Identifying a poisoning from clinical patterns
The team compared four machine-learning classifiers using 201,031 NPDS records from 2014–2018. The models distinguished eight selected drugs or drug classes with overall accuracy of 77–80%. This was a retrospective pilot restricted to single-agent exposures, not a validation of diagnosis in unselected emergency patients.
First author; corresponding author in the publisher record.
Full citation & links
Omid Mehrpour, Christopher Hoyte, Heather Delva-Clark, Abdullah Al Masud, Ashis Biswas, Jonathan Schimmel, Samaneh Nakhaee, Foster Goss. Classification of acute poisoning exposures with machine learning models derived from the National Poison Data System. Basic & clinical pharmacology & toxicology. 2022;131(6):566-574. DOI: 10.1111/bcpt.13800.