The problem The reason much of the data in clinical studies remains underused is because data sets exist in closed silos and on separate servers. There are no established ways to provide comprehensive descriptions about data (meta data), which would make the research easier for others to discover. There is also a lack of widely used protocols for using data and protecting participants’ anonymity.

Prototyping for research and collaboration Our aim was to design, prototype, and test an end-to-end Vivli experience for two types of user:

• Requesters, who would use Vivli to search for studies and request access to individual patient data
• Submitters, who would use Vivli to add studies to the Vivli catalogue for requesters to discover and use

We used prototypes as stimulus material in expert research that we conducted with thought leaders in information sharing platforms in healthcare. We also used our prototypes with Vivli’s potential users working all over the world in various medical domains, such as: randomised clinical trials, evidence synthesis, clinical informatics, and secondary analysis. Our findings helped us improve the prototype and refine the Vivli experience. The prototypes were also used as a tool for strengthening the Vivli consortium and communicating Vivli’s vision to stakeholders and collaborators. The following video demonstrates the functionality of the Search and Request prototype.

Project learnings During our research we identified the challenges a data platform such as Vivli is going to face. Here are just a few of our findings that informed design decisions: How can we improve discoverability of clinical studies through a better search interface? Existing platforms offer limited search capabilities, restricted to a minimal selection of high-level search terms. Vivli will provide access to rich metadata annotation, allowing requesters to create rich search queries. We designed an interactive query tool augmented by Cochrane’s PICO ontology (Population, Intervention, Comparator, Outcome) to help researchers formulate clinical questions and contextualised queries.

We also found that researchers often waste time requesting data that turns out to be of low quality or lacking relevant variables. So we designed a data quality score and a list of high-level data schema that display alongside each query result, before a request is made.

How can an online platform incentivise researchers to share data? We found that academic researchers who conduct clinical trials find limited value in making their data available for secondary analysis. This is unless data sharing is rewarded in academia’s main currency — publication citations. To incentivise better data sharing amongst academics, we designed a mechanism for tracking data usage and citations on the submitter’s account.

How can we ensure that data curation is as fast, accurate and complete as possible? Searches for studies are only as good as the study description that is curated during study submission. We designed a curation process through which submitters use automatic extraction of metadata elements from public registries and the study protocol. This is combined with manual refinement of metadata annotation and aided with a preview of the curated study as seen by requesters in search results.

How can the user interface strengthen transparency of the data sharing experience? Because of the sensitive nature of clinical study data, sharing is subject to a review process, legal arrangements, and data processing that takes time and is handled by multiple stakeholders. This is another hurdle for requesters so we designed a range of mitigating features that improve the transparency of the review process. These include live visualisations of the request status, tailored wait estimations, and a simplified online data usage agreement.