The new heoro.com database launched successfully at ISPOR DC last week. We’ve developed an ontology of the 4,250 PRO tools we identified from more than 22,200 PRO study abstracts in the database.
Creating a new ontology is time-consuming work. We based the disease ontology in heoro.com on MeSH terms from PubMed, the source of the 100,000-plus abstracts we identified from a systematic search from 2005. Our intervention ontology also draws on MeSH terms for treatments, combined with lists of pharmaceutical products licensed in a number of countries including the US and UK, and is updated as studies on new products are published.
With our patient-reported outcome (PRO) instrument ontology, though, we started from scratch:
First, we identified all the instruments measuring quality of life, utilities, and other patient, family, caregiver or clinician outcomes, experiences and views from the 3,915 abstracts we identified in our training set of 10,000 abstracts moderated by experts in clinical medicine and health economics and outcomes research (HEOR).
Then, we developed unique natural language processing software, with around 95% accuracy for identifying PRO study publications in the 90,000 unmoderated papers in the database. The software identified a further 18,339 abstracts that were likely to be PRO studies.
Finally, our experts supplemented and indexed the list of PRO instruments by manually going through all 18,339 abstracts. The tables below summarise what we found.