PRISMA2020: An R package and Shiny app for producing PRISMA 2020‐compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis

Abstract Background Reporting standards, such as PRISMA aim to ensure that the methods and results of systematic reviews are described in sufficient detail to allow full transparency. Flow diagrams in evidence syntheses allow the reader to rapidly understand the core procedures used in a review and examine the attrition of irrelevant records throughout the review process. Recent research suggests that use of flow diagrams in systematic reviews is poor and of low quality and called for standardised templates to facilitate better reporting in flow diagrams. The increasing options for interactivity provided by the Internet gives us an opportunity to support easy‐to‐use evidence synthesis tools, and here we report on the development of a tool for the production of PRISMA 2020‐compliant systematic review flow diagrams. Methods and Findings We developed a free‐to‐use, Open Source R package and web‐based Shiny app to allow users to design PRISMA flow diagrams for their own systematic reviews. Our tool allows users to produce standardised visualisations that transparently document the methods and results of a systematic review process in a variety of formats. In addition, we provide the opportunity to produce interactive, web‐based flow diagrams (exported as HTML files), that allow readers to click on boxes of the diagram and navigate to further details on methods, results or data files. We provide an interactive example here; https://prisma-flowdiagram.github.io/. Conclusions We have developed a user‐friendly tool for producing PRISMA 2020‐compliant flow diagrams for users with coding experience and, importantly, for users without prior experience in coding by making use of Shiny (https://estech.shinyapps.io/prisma_flowdiagram/). This free‐to‐use tool will make it easier to produce clear and PRISMA 2020‐compliant systematic review flow diagrams. Significantly, users can also produce interactive flow diagrams for the first time, allowing readers of their reviews to smoothly and swiftly explore and navigate to further details of the methods and results of a review. We believe this tool will increase use of PRISMA flow diagrams, improve the compliance and quality of flow diagrams, and facilitate strong science communication of the methods and results of systematic reviews by making use of interactivity. We encourage the systematic review community to make use of the tool, and provide feedback to streamline and improve their usability and efficiency.

making use of interactivity. We encourage the systematic review community to make use of the tool, and provide feedback to streamline and improve their usability and efficiency.
1 | INTRODUCTION 1.1 | Evidence synthesis reporting standards Evidence syntheses (e.g., systematic reviews and evidence maps) typically aim to reliably synthesise an evidence base, and are based on state-of-the-art methodologies designed to maximise comprehensiveness (or representativeness), procedural objectivity, and reproducibility, whilst minimising subjectivity and risk of bias (Collaboration for Environmental Evidence, 2018;Higgins et al., 2019). Reproducibility is made possible through a high degree of transparency when reporting the planned or final methods used in a review protocol or final report.
Reporting standards, such as PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses [Moher et al., 2009]) and ROSES (RepOrting standards for Systematic Evidence Syntheses [Haddaway, Macura, et al., 2018]), aim to ensure that review methods and findings are described in sufficient detail.
In 2009, the PRISMA statement-a reporting guideline designed primarily for systematic reviews of health interventions-was released Moher et al., 2009). The guideline was developed by a consortium of systematic reviewers, methodologists and journal editors to address evidence of incomplete reporting in systematic reviews (Moher et al., 2007), with recommendations formed largely based on expert consensus obtained via Delphi surveys and consensus meetings. The PRISMA statement has been widely endorsed and adopted by journals, and evidence suggests use of the guideline is associated with more complete reporting of systematic reviews (Page et al., 2016). However, to address the many innovations in methods for systematic reviews, changes in terminology, and new options to disseminate research evidence that have occurred since 2009, an update to the guideline (referred to now as PRISMA 2020 [Page et al., 2021]) has recently occurred.

