Publicly available machine learning models for identifying opioid misuse from the clinical notes of hospitalized patients.
Brihat SharmaDmitriy DligachKristin SwopeElizabeth Salisbury-AfsharNiranjan S KarnikCara JoyceMajid AfsharPublished in: BMC medical informatics and decision making (2020)
We demonstrate good test characteristics for an opioid misuse computable phenotype that is void of any PHI and performs similarly to models that use PHI. Herein we share a PHI-free, trained opioid misuse classifier for other researchers and health systems to use and benchmark to overcome privacy and security concerns.