been adopted for use in qualitative evidence syntheses. 6 It uses a deductive approach of mapping data from primary research studies onto a framework constructed of pre-identified themes, concepts, theories or ideas. This initial framework is then revised using an inductive, iterative analysis which can help incorporate any as yet un-coded data, resulting in the final framework being a mix of initial and revised themes. 6 Ritchie and Spencer 7 outlined five stages of using framework analysis within primary qualitative research: 1. Familiarisation: the process by which researchers immerse themselves in the data to understand the range of views and sources which are represented and begin to note key ideas or themes. At this stage, depending of the volume of material and time available for synthesis, the researcher may prioritize certain material for further analysis. 2. Identifying a thematic framework: Informed through the familiarisation process, the researcher will identify key concepts and issues arising out of the primary material to be analyzed and apply it to the first few transcripts. This framework can then be revised in light of additional material gained through applying the framework to additional transcripts. 3. Indexing: The process of applying the chosen framework to the textual data. 4. Charting: Identifying the range of different experiences and views within each theme, noting any associations and patterns. 5. Mapping and Interpretation: Considering emerging associations and themes in light of the review question to create and explain key concepts representing core issues within the data. This stage should also note any contrast/diversification between ideas.
Within qualitative evidence syntheses, this method is useful for rapidly synthesizing and organizing qualitative primary research data into a coherent output in timepressured situations. 8 Figure 1 illustrates how the five stages 1 originally outlined by Ritchie and Spencer 7 can be integrated within the systematic review process as defined by Gough et al. 9 There is some flexibility in how the stages outlined by Ritchie and Spencer 7 can be applied to the different stages of the systematic review process. The methods selected for framework synthesis can be influenced by several factors, including the aims of the review and the stages of the review in which framework synthesis was applied. 10 The selection of an initial framework usually follows one of two main approaches. An initial framework can be based upon researchers' previous experience and understanding of the relevant background literature within their topic area. 11 As part of this process, the researcher may have conducted extensive literature searching and/or consultation with topic experts and other key stakeholders to develop a framework which contains key components suggested as relevant and important in previously conducted research. 10 Alternatively, a method known as a "Best-Fit" framework synthesis can be used. This method involves selecting a pre-existing theory or model on which to base the framework 8,12 and can comprise pre-existing themes, concepts or ideas from existing research or legislation, from one or multiple sources. However, in circumstances where neither of these approaches is possible, adaptation of the method may be necessary. This paper presents a worked example of novel approach to framework synthesis termed "Rapid Best-Fit," based upon a revised Best-Fit methodology focusing on the development and evolution of a framework where no existing or accepted theoretical model could be identified prior to synthesis, and where time constraints prevented us from constructing our own framework.

| SYSTEMATIC REVIEW METHODS
Our example is drawn from a recent Department of Health and Social Care commissioned systematic review intended to inform the Independent Review of the Mental Health Act (1983). 13 This was a rapid review of qualitative evidence, conducted within 6 weeks, focussing on Highlights • Best-Fit framework synthesis can be used to rapidly make sense of qualitative research eligible for inclusion within a systematic review. This type of synthesis uses a framework based upon previous knowledge or existing theory to understand and integrate qualitative data within the primary research studies. • However, there may not be enough existing relevant theory to create an initial framework and, as may be the case for rapid reviews, researchers may not have the time to develop their own framework based upon the background literature. • This paper describes the development and use of the "Rapid Best-Fit" approach within an 6 week rapid review. We demonstrate how an initial framework can be based on the research objectives of a systematic review, and developed using themes from the primary studies contributing the most relevant data to the research objectives and thematic synthesis. • This approach may be useful to researchers, health and social care clinicians and policy makers undertaking systematic reviews in time-pressured situations to inform their decision making the experiences of service users, carers and professionals of the Nearest Relative provisions of the Mental Health Act. 14 A protocol was registered on the PROSPERO database (PROSPERO CRD42018088237).
The methods used to identify and select evidence followed the best practice approach recommended by the University of York's Centre for Reviews and Dissemination 15 and are detailed in full within the main project report. 16 Table 1 below reports the key decisions made during this review to manage the tension between our desire to maintain a robust, transparent and reproducible approach, while also ensuring this systematic review was deliverable within the 6-week timeframe. Table 1 highlights how our decision to use a framework synthesis approach to integrate the qualitative data from the primary studies included in our review was key to ensuring our review could be completed within 6 weeks. Below we focus in detail each stage of our synthesis in line with each of the five stages of framework synthesis identified by Ritchie and Spencer 7 within the model described by Brunton Prior to finalizing the aim and research objectives, we undertook a 2-week scoping process to immerse ourselves in the literature, familiarize ourselves with the Mental Health Act (1983) 14 and other key legislation, as well as key issues associated with the Nearest Relative role. We also liaized closely with the Department of Health and Social Care (DHSC) to ensure that the aims of the review reflected their needs, the quantity and type of research available to review and our remaining timeframe of 6 weeks.