| Review flow diagrams
Flow diagrams in evidence syntheses allow the reader to rapidly understand the core procedures used in a review and examine the attrition of irrelevant records throughout the review process. The PRISMA flow diagram, published in 2009, is a static infographic describes the sources, numbers and fates of all identified and screened records in a review (for more details, see the original flow diagram (Moher et al., 2009) and an update from 2014 [Stovold et al., 2014]). A recent assessment of the quality and use of flow diagrams in systematic reviews found that only 50% of identified reviews made use of flow diagrams, with their quality generally being low and not significantly improving over time (quality defined by the presence of critical data on the flow of studies through a review [Vu-Ngoc et al., 2018]): as a result, the authors called for a standardised flow diagram template to improve reporting quality.
Several changes were made to the original PRISMA flow diagram in the 2020 update (Page et al., 2021). The 2020 template: (i) recommends authors specify how many records were excluded before screening (e.g., because they were duplicate records that were removed, or marked as ineligible by automation tools); (ii) recommends authors specify how many full text reports were sought for retrieval and how many were not retrieved; (iii) gives authors the option to specify how many studies and reports included in a previous version of the review were carried over into the latest iteration of the review (if an updated review); and (iv) gives authors the option to illustrate the flow of records through the review as separated by type of source (e.g., bibliographic databases, websites, organisation and citation searching). Also, the phrase 'studies included qualitative synthesis' has been replaced with 'studies included in review', given the former phrase has been incorrectly interpreted by some users as referring to syntheses of qualitative data. Furthermore, the recommendation to report in the flow diagram the number of studies included in quantitative synthesis (e.g., metaanalysis) has been removed, given a systematic review typically includes many quantitative syntheses, and the number of studies included in each varies (e.g., one meta-analysis might include 12 studies, another might include five).

| Transparency and Open Science in evidence syntheses
Broadly speaking, the Open Science movement aims to promote research integrity, experimental and analytical repeatability and full transparency, from project inception to publication and communication. Various definitions and frameworks for Open Science have been proposed (e.g., Open Data, Open Methods, Open Access, Open Source proposed by Kraker et al., 2011, and44 components by Pontika &Knoth, 2015).
The application of Open Science principles to evidence synthesis has been explored by Haddaway (2018), defined as Open Synthesis: the concept has since been expanded to cover 10 proposed components (Open Synthesis Working Group 2020). Open Synthesis is important and beneficial for a number of key reasons (Haddaway, 2018): (1) there is a need to be able to access and verify methods used in reviews and allow interrogation of the fate of each record in the review process; (2) to reduce research waste, data between different elements of a review (e.g., the methods or the results for any one step) may help to facilitate the validation and assessment of systematic reviews and make it far easier to access and reuse their methods, data and code. Such interactivity could be achieved through hyperlinking within static digital files, such as PDF (portable document format) files, or through web-based visualisations that would facilitate updating or 'living reviews' (Blank & Reisdorf, 2012).

Furthermore, by embedding and nesting relevant information
behind an interactive visualisation such as a flow diagram, review authors could make use of a key concept in science communication: that of simplification. Simplification is a key principle in audio-visual science communication (Elliott et al., 2017) and relies on prioritisation of information rather than 'dumbing down' (Finkler & Leon, 2019).
Extensive detail on the methods employed and on the reporting of information sources, data inputs and outputs could be accessed via hyperlinks, with core information placed front-and-centre. This

| Objectives
This project had the following aims: 1) to develop a novel package for the R programming environment (R Core Team, 2018) for producing systematic review flow diagrams that conform to the latest update of the PRISMA statement (Page et al., 2021); 2) to adapt this code and publish a free-to-use, web-based tool (a Shiny app) for producing publication-quality flow diagram figures without any necessary prior coding experience; 3) to allow users to produce interactive versions of the flow diagrams that include hyperlinks to specific web pages, files or document sections.
The project was produced collaboratively as part of the Evidence

