Menu Back to Poster-Presentations-Details

P238: Optimization of Content Retrieval for Medical Information Scientific Responses by the Addition of Medical Conference Evidence

Poster Presenter

Joseph Laudano

Medical Director
IQVIA
United States

Objectives

To identify and quantify potentially useful evidence for the development of scientific response documents (SRDs) from a scientific conference publication database not presently available from PubMed (PM) alone.

Method

A common medical inquiry from HCPs was selected for the test case. A medical conference database (MCD) was used as the test corpus for evidence retrieval. A Boolean search strategy was applied to the MCD with a content gap analysis (MCD vs PubMed) performed for relevant retrieved abstracts.

Results

We tested the hypothesis that there is additional and potentially useful evidence available from medical conferences for the development of SRDs not presently retrievable from PubMed alone. A typical medical information inquiry sought by healthcare professionals, “information on long-term safety and efficacy” 1, was selected as the test query. The query was converted to a Boolean search expression and applied to the test corpus which consists of over 26MM medical conference and PubMed abstracts, for the period January 1, 2019 through March 1, 2024. The subjects of inquiry for the study were all 37 newly approved drugs for the year 2022. An overall total of 10,582 scientific documents were retrieved with product mentions from the Data Lake (6,568 from scientific conferences and 4,014 from PubMed). For the selected medical inquiry applied to the MCD Data base, a total of 529 scientific documents were retrieved with 26 conference abstracts removed after being identified as encore presentations, leaving a total of 503 abstracts (347 from medical conferences and 156 from PubMed). The mean number of documents retrieved from the conference database for all 37 products was 9.38, range 0-58, median 29.0. From PubMed, a mean number of 4.22 documents were retrieved with a range 0-31, median 15.5. Among the 37 products reviewed, abrocitinib had the most documents retrieved from both the IM and PubMed databases (58 and 31 respectively). Abstracts from conferences scored higher (74.2) than those from PubMed (50.6) on relevance rank based on Term frequency-inverse document frequency (TF-IDF)/Jaccard index. Five products, oteseconazole, eflapegrastim, hyperpolarized Xe-129, nivolumab and relatlimab-rmbw, olutasidenib and vonoprazan/amoxicillin/clarithromycin had no documents retrieved from either database for the query.

Conclusion

Medical Information Professionals rely heavily on the ability to identify the most up-to-date, relevant and reputable evidence in order to effectively provide accurate and unbiased responses to healthcare professionals’ inquiries. Providing access to the latest abstract data presented at medical conferences not yet found in PubMed or other databases may afford Medical Information Professionals a more complete and up-to-date source of evidence for the development of SRDs. In this study, for the selected medical information inquiry and pharmaceutical products, we identified relevant and useful evidence which had been presented at the conference level but was not found in PubMed. Although limited to our selected inquiry topic and product subjects, this finding suggests that healthcare and research professionals may be limited to a subset of scientific evidence available to Medical Information Professionals, potentially impacting important healthcare decision making and research. Further investigation is necessary to determine the importance and significance of using conference evidence in this regard on a larger scale and, short of the time to appear in print, the reasons why they may or may not be submitted for eventual publication. Studies using additional medical information inquiries and product subjects are warranted.