To truly combat traditional trial inefficiencies, including time-consuming and costly patient recruitment efforts, the industry needs NLP and other AI solutions to extract valuable information from unstructured data like patient records,” said Infolytx Chief Technology Officer Dr. Zunaid Kazi.
NEW YORK (PRWEB) November 05, 2018
Infolytx, an AI and machine learning company, was asked to present its Natural Language Processing or NLP-based Electronic Health Records (EHR) Information Extraction framework solution to improve patient identification for clinical trials at the 2018 American Medical Informatics Association (AMIA) Annual Symposium, as part of the National NLP Clinical Challenge (n2c2) co-sponsored by Harvard Medical School (HMS) Department of Biomedical Informatics and George Mason University (GMU) School of Engineering. During the day-long, collaborative n2c2 Shared Task and Workshop on Friday, November 3, Infolytx shared its unique NLP-based solution with industry and academic institutions from across the globe as part of the workshop goal to collectively help resolve ongoing challenges in using NLP to optimize clinical research processes.
Challenge organizers tasked participants with providing a solution to the question: “Can NLP systems use narrative medical records to identify which patients meet selection criteria for clinical trials?” Infolytx and other participants, including IBM Research, Kaiser Permanente, and Stanford University, were required to thoroughly evaluate more than 200 patient records and compare each patient to a list of 13 specific selection criteria to identify individuals eligible for clinical trial participation. Challenge participants needed to scrub patient records using their NLP solution to find individuals who met all 13 requirements, which included: the ability to make their own medical decisions, major diabetes-related complications, and advanced cardiovascular disease.
Infolytx’ data extraction framework enabled automated, scalable, and unbiased selection of patients who met all 13 selection criteria for clinical trials. In vetting more than 200 patient records, the company’s framework provided a micro F1 of 86%.
“To truly combat traditional trial inefficiencies, including time-consuming and costly patient recruitment efforts, the industry needs NLP and other AI solutions to extract valuable information from unstructured data like patient records,” said Infolytx Chief Technology Officer Dr. Zunaid Kazi. “In this case, we can use this data to target the appropriate patients for a study to help keep the clinical development process moving quickly and with quality—improving health outcomes. This is just one critical way domain expertise combined with tech-driven solutions can help the industry and ultimately, patients.”
NLP and Clinical Development Acceleration
Patient identification and recruitment are some of the biggest challenges to advancing treatment development and making much-needed therapies available to patents. With increasingly complex selection criteria due to regulatory requirements, medical researchers have to examine patient records, including extensive notes and narratives, to recruit patients—limiting them to patients who seek out trials themselves or via physician recommendations. NLP and other tech-driven solutions can help mine clinical data in EHRs for keywords and phrases to better support needs of selection criteria to improve patient identification processes for clinical trials.
“Analyzing data—structured or unstructured—is a key focus area for us, and the invitation by the n2c2 Challenge organizers to share our insights further demonstrates our team’s competence to support and accelerate the efficacy of a variety of NLP applications” noted Infolytx CEO Badrul Husain.
About the National NLP Clinical Challenge (n2c2)
Harvard Medical School (HMS) Department of Biomedical Informatics and George Mason University (GMU) Volgenau School of Engineering co-sponsored the 2018 National NLP Clinical Challenge to task participants with providing a solution to the question: “Can NLP systems use narrative medical records to identify which patients meet selection criteria for clinical trials?” The task required NLP systems to thoroughly evaluate patient records and compare each patient to a list of 13 specific selection criteria to identify target patients eligible for clinical trial participation. NLP systems that automatically assess which patients will meet complex criteria for studies can both reduce time and cost related to patient recruitment and help remove bias from trials.
Challenge co-chairs and AMIA NLP Processing Working Group Workshop organizers Özlem Uzuner, Associate Professor, information Sciences and Technology, at GMU, and Amber Stubbs, Assistant Professor, Math & Computational Sciences, at Simmons University, selected Infolytx along with four other participants—from the over 100 Challenge submissions—to orally present its unique NLP solution. Another 13 participants, including Stanford University School of Medicine (co-authors with Oracle and National Institutes of Health), Med Data Quest, and Kaiser Permanente, shared poster presentations at the 2018 AMIA annual symposium on November 2, 2018. For more information on the Challenge, visit the n2c2 page.
Infolytx is an AI and machine learning company building solutions to help healthcare and other high-tech industries understand and monetize their data. Data that is disparate, structured or unstructured can be integrated, mined, analyzed, visualized, and presented by our expert data engineers and scientists.
Today’s competitive marketplace demands that savvy companies take advantage of every asset and tool at their disposal – Infolytx can help companies tap into their data sets to gain crucial insights. With our cutting-edge technologies and world-class methodologies, we provide companies with significant competitive advantage. For more information, please visit the Infolytx site.