Communications of the ACM
Google, UCSF Collaborate on Machine Learning Tool to Help Prevent Harmful Prescription Errors
Machine learning experts have developed a machine learning model that can anticipate normal physician drug prescribing patterns, using a patients electronic health records as input.
Credit: Akio Kon/Bloomberg/Getty Images
A study by Google Health experts and the University of California, San Francisco's computational health sciences department detailed a machine learning model that predicts normal physician drug prescribing patterns, using a patient's electronic health records (EHRs) as input.
The model looks for any prescriptions that appear "abnormal for the patient and their current situation."
The researchers paired the EHR's historical data with current state information, resulting in models that attempt to accurately predict a course of prescription for given patients.
Google said the best-performing model was accurate "three quarters of the time," or matched up with what a physician decided to prescribe in most cases. It also had greater accuracy (93%) in predicting at least one medication that fell within a top 10 list of a physician's most likely medicine choices for a patient, even if its top choice did not match the physician's.
From TechCrunch
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found