| RESULTS
In the following pages, we summarise the functionality of the R package (requiring knowledge of coding in R) and Shiny app (a pointand-click web app with no need for coding), providing a summary in lay terms, along with a more detailed description for the code-savvy ('Code detail' boxes). Functions are indicated by courier font, whilst packages are indicated by italics.
3.1 | The PRISMA2020 R package-For users with coding experience

| Data import and cleaning
The data needed for the PRISMA_flowdiagram() function can be entered either directly as a set of numbers or R objects, but data upload can be facilitated by using a template comma separated value (CSV) file (see Table 1). We recommend the use of a CSV file as opposed to manually inputting numbers, as this allows for better reproducibility/transparency, as the underlying CSV can be shared.
This file can be edited to a large extent and the edits incorporated into the text, numbers, hyperlinks and tooltips used to make the plot. Users can alter the following columns: 'boxtext' (the text that appears in the flow diagram), 'tooltips' (the text that appears as the mouse hovers over the box), 'url' (the url to which the hyperlink leads on clicking), and 'n' (the number appearing in the flow diagram).
The function PRISMA_read() reads in a template CSV file containing data to display in the flow diagram, including text contents, quantitative data (i.e., the number of records in each box), tooltips (i.e., the text that appears when the mouse hovers over a box), and hyperlinks for 'on click' functionality. The output is a list of named objects that can be read directly into PRISMA_flowdiagram(). The function PRISMA_flowdiagram() produces a PRISMA 2020style flow diagram for systematic reviews. In summary, boxes are placed at specific locations across the graph, and they are automatically connected with arrows according to a specified set of connections.
Code detail: PRISMA_flowdiagram() uses the grViz() function from the DiagrammeR package (Iannone & Iannone, 2020) to plot a DOT graphic using layout = neato to explicitly place 'nodes' (boxes) at a particular location and splines='ortho' to specify axis-aligned edges are drawn between nodes. The label (including data) and tooltip for each node is read in within the main text of the function by using paste() to combine DOT strings and R objects.
Along with the text, data, tooltips and hyperlinks, users can specify whether to plot the 'previous studies' arm or the 'other studies' arm of the flow diagram by specifying these options within the PRISMA_flowdiagram() function.
In addition, the font, box fill, box line colour and line arrow head/ tail can be altered as desired.
Code detail: Since text rotation is not supported in DOT or DiagrammeR, the vertical labels for the left-hand blue bars are added via JavaScript appending using the appendContent() and onStaticRenderComplete() functions from the htmlwidgets package (Vaidyanathan et al., 2018) to append a block of JavaScript to the HTML output.

| Saving the output as a file
The PRISMA_save() function allows for the flow diagram to be saved as an HTML file (with interactivity preserved, either as a standalone file with scripts and styles embedded, or as a ZIP archive in which these are stored separately), or as a PDF, PNG, SVG, PS or WEBP file (without interactivity). This function takes the plot produced by PRISMA_flowdiagram() and saves the file. A default option for filename is provided, but this can be overridden along with the filetype which is calculated from the file extension by default.
Code detail: When saving as HTML, the PRISMA_save () function uses the savewidget() function from the htmlwidgets package (Vaidyanathan et al., 2018). When saving as other formats, the internal function PRISMA_-gen_tmp_svg() is used, this first uses savewidget () to create an HTML file in a temporary directory and then uses the various XML manipulation functions from the xml2 package (Wickham et al., 2018) to step through the HTML, using xpath (Dyck et al., 2014) to find the SVG embedded within the HTML.
As JavaScript is not supported in SVG files, the xml2 package is again used to add a rotate transformation and programmatically alter the x and y coordinates to create the blue vertical labels. Following this, the temporary SVG is either copied to its final destination, or the rsvg (Ooms, 2017) package is used to convert it into the desired output format.

