Objective: Pharmaceutical documents such as the common technical document, package inserts (PIs), and interview forms (IFs) are available at the website of the Pharmaceuticals and Medical Devices Agency. However, because these documents were created with an emphasis on human readability in paper form, it is difficult to use the information included and interoperate these documents with computers. Using IFs, we will investigate how to structure pharmaceutical documents in the AI era to achieve both human and machine readability.
Design/Methods: The IFs of arbitrary selected ten drugs were structured into Resource Description Framework (RDF) according to the Drug Interview Form Description Guidelines 2018 (updated version in 2019). The data were manually extracted from the IFs and entered into a spreadsheet before being converted to RDF by a written script. The PIs were converted to RDF in addition to the IFs. To examine the linkage with external databases, IDs in ChEMBL, which is a manually curated database of bioactive molecules with drug-like properties, were embedded in the RDF.
Results: We demonstrated that the conversion of IFs and PIs into RDF makes it possible to easily retrieve the corresponding part of the PIs cited in the IFs. Furthermore, we quickly obtained the relevant data from ChEMBL, demonstrating the feasibility of linking IFs with an external database. Our attempt to RDFization of IFs is expected to encourage the development of web applications for healthcare professionals and the development of datasets for AI development.
Conclusion: We could easily interoperate IFs with other pharmaceutical documents and an external database by converting IFs into RDF following the description guidelines. However, problems such as how to deal with items that were not described in the description guidelines were indicated. We hope that discussions will grow based on this effort and that related industries will move toward accomplishing effective use of these documents.
