Building and evaluating machine learning models to classify appropriateness of antimicrobial prescriptions
Project Details
This project aims to utilise the Patron dataset to explore the feasibility of a passive GP NAPS and develop algorithms using machine learning to automate compliance and appropriateness antibiotic assessment.
Patron ID: PAT063
Project Lead:
The GP-NAPS is currently the most comprehensive audit of antibiotic prescribing in general practice in Australia, however, the methodology is not sustainable at scale. It requires a trained auditor to visit each general practice and review individual patient EMRs to determine appropriateness of each prescription, completing paper-based audit tools.
There is a need to develop passive automated audit but this will require improved documentation of the indication for antibiotics in the EMR and data entry of fields that are important to assess prescribing appropriateness, such as weight, age and co-morbidities, as well as the availability of pathology data such as eGFR. We will utilise the Patron dataset to explore the feasibility of a passive GP NAPS. i.e. using data extracted from general practice records, rather than onsite auditing of patient files. We will also apply machine learning techniques to automate the assessment of appropriateness of prescriptions in the Patron dataset.