Research Protocol
Draft. The eligibility criteria and limits can still change. The protocol will be registered on OSF Registries before screening begins, because PROSPERO does not accept scoping reviews. Registration link: to be added.
Background
This review maps the factors reported to influence the sharing and reuse of research data in health research, including barriers, facilitators, and governance mechanisms.
This review updates the 2019 review of research data sharing (Chawinga and Zinn 2019) with the scoping review methods of Cavallaro et al. (2026). The 2019 review shares this review’s focus on research data sharing, which makes it a strong candidate for an update. It looks at differences across countries. This review looks at differences across research domains, with special attention on healthcare.
Earlier reviews address parts of this topic. Reviews in health and medical research examine incentives (Rowhani-Farid et al. 2017), principles and norms (Kalkman et al. 2019), and governance mechanisms such as harmonized data access agreements (Saulnier et al. 2019). Reviews in public health examine barriers to public health data sharing but exclude research data sharing (van Panhuis et al. 2014).
This review compares health research data sharing with data sharing in general and in other fields. Beyond health, cross-disciplinary reviews cover what drives and inhibits researchers in sharing and reusing data (Fecher et al. 2015; Zuiderwijk et al. 2020; Perrier et al. 2020).
The taxonomy of factors this review produces serves the CITF-emails study, where it supports manual annotation for supervised machine learning on the emails and identifies factors that the emails cover only in part.
Objective and questions
The objective is to map the factors reported to influence the sharing and reuse of research data in health research, and to compare them with the factors reported for other fields and for public health data.
- Which factors (barriers, facilitators, and governance mechanisms) are reported to influence the sharing and reuse of research data in health research?
- How can these factors be organized into a taxonomy suitable for annotating correspondence about data sharing?
- How do the factors reported for health research differ from those reported for research in other fields and for public health data?
Methods
The review follows the scoping review framework of Arksey and O’Malley (2005) as refined by Levac et al. (2010) and the JBI guidance (Peters et al. 2020). The method draws on the protocol of Cavallaro et al. (2026), which pairs machine learning screening with automated deduplication. It improves on that protocol in deduplication and seed generation. The review is reported according to PRISMA-ScR (Tricco et al. 2018). The searches are reported according to PRISMA-S (Rethlefsen et al. 2021). The flow diagram adapts the PRISMA 2020 template (Page et al. 2021), with “sources of evidence” in place of “studies”.
Eligibility criteria
The criteria follow the population, concept, and context (PCC) structure that JBI recommends for scoping reviews.
| Element | Criterion |
|---|---|
| Population | No restriction. Sources may concern researchers, data stewards, institutions, or governance bodies. |
| Concept | Factors that influence the sharing and reuse of research data, including barriers, facilitators, and governance mechanisms. Research data are data collected for research purposes under ethics oversight. |
| Context, tier 1 | Health research. All source types are eligible. |
| Context, tier 2 | Research in other fields and cross-disciplinary research. Reviews and syntheses only. |
Sources are excluded when they concern only public health data or other health systems data, or only clinical care data. The rule for routine or clinical data reused for research is not yet set. Each source is assigned a tier at screening from its field.
The search covers 2015 to 2025. Language limits are not yet set. Any limit will be stated and justified, as PRISMA-S item 9 requires.
Information sources and search
Four queries run in Scopus, Web of Science, and PubMed for 2015 to 2025. All four build on Chawinga and Zinn (2019). The base RDS query (Q1) takes their core term, “research data sharing”. Three concept queries (Q2 to Q4) extend that term with the vocabulary the field uses around it. Each pairs its own key term with “research data” or “data sharing” using AND, so a record must name both the concept and the data practice in its title or abstract. The queries use no health terms. Each database applies its own document type filter. Scopus and Web of Science limit results to articles and reviews, and PubMed limits results to journal articles. The search strategy page lists the key terms, the strings, and the counts.
Citation searching of the included sources supplements the database searches. The final strings are tested against a list of known relevant sources. Every search is recorded on the search strategy page.
Selection of sources
Records are deduplicated with ASySD, which, unlike the Zotero workflow of Cavallaro et al. (2026), runs from a script and makes the step reproducible. The records are then screened on title and abstract in ASReview, and again on full text, against the screening guide. Exclusion reasons are recorded at full text. Counts at every stage feed the flow diagram.
ASReview ranks records by predicted relevance and starts from a seed sample. The seed is the base RDS query (Q1), which returned 308 records against 4,192 for the four queries combined (7.3%), modestly above the conventional 5%. A seed from a fixed keyword is reproducible. Cavallaro et al. (2026) select their seed sample ad hoc from the first 5% of deduplicated records.
Data charting
Each included source is charted once into the form defined in the codebook, with one row for the source and one row for each factor it reports. The form is piloted on a sample of sources and revised before full charting begins.
Synthesis
The review codes factors inductively. The resulting taxonomy is a primary output. Counts by tier, field, source type, and year describe the evidence base, and the factor categories are compared across fields.
Reporting and data
The checklists for PRISMA-ScR and PRISMA-S record where each item is reported. The flow diagram is generated from the screening counts. Search exports, screening decisions, charted data, and all code are in the repository.
Open decisions
- Language limits.
- The rule for routine or clinical data reused for research.
- The definition of research data, by purpose or by consent.
- The number of screeners, calibration on a sample, and the handling of disagreement.
- Critical appraisal of included sources, which PRISMA-ScR treats as optional.
- Whether any step uses a large language model. Any such step will be reported following the PRISMA 2020 guidance on automation tools and the RAISE guidance.