Ontologies are broadly used across healthcare and life sciences. As a healthcare data company covering pharma, from early clinical development to market access as well as healthcare delivery, IQVIA makes an extensive usage of ontologies: for data harmonization (e.g.: OMOP [1]), regulatory submission support (e.g.: CDISC [2]), to structure data through NLP pipelines and to drive analytics (e.g.: population definitions).
The "ontology architecture" of IQVIA can be considered made of three parts, as illustrated below (for a sample of geographies and domains). andrea.splendiani@iqvia.com (A. Splendiani); emiliano.reynares@iqvia.com (E. Reynares); anthony.reckard@iqvia.com (A. Reckard) 0000-0002-3201-9617 (A. Splendiani); 0000-0002-5109-3716 (E. Reynares)
At the bottom, IQVIA makes use of ontologies that are "standard": either public or provided by third parties, these are shared among different actors of the healthcare ecosystem and constitute the basis for interoperability. The breadth of operations of IQVIA consider ontologies (or reference data) from many different perspectives. For instance, products may make use of as diverse resources as ChEBI [3] (research), ATC [4] (therapeutic classification), RxNorm [5] (products) and more.
In addition to this, IQVIA collects and integrate reference data from its massive health record ingestion processes. This results in extended resources, such a regional name variation for products, or typical references in prescriptions, or even products for which standard nomenclatures are not present (e.g.: lab supplies).
It is interesting to note that while ontologies represent the standard "as proposed", this extensive collection results in a view of the standard "as used", including an understanding of what granularities exist, what is updated when and so on.
Finally, as an analytics company, IQVIA goes the "last mile" to complement current ontological resources to support analytics. This involves developing knowledge graphs that allow for navigation across ontology versions, even those that are over a decade old, to support the creation of longitudinal patient data. Additionally, custom ontologies are on development to mine patientreported outcomes or social media for sentiment analysis.
This includes for instance the development of knowledge graphs enabling the navigation across ontology versions (sometimes beyond a decade) to support patient's longitudinal data creation, or the development of custom ontologies to, for instance, mine patients reported outcome or social media for sentiment analysis.
With this contribution we intend to present our assets and expertise in the ontology space, and invite the interested parties to kick off a discussion on how to leverage such resources (e.g.: extensive observation of ontology use in the field) and advance the current state of the art.