| The Shiny app-A web-based application for all
Shiny is a package within the R environment that allows users to construct standalone web-based applications based on R functions (Beeley, 2013). The 'app' can be interacted with by entering data, running functions with user-specified settings to plot figures, and downloading the resultant figures in a variety of formats.
The PRISMA2020 Shiny app is available free-of-charge The createLabels() function is called just before the plot is re-rendered, this registers a MutationObserver that waits for the nodes to be created. Once the nodes are visible to the DOM (Document Object Model, a programming interface for HTML), the renderLabel () function is called, once for each node, to add the labels.

| Interactivity
The interactivity here represents an additional step to cross link and host the interactive PRISMA flow diagram with the relevant texts and data. As described above, such interactivity allows for radically improved transparency and may be of particular value for highly complex review methods. Including interactivity obviously corresponds to additional effort on the behalf of review authors, but has clear benefits for transparency and communication. Interactivity is provided by both the R package and the Shiny app.
Interactivity in the PRISMA2020 flow diagrams is provided in two ways. First, mouse-over tooltips appear as the user's mouse is moved over a particular box. These popup boxes can contain userspecified text providing more information as desired. For example, a short elaboration of the numbers of text in each box to clarify meanings. Alternatively, tooltips can provide an explanation of the information that will be hyperlinked to on clicking. Second, the boxes can be given hyperlinks so that the user can follow a predetermined link. These links can be anchors within a document or webpage, or datafiles or web pages stored on external or local repositories (e.g., supplementary files on a data repository such as figshare or Zenodo).

| Case study
We have prepared a case study that demonstrates possible interactivity that can be employed in a web-based PRISMA2020 flow diagram (see Figure 4). The example website is available at https:// prisma-flowdiagram.github.io/.
The website is based on data from an ongoing systematic review into ambulance clinician responses to adult male victims of intimate partner violence (Mackay et al., 2021). The site uses a flowchart generated from this software, alongside bootstrap (https:// getbootstrap.com) to make a fully interactive experience, enabling users to interrogate various aspects of the review. As the review is currently underway, the site will be updated as the review progresses.

| DISCUSSION
The PRISMA 2020 update represents a significant development of the PRISMA statement, increasing the usability and level of detail needed in systematic reviews. The PRISMA2020 flow diagram similarly provides a clearer and more detailed template. To our knowledge, our tool is the only one catering for the PRISMA 2020 update. Previous tools offer far less flexibility and usability, for example, those provided by the theta Collaborative (http://prisma. thetacollaborative.ca/), Jack Wasey's R package PRISMAstatement (https://rdrr.io/cran/PRISMAstatement/), and 'plot_PRISMA()' from the metagear R package (https://rdrr.io/cran/metagear/).
We have developed a user-friendly tool for producing PRISMA 2020-compliant flow diagrams for users with coding experience and, importantly, for users without prior experience in coding by making use of Shiny (https://estech.shinyapps.io/prisma_ flowdiagram/). This free-to-use tool will make it easier to produce clear and PRISMA 2020-compliant systematic review flow diagrams.
Significantly, users can also produce interactive flow diagrams for the first time, allowing readers of their reviews to smoothly and swiftly explore and navigate to further details of the methods and results of a review.
In addition, the ability to produce flow diagrams using code in a data-driven approach carries with it a number of benefits, including: facilitating Open Science (specifically Open Code); reducing the risk of transcription errors; and, opening up possibilities for reproducible documents such as executable research articles (Tsang & Maciocci, 2020) and communicating the results of living systematic reviews (Elliott et al., 2017).
We believe this tool will increase use of PRISMA flow diagrams, improve the compliance and quality of flow diagrams, and facilitate strong science communication of the methods and results of systematic reviews by making use of interactivity. We encourage the systematic review community to make use of the tool, and provide feedback to streamline and improve usability and efficiency.

CONFLICTS OF INTEREST
Matthew Page co-led the development of the PRISMA 2020 statement and Luke McGuinness is a coauthor of the PRISMA 2020 statement, but they have no commercial interest in the use of this reporting guideline.