@inproceedings{128dff9168da4dfe8fe18e3286ed4c53,
title = "Associations Between Aggregate NLP-Extracted Conflicts of Interest and Adverse Events by Drug Product",
abstract = "This study evaluates associations between aggregate conflicts of interest (COI) and drug safety. We used a machine-learning system to extract and classify COI from PubMed-indexed disclosure statements. Individual conflicts were classified as Type 1 (personal fees, travel, board memberships, and non-financial support), Type 2 (grants and research support), or Type 3 (stock ownership and industry employment). COI were aggregated by type compared to adverse events by product. Type 1 COI are associated with a 1.1-1.8% increase in the number of adverse events, serious events, hospitalizations, and deaths. Type 2 COI are associated with a 1.7-2% decrease in adverse events across severity levels. Type 3 COI are associated with an approximately 1% increase in adverse events, serious events, and hospitalizations, but have no significant association with adverse events resulting in death. The findings suggest that COI policies might be adapted to account the relative risks of different types of financial relationships.",
keywords = "Conflict of Interest, Health Policy, Natural Language Processing",
author = "Graham, {S. Scott} and Majdik, {Zoltan P.} and Barbour, {Johua B.} and Rousseau, {Justin F.}",
note = "Publisher Copyright: {\textcopyright} 2022 International Medical Informatics Association (IMIA) and IOS Press.; 18th World Congress on Medical and Health Informatics: One World, One Health - Global Partnership for Digital Innovation, MEDINFO 2021 ; Conference date: 02-10-2021 Through 04-10-2021",
year = "2022",
month = jun,
day = "6",
doi = "10.3233/SHTI220106",
language = "English (US)",
series = "Studies in Health Technology and Informatics",
publisher = "IOS Press BV",
pages = "405--409",
editor = "Paula Otero and Philip Scott and Martin, {Susan Z.} and Elaine Huesing",
booktitle = "MEDINFO 2021",
}