gms | German Medical Science

In the production of a systematic review or HTA, reivew authors have categorized included studies according to user-defined PICO (or SPICE, SPIDER, etc) ontologies, categorized excluded studies according to exclusion criteria, extracted data from a small amount of studies, and critically appraised these same studies. All of this “assessment data” can theoretically be used by future reviewers, rather than future reviewers having to redo this work. With the exception of clinical trials databases, most studies do not submit this data in any structured way.

Our IT department funds a “Knowledge Digitalization” portfolio, under the auspices of which we have started a project to purchase a software solution to facilitate storage, sharing, and re-use of data during the review process. If we are successful, the project will result in the design and purchase of a solution that will allow researchers in different organizations, and who have used different review software or other tools, to re-use each other’s PICO categorizations, extracted data, and even critical appraisal.

Anticipated gains are increased production and improved quality of reviews, an estimated 32-55 hours’ saved per review, and further support of FAIR principles and contribution to a shared good.

We will discuss the transformation of this project from a simple procurement of a repository, to designing specifications for an entirely new software. We approached the procurement process creatively by first inviting fifteen organizations (including HTA organizations, research groups, and private AI and software vendors) to help us improve our product specifications. Their input was vital and pointed out both possibilities as well as unrealistic expectations, particularly regarding automatic extraction of unstructured data.

We will conclude with a demonstration of the solution, our initial experiences in implementation, and costs and benefits.

Keywords: metadata, data-sharing, machine learning, research waste, FAIR principles