| Systematic review aim and objectives
The agreed aim of our qualitative evidence synthesis 16 was to summarize and synthesize experiences of the Nearest Relative provisions of the Mental Health Act 14 from the perspectives of service users, family members, carers and relevant professionals. Our five research questions were identified and refined through consultation with the DHSC, our policy customer, which were to explore: 1. Experiences relating to the identification of the Nearest Relative in the care of an individual who has been compulsorily detained under the Mental Health Act (1983); 2. Experiences of requesting displacement of the assigned Nearest Relative, including the process of going through a tribunal and issues associated with this, such as influences on ongoing care; 3. Issues relating to service user confidentiality and information-sharing, relating to all aspects of compulsory detention; 4. Issues relating to decisions about care during detention and after discharge, including discharge to a Community Treatment Order; 5. Issues relating to service users having access to support from individuals who they want to be involved with or informed about their care.

| Search for relevant research
The search strategy was informed by our scoping searches conducted within MEDLINE, Google Search and Google Purposive sampling approach used to prioritize best available evidence for inclusion. Studies containing at least half a page of qualitative data directly relating to the research objectives were prioritized for inclusion in the framework synthesis.
Data extraction and quality appraisal -• Data extraction and quality appraisal completed by one reviewer and checked by a second. • Quality appraisal was undertaken by using the complete version of the Wallace checklist.
Synthesis -• Data relevant to research objectives extracted into initial framework by one reviewer and checked by a second. • Selection of themes to revise the initial framework was made by one reviewer and checked by a second. • The final revisions to the framework following inductive thematic synthesis were made by one reviewer and then checked with the review team and one carer with experience of the NR provisions.
• Choice of framework synthesis to synthesize first and second order data from included primary studies. • Initial framework based upon initial research objectives of our systematic review. • Revisions to the initial framework were based upon themes from the primary studies which contained the highest quantity of first or second-order construct data relevant to multiple research objectives. While an inductive thematic synthesis would have suited an exploration of experiences within this topic area, it was not possible to use this approach due to the limited time available. We considered confining our synthesis to content analysis, 21 to produce a description of the themes across the included studies but felt this approach would not fully reflect the views and experiences captured within the primary studies. Instead, we decided a framework synthesis represented the most pragmatic way to meet the aims of the review within the timeframe. We initially sought to draw upon existing theories or models when designing a framework as outlined by Booth and colleagues in their description of a "Best-Fit" approach. 12 Unfortunately, neither the background scoping nor the studies included in the review yielded a single model or theory we felt we could use as a basis for the synthesis. The theories and models which we identified within studies were either heavily influenced by a very specific theoretical approach not generalizable to broader contexts, 22 or were not comprehensive enough to be applicable to the majority of the research objectives. 23 We were also uncertain if these frameworks could be applied to people with experiences of the different iterations of the Mental Health Act (1983) used in Scotland and Northern Ireland.
As described above, another option would have been to construct a new framework based upon our own understanding of the literature. 6 While the period of intense background reading and familiarisation had equipped us with a working knowledge of the Mental Health Act and its Nearest Relative provisions, 14 we were concerned that our understanding of the important issues within the area may be incomplete thereby introducing bias to the initial framework. Although the framework synthesis approach does provide the opportunity to revise the framework in light of new insights, we were concerned that an inadequate initial framework could result in relevant material being missed during the initial data extraction phase. We had a 2-week period available to us for synthesis, which prohibited any in-depth re-reading of original data sources after the completion of data extraction.
To ensure we extracted all relevant data from the included studies, we based our initial framework directly upon the five research objectives as agreed with the policy customer. A framework was constructed using Microsoft Excel where each research objective was allocated a separate sheet, within which each paper was given a separate row. Each column depicted a different type of participant perspective, including the views of service users, carers, Nearest Relatives, policy makers and solicitors.
Within her systematic review, Brunton 10 conceptualizes inclusion screening and data extraction as contributing toward the "Indexing" process of framework synthesis. Our processes for screening, prioritisation and initial data extraction, including quality appraisal, are described in full within main project report. 20 In addition to these stages, and consistent with the approach used by Brunton 10 we extracted summary data to describe the main features of each of our included studies and key data to determine each study's relevance to our review question. This enabled us to prioritize papers for further synthesis.

