Voices of experience

Background Journals in health sciences increasingly require or recommend that authors deposit the data from their research in open repositories. The rationale for publicly available data is well understood but many researchers lack the time, knowledge, and skills to do it well, if at all. There are few descriptions of the pragmatic process a researcher author undertakes to complete the open data deposit in the literature. When the author’s manuscript for a mixed methods study was accepted by a journal that required shared data as condition of publication, she proceeded to comply despite uncertainty with the process. Purpose The purpose of this study is to describe the experience of an information science researcher and first-time data depositor to complete an open data deposit. The study illustrates the questions encountered and choices made in the process. Methods To begin the data deposit process, the author found guidance from the accepting journal’s policy and rationale for its shared data requirement. A checklist of pragmatic steps from an open repository provide a framework that the author used to outline and organize the process. Process steps included organizing data files, preparing documentation, determining rights and licensing, and determining sharing and permissions. Choices and decisions included which data versions to share, how much data to share, repository choice, and file naming. Processes and decisions varied between the quantitative and qualitative data prepared. Results The author deposited data in two datasets and documentation for each in Figshare open repository, thus meeting the journal policy requirements to deposit sufficient data and documentation to replicate the results reported in the journal article and also meeting the publication deadline to include a Data Availability Statement with the published article. Conclusion This experience illustrated some practical data sharing issues faced by a librarian author seeking to comply with a journal data sharing policy requirement for publication of an accepted manuscript. Both novice data depositors and data librarians VOICES OF EXPERIENCE Hypothesis Vol. 32 No.1 Fall/Winter 2020 may find this individual experience useful for their own work and the advice they give to others. BACKGROUND Journals in health sciences increasingly require or recommend that authors deposit the data from their research in open repositories. In October 2019, the Journal of the Medical Library Association (JMLA) announced its policy to require authors to deposit deidentified research data for all original investigations and case report articles accepted for publication in October of 2019 [1]. Although many library and information science journals recommend data sharing, JMLA was the first and still possibly the only one to require it. The rationale for publicly available data—that it fosters scientific progress and enables replication and reproducibility of research—is well described in the literature and in the JMLA editorial justifying its requirement [2]. Data sharing workflows for researchers and for repositories are described at a conceptual level by Austin et al [3] with little process detail. Educational methods for teaching research data management methods to students and researchers are described in a review by Corti and Van den Eyden. [4]. They recommend skills training and active learning methods but do not cover specific content on how to share data and what data to share. While most researchers support data sharing and agree with its value, studies show that many researchers have uncertainty and concerns with the pragmatic aspects of data sharing [5-7]. A 2018 survey by Stuart et al of 7,000 researchers from various disciplines and experience levels found that 76% rated the importance of discoverable data highly, while 46% reported organizing data and presenting data as problematic. Lack of time was a problem for 35%, and repository choice a problem for 33% [8]. A survey of clinical trial investigators, by Tannenbaum et al, found that they spent a median of 18 hours (IQR 8 – 40) per data set shared [9]. A meta-synthesis of qualitative studies of researcher experience with data sharing found that “researchers lack time, resources and skills to effectively share their data in public repositories”[10]. I submitted a manuscript to JMLA after its data sharing policy was announced, but prior to when it went into effect. Technically I was not required to share the data for manuscript. But when my manuscript was accepted for the July 2020 issue, I decided to share the data from my research motivated by belief in the value of data sharing and by the journal requirement policy. Like other researchers, I was uncertain about how to organize and prepare the data, and was apprehensive about the time it would take, but I decided to do it. The accepted manuscript (now a JMLA article) is a mixed methods assessment of a clinical evidence technology based on a survey of 32 primary care providers (PCPs) who participated in an earlier randomized trial concerning their use (or non-use) VOICES OF EXPERIENCE Hypothesis Vol. 32 No.1 Fall/Winter 2020 of the technology during that trial, and interviews with 11 PCPs in the intervention arm [11]. While advantages and obstacles to researcher data sharing are well described, few studies describe researchers’ data sharing experiences in detail. The purpose of this case study is to describe my experience as an information science researcher and first-time data-depositor, and illustrate the practical questions and issues that I encountered in the process. METHODS To begin the open repository deposit process, I reviewed the journal’s requirements. The JMLA data sharing policy states ”The JMLA requires authors of Original Investigation, Case Report, and Special Paper articles to (1) place the deidentified data associated with the manuscript in a repository and (2) include a Data Availability Statement in the manuscript describing where and how the data can be accessed” [1]. The JMLA editorial announcing the policy by Akers et al elaborated on the requirements stating that “at least minimal data needed to support or replicate results”, and “documentation describing the contents of the data files” must be deposited [2]. A data availability statement, including a URL or DOI for the data, must be provided by the author and included with the published article. The editorial offered guidance for authors depositing data for the first time and recommended a selection of resources and references that could be consulted for help, including Library-based web-sites and LibGuides, and webpages of open repositories and noted that help is available from health science libraries and librarians [2]. I sought practical guidance that provided specific data sharing information and advice, and not data sharing background or rationale from among the editorial references. Three references that provided pragmatic methods for organization and presentation of data were Washington University Libraries’ webpage “Preparing data for deposit” [12], the 2019 Journal of eScience Librarianship Presentations article entitled “Best Practices for Data Sharing and Deposit for Librarian Authors”, by Regina Raboin et al [13] , and the Inter-University Consortium for Political and Social Research (ICPSR) Guide which includes a section entitled “Preparing data for sharing” [14]. I found clear steps to follow in the Digital Research Materials Repository (DRMR) Deposit website created by the Washington University St. Louis Libraries (available at https://libguides.wustl.edu/drmr/dataprep) [12]. There, a checklist and a template README file form are available as downloadable and printable tools to address the organization, presentation, documentation and other tasks. The Checklist outlines four major steps and multiple itemized sub-steps. The WUSL Data Research Materials Repository (DRMR) checklist tasks are to organize your deposit, prepare documentation, determine deposit rights & licensing, and determine sharing and permissions. I used the DRMR checklist as a process framework VOICES OF EXPERIENCE Hypothesis Vol. 32 No.1 Fall/Winter 2020 to move forward with the data deposit but revised it adding “review journal policy and requirements” as first step described above and a fifth task, “choose your data repository”. The process was recursive in that I worked on several checklist steps at once or went back and forth between them rather than proceeding in a consecutive workflow. [See Figure 1] Figure 1: Author's process to deposit research data in an open repository to comply with journal policy. Step 1: Organize Your Deposit After reviewing the journal’s policy and guidelines, the first step, to “Organize your deposit” seemed simple, but it was on this step that I encountered the most decision points and spent the most time. Sub-items in this category included gathering and choosing the dataset files to share, naming files and labelling variables consistently, and transforming datasets and files to -non-proprietary file formats. The JMLA policy names appropriate data types as “including but not limited to spreadsheets, text files, interview recordings or transcripts, images, videos, output from statistical software, and computer code or scripts”[2]. I had several of these data types and felt uncertainty concerning which versions of similar data files and how much of the total available data to prepare. VOICES OF EXPERIENCE Hypothesis Vol. 32 No.1 Fall/Winter 2020 Step 1a: Data version Owing to the mixed methods design of the research, there were both quantitative and qualitative data files, and version was an issue in each type. For the quantitative data, potential data files included raw deidentified survey response data output from REDCap [15], and the same survey data cleaned of extraneous variables and processed for analysis from the Stata statistical software [16]. Raboi