Text

theses

Background: The adoption of computing technology in modern medical devices is ubiquitous. However, limited research currently exists on the role of computing technology on medical device failures and patient safety. The U.S. Food and Drug Administration (FDA) collects and publishes reports of medical device events as a part of its Medical Device Reporting program, but the problem codes assigned to events in the published database do not appear to be reliable to identify computing technology-related events. Methods: A supervised machine learning technique was designed and implemented to classify over 11 million natural-language narratives of medical device events reported to the FDA between 2007 and 2016 to identify events related to computing technology. The result of the classification was then used to analyze the events from several dimensions. Results: A total of 5,110,200 reports of medical device events were submitted to the FDA between 2007 and 2016. Of these, 1,155,516 (22.61%) were related to computing technology. Number of computing technology-related medical device events reported to the FDA jumped nearly 7-fold from 37,679 in 2007 to 262,407 in 2016. Nearly all (99.36%) of these reports were submitted to the FDA by the manufacturers of devices, even though patients were the original reporters of the issues leading up to the submission in nearly a third (32.46%) of the events. A total of 3,449 medical device events related to computing technology were associated with patient deaths in the 10-year period. Also, events published by the FDA on its Manufacturer and User Facility Device Experience (MAUDE) database were found to be missing problem codes in 62.74% of a sampled set (N=102) of events related to computing technology and inaccurate in 26.47% of the sampled events. Conclusions: Computing technology-related events constitute a significant portion of medical device events reported to the FDA every year. Overall, these events are on an increasing trend on an absolute basis. Manufacturers are the submitters of nearly all of the computing technology-related medical device events reported to the FDA. Medical device events related to computing technology can cause serious adverse patient events, including death. Problem codes assigned to computing technology-related medical device events in the MAUDE database published by the FDA are inaccurate at a significant rate and should not be used in research.

ETD_8760

Ph.D.

Includes bibliographical references

by Deepak Khanal

doi:10.7282/T3TX3JPK

ETD doctoral

The author owns the copyright to this work.