| Prioritisation of articles for further synthesis
Following Full-text screening, 35 papers from 20 studies were identified as eligible for inclusion in the review. We were not able to include all of the papers that met our inclusion criteria in the framework synthesis within the time available. Thus, a pragmatic approach was required to prioritize certain papers for further synthesis. All eligible studies were read and independently rated by two reviewers (Initials removed for anonymity) according to the quantity of relevant first and second order data they contained (high, medium or low), with disagreements resolved through discussion. The country and participant viewpoints represented within each study were tabulated, an example of which is shown in Table 2.
Following this process, 22 papers from 12 studies, containing at least half a page of qualitative data directly applicable to the research objectives, were prioritized for inclusion within the framework synthesis. We checked that the prioritized studies represented participant views from across England, Scotland and Northern Ireland. We found no studies conducted in Wales. Details of the nonprioritized studies including country, research aim, population sample, data collection method and analytical approach were summarized in a table. Studies based upon data collected through surveys/questionnaires were not prioritized due to the absence of data stemming from open questions.

| Data extraction and quality appraisal
During the first stage of data extraction, we extracted summary data into a Microsoft Excel data extraction form by one reviewer (LS, MN) and checked by a second (MN, LS). The Wallace checklist was used to appraise the quality of the studies prioritized for further synthesis, and was not used as a basis to exclude studies from this review. 44 Instead the quality assessments inform the amount of confidence reviewers could place in the findings of each study.
For the 22 prioritized papers, we extracted participant quotes and author interpretations from the results sections of each of the prioritized articles using the initial framework based upon our research objectives. Articles based upon studies conducted in England, Scotland and Northern Ireland were considered separately to allow for the identification of unique themes relevant to different iterations of the Mental Health Act (1983). An example of a completed stage 1 framework synthesis data extraction sheet is provided in Table 3.
Our objectives were very broad and thus allowed us to capture the majority of the data relevant to our research questions (as determined by two reviewers). Where there was any uncertainty about relevance of the first/secondorder construct data to our research questions, this data was extracted into a spreadsheet of similar structure to Jankovic et al, 2011 24 Smith, 2015 25 Stroud et al, 2015 26 Rapaport, 2012 27 Gault, 2009 28 Rugkasa and Canvin, 2017 29 Stroud et al, 2014 30 Rapaport, 2004 31 Henderson, 2002 24 Rapaport, 2012 27 Rapaport, 2003 32 Gault et al, 2013 33 Gault, 2009 28  that outlined in Table 3. This content was then considered for inclusion/informed the revision of the framework across the synthesis process where appropriate.

| Charting
As previously described, in this stage of the process we used the themes from articles which contributed the most data to the five research objectives to revise the initial framework allowing for the formation of preliminary subthemes which were then revised using an inductive approach.

| Synthesis: revision of initial framework
Two reviewers (LS, MN) independently identified eight papers from three of the prioritized studies 22,45,46 which contained the highest number of lines of first and second-order data which were relevant to multiple research objectives and represented the experiences/views of different participant groups. These three studies focused on views on the implementation of Mental Health Act (1983) in both England 22 and Scotland. 45,46 The themes from each of these three studies were grouped together under the relevant research objectives by one reviewer (LS), as illustrated by Table 4. This selection of themes and their allocation to an appropriate research objective was then checked by a second reviewer (MN). The data extracted into the first iteration of the framework was reallocated under the new themes, with color codes used to differentiate between different participant groups, see Table 5 for an example. Studies from England and Scotland were considered within the same framework and any data which were not captured by the new themes was noted separately. Within each of the broad themes, similar or related ideas were grouped together to form preliminary descriptive subthemes.

| Synthesis: final framework revisions using thematic synthesis
During the third revision of the framework, we revised the structure of the descriptive themes and subthemes using an inductive, iterative process. These changes aimed to capture data from the included studies which had not yet been synthesized within the framework, which meant there was the opportunity to ensure data from studies conducted within Northern Ireland was accounted for by the final framework. The content contained within some subthemes changed in response to a new structure, with the names and placement of subthemes being changed where appropriate to reflect new content. Some theme names were changed to better reflect the content of their subthemes. Feedback on emerging subthemes and themes was sought from the review team (MN, SB, RA, JTC) throughout the synthesis process to ensure the names of developing subthemes and themes accurately reflected their content. Figure 2 summarizes our process of revising the synthesis framework during the review. 2

| MAPPING AND INTERPRETATION (SYNTHESIS, INTERPRETING/ COMMUNICATION)
This process corresponds to the Interpretation and Communication stage within systematic reviews described by Gough et al. 9 The findings of the review were considered in light of the original research aims and questions, and presented in a way which was intended to be accessible to the policy maker and other potential target audiences. 10

| Synthesis: presentation of findings
Following the finalization of the framework, we created a table that outlined which themes and subthemes contributed toward each of the five research objectives. The table F I G U R E 2 Summary of framework synthesis process also presents the participant viewpoint (eg, service user, carer, Nearest Relative or professional), see Table 6 for an example. The full table can be viewed within the main project report. 20 This table allowed us to visually depict how our final themes/subthemes were founded on the primary data contained within our prioritized articles. It also facilitated the narrative write-up of our four themes and their associated subthemes, again grouped under each of our research objectives.
The narrative write up of the themes and subthemes incorporated a summary at the end of each section, which made use of accessible language and bullet points. This was intended to support policy makers, and other readers, to quickly access the main points arising from the synthesis. Publication of the main report was also accompanied by the release of a Briefing paper, a four page document comprising a one page summary of the background, methods, results and recommendations arising from the review, followed by three more pages providing further detail. This document was intended to provide an accessible overview of the systematic review to commissioners, policy makers and clinicians with an interest in the topic area.

| Interpretation and communication: relationship between the themes
We developed a model depicting possible relationships between the themes and subthemes. This model considered overlaps in the data within the final framework, where the first or second order data from the prioritized articles could potentially be considered under multiple subthemes or where the concepts within one subtheme could potentially influence those within another, as illustrated by Figure 3. 3 Feedback on the final synthesis was sought from one carer with experiences of the Nearest Relative provision of the Mental Health Act (1983) to explore how the content of the synthesis reflected their own experiences. The carer was an individual known to the research team, who had prior experience of conducting systematic reviews and knowledge of qualitative techniques. Their existing relationship with the research team and prior knowledge of qualitative systematic review techniques made it easier for the review team to engage with them over such a confined time period. Their feedback largely corroborated the preliminary themes from the framework synthesis and clarified key issues in the write up of findings within the main report.

| DISCUSSION
Our "Rapid Best-Fit" approach to framework synthesis allowed us to conduct a qualitative evidence synthesis exploring the views and experiences of service users, carers and professionals of the Nearest Relative role of the Mental Health Act (1983) within 6 weeks. Within the context of a rapid review, it was not possible to construct a new conceptual model to structure the synthesis and we were unable to identify a suitable existing model or theory. We, therefore, based our initial framework upon the five research objectives that were co-produced with the policy customer. This framework was then revised using themes from articles which contained the most lines of first and/or second order data in terms of addressing the research objectives and representing views of different iterations of the Mental Health Act (1983) across England and Scotland. The final stage of the framework synthesis involved re-examination of the content of each existing theme and subtheme using thematic synthesis and revising our third-order interpretation where necessary, to reflect the ideas and concepts contributing to them. This process resulted in four themes grouped according to our initial research objectives. Three of these themes were descriptive in nature, being based directly upon our research objectives, with their associated subthemes providing more focused detail and interpretation of the first and second-order data represented within. The inductive thematic synthesis conducted as part of stage 3 of the development of our framework was particularly applicable to the themes based upon our final two research objectives: "Explore issues related to care during detention and after discharge" and "Explore issues related to service users having access to support from carers." We felt that converting these two objectives into descriptive themes would not be very useful for our intended audience. Hence, the two themes relevant to these research objectives, "Enabling use of the Nearest Relative/Named Person role" and "Importance of maintaining relationships," are more interpretative than the first two themes, while still being directly informed by our research objectives.
As far as we are aware, this qualitative evidence synthesis represents the first use of a framework synthesis where the initial framework was based directly upon the research objectives identified through consultation with the commissioners of our research. Existing framework synthesis approaches use either a framework developed through a combination of understanding of background literature, stakeholder input and/or researcher knowledge (eg, 9 ) or adapting an existing model or theory (eg, 12 ), while the initial stage of our framework was based directly on the needs of our commissioners.

| Methodological strengths and limitations of this review
A full thematic synthesis would have been the authors' preferred approach to synthesis to ensure a full exploration of people's experiences relating to this topic. The initial deductive stage of framework synthesis could have potentially limited the extent to which this could be carried out. Despite the interpretation used within stage 3 of refining the framework, the results of the review are descriptive in nature and it was not possible to include all of the extracted and coded data from the prioritized papers in the framework in the narrative write-up of the synthesis.
However, the "Rapid Best-Fit" approach allowed us to synthesize a high volume of qualitative data in a way which addressed our research objectives, within a limited time-frame. Structuring our initial framework according to the research objectives ensured all data relevant to our research aim was captured during the early stages of data extraction, which expedited the synthesis. This approach also meant that a focus on the overall aims and context of our review could be maintained through subsequent revisions of the framework, ensuring our final themes and subthemes were directly related to the requirements of the policy customer.
The initial framework allowed us to rank the papers prioritized for further synthesis according to the quantity of data they contained relevant to each of the research objectives. We recognize that this approach potentially risked prioritizing lengthy reports containing more descriptive data and potentially overlooking studies with smaller amounts of more interpretative data which could still be useful in informing the synthesis. However, it did enabled us to select the themes to revise the framework in a systematic manner and begin the process of tailoring the framework to reflect both the content of the articles within the synthesis as well as the research objectives. This approach also ensured the framework was able to capture the experiences of the Nearest Relative role across different iterations of the Mental Health Act (1983) in England and Scotland, although no studies carried out in Northern Ireland were prioritized at this stage due to the themes and subthemes contained within these articles not being directly applicable to the research objectives of this review. This meant that themes from articles exploring the views of individuals with regard to the Mental Health (Northern Ireland) Order (1986) 49 were not incorporated into the second iteration of the framework, which could have meant the views of this individual group were masked or overlooked within a framework based upon the experiences of individuals of legislation from England and Scotland. However, the thematic synthesis conducted during the third revision of the framework provided the opportunity to identify material which was incongruent with existing themes and subthemes and allow researchers to incorporate views on the Mental Health (Northern Ireland) Order (1986). 49 We could have used existing policy and/or legal documents, such as The Mental Health Act Code of Practice 2015 to create an initial framework. 50 However, this may have biased the framework toward what some people think should happen instead of what actually happens in practice and therefore not representing the views and experiences of those concerned. The lead author (LS) led the framework synthesis, drawing on their prior experience as a qualitative reviewer and qualified Clinical Psychologist to interpret the data and organize it into themes. The potential for bias was reduced through the deductive and descriptive nature of the initial stages of the synthesis, the checking of extracted data by a second reviewer (MN) and involving other members of the team during the process of revising the initial framework within the synthesis stage of the review.

| FURTHER REFLECTIONS
Within the context of a rapid review in response to an urgent policy need, where qualitative research is required to answer the research question, we found the rapid best fit approach a helpful method to use to ensure our findings were in line with the priorities of our stakeholders. In our opinion, this approach could be suited for other topics, where a similar quantity and quality of qualitative evidence is available for synthesis.
Our research priorities were relatively straightforward to identify as we only had one main point of contact who represented the DHSC and was familiar with their priorities. However, in a review with multiple stakeholders identifying clear research aims and/or objectives within the context of a rapid review may be challenging and potentially lead to misunderstanding or important information being overlooked while developing the protocol. In this instance, information relevant to the research aim may be inadvertently excluded from the review unless the study authors choose to document a deviation from protocol. This in itself is not necessarily problematic, providing this deviation is documented clearly and applied systematically throughout the review process. From a practical perspective, however, it may not be possible to apply any changes to protocol retrospectively as these changes may require additional time and resources which may not be available when conducting a rapid review. In which case, any relevant but overlooked information would have to be acknowledged within the limitations section of the review and/or highlighted within an appendix.
In our experience, it is vital to ensure that an adequate period is dedicated to the scoping of the review topic and that key stakeholders are fully involved with this process and provided with the opportunity to approve the final protocol, before title and abstract screening commences. Within this review, using this approach ensured that our methods remained consistent with our research objectives, something which may otherwise be challenging while undertaking a fast-paced rapid review.
Some researchers may regard the use of qualitative synthesis within rapid reviews as inappropriate, as the constricted time-frame prevents full immersion within the primary data, potentially resulting in a more aggregative, descriptive synthesis. For the generation of new theory/insight, an entirely inductive approach such as metaethnography or thematic synthesis would be more appropriate. 51 While opportunities for some inductive synthesis was incorporated within the final stage of this technique, the initial stages of the synthesis were constrained somewhat by the priorities of our stakeholders and the themes selected to revise our initial framework. While a full thematic synthesis would not have been possible within the limited timeframe for our review, we wonder what new understanding would be gained through synthesis of the data from our included studies using solely inductive methods.
The initial stages of the rapid best-fit framework approach requiring the use of deductive approach using a framework developed a priori could potentially be somewhat easier to utilize by researchers with limited experience of qualitative evidence synthesis. However, it is possible that the naïve researcher may find it more challenging to conduct the thematic synthesis within the latter stages of this approach, which requires additional interpretation and identification of possible relationships across different themes without additional support from a person with experience of qualitative evidence synthesis, especially within a constrained timeframe.

| CONCLUSIONS
In the absence of an initial theory-informed framework, the "Rapid Best-Fit" method developed and applied within this review allowed us to organize the data from our included papers into four themes which reflected both our research objectives and the views and experiences of the individuals within the primary studies. This outcome is consistent with those emerging from other methods of framework synthesis, although further examples of using this approach are needed to determine if this method can be replicated across different research areas. Overall, the close collaboration between the research team and policy customer while defining our research objectives ensured that the "Rapid Best-Fit" approach allowed for the incorporation of qualitative data within fast-paced decision making and ensured the final output was useful and relevant to the needs of the commissioners of the review.