Overcoming challenges in extracting prescribing habits from veterinary clinics using big data and deep learning

Area of Science:

• Veterinary Medicine
• Data Science
• Pharmacology

Background:

• Antimicrobial stewardship is vital for combating resistance.
• Previous studies used VetCompass Australia and Natural Language Processing (NLP) