An automated process that combines natural language processing and machine learning has identified people who inject drugs (IDUs) in electronic health records more quickly and accurately than current methods that rely on manual review of records.
Currently, people who inject drugs are identified by International Classification of Diseases (ICD) codes that appear in patient records. electronic health health care provider records or extracted from these notes by trained coders who review them for billing purposes. But there is no specific ICD code for this injecting drug usetherefore, providers and coders must rely on a combination of non-specific codes as a proxy for PIN identification – a slow approach that can lead to inaccuracies.
Researchers manually reviewed 1,000 records from 2003 to 2014 of people admitted to Veterans Administration hospitals with bacteremia caused by Staphylococcus aureus, a common infection that develops when the bacteria enter skin openings, such as injection sites. They then developed and taught how to use the algorithms natural language processing and machine learning and compared them to 11 proxy ICD code combinations for PIN identification.
Limitations of the study include potentially poor provider documentation. In addition, the dataset used was from 2003 to 2014, but since then the epidemic of injecting drug use has shifted from prescription opioids and heroin to synthetic opioids like fentanyl, which the algorithm might miss because there aren’t many examples of the drug in the dataset where it learned the classification. Finally, the findings may not be generalizable to other circumstances, given that they are based entirely on Veterans Administration data.
The use of this artificial intelligence model significantly accelerates the process of identifying PINs, which can improve clinical decision-making, health care research and administrative oversight.
“Using natural language processing and machine learningwe could identify people who inject drugs in thousands of notes in a matter of minutes, compared to what it would have taken a reviewer weeks to do,” said lead author Dr. David Goodman-Meza, associate professor of medicine in the Division of Infectious Diseases at the David Geffen School of Medicine at the University of California, Los Angeles. health care systems to identify PINs to better allocate resources such as needle and substance abuse services and mental health treatment programs for people who use drugs.”
The study is published in a peer-reviewed journal Open forum on infectious diseases.
David Goodman-Meza et al. Natural language processing and machine learning to identify people who inject drugs in electronic health records, Open forum on infectious diseases (2022). DOI: 10.1093/ofid/ofac471
Citation: Artificial intelligence tools quickly detect signs of injection drug use in patient health records (2022, September 22) Retrieved September 22, 2022, from https://medicalxpress.com/news/2022-09-artificial-intelligence-tools -quickly-drug .html
This document is subject to copyright. Except in good faith for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.