abstract  illustration

Using Natural Language Processing to Extract Predictors of Medical Device Adverse Events

March 6, 2020

Pinar Karaca-Mandic, PhD, an economist and academic director of the Medical Industry Leadership Institute (MILI) in the Carlson School of Management, examines how medical technologies enter and exit markets. As director of MILI, she works with academics, physicians, and industry experts to understand and improve health care. She is the principal investigator for a recently funded BOLD Ideas project that seeks to employ natural language processing to extract predictors of medical device adverse events.

“Minnesota is a global hub for the medical device industry, and this project will provide a key ingredient to assessing medical device safety and whether device regulation may hinder health,” said Karaca-Mandic. “The FDA allows many new medical devices to enter the market, not because of their demonstrated clinical efficacy, but because of their similarity to older devices already on the market. However, these so-called predicate devices may have originally been approved by the FDA based on now outdated clinical data.”

There is growing concern that devices approved based on similarity to predicate devices, rather than demonstrated efficacy, are unsafe. Recent widespread recalls of textured breast implants, transvaginal meshes, and metal-on-metal hip implants stemmed from devices entering the market based solely on comparisons to predicate devices. However, it is not known how the safety of new devices are associated with age and other features of the predicate devices. This is because the FDA does not provide research-ready data on predicate devices in approval of new devices. Instead, information on predicate devices is inconveniently stored across thousands of pdf documents.

Karaca-Mandic is leading an interdisciplinary team from the Carlson School of Management and the School of Public Health to extract information on predicate devices from these documents and create a database. This project brings together experts in natural language processing and the medical device industry including Soumya Sen, PhD, an information scientist and director of research at the Management Information Systems Research Center; Yi Zhu, MS, a PhD student in the Carlson School with experience on natural language processing projects; and Alexander Everhart, a PhD candidate in the School of Public Health who works as an economist at Medtronic.

The team will consult experts from MILI and the Earl E. Bakken Medical Devices Center (MDC). MILI's executives-in-residence include regulatory experts from industry leaders Medtronic, C.R. Bard, and Boston Scientific, while the MDC regularly counsels inventors and start-ups on device regulations. The team will develop their methodological approach and disseminate results in consultation with these experts to insure their research benefits patients, companies, and regulators alike.

This initiative received funding from the Office of Academic Clinical Affairs BOLD Ideas Grant program which supports interdisciplinary teams seeking to tackle the complex issues inhibiting the health and wellbeing of our communities.

Other News

Cuong Pham
For Men’s Health Month in June, Cuong Pham from the Community-University Health Care Center answers questions about why it is important for…
health communication network
Assistant Professor David Haynes of the University’s Institute for Health Informatics is trained as a health geographer. Through geographic…
Group of older people with arms linked
With more people living longer, we are approaching the tipping point at which our state, country, and world are impacted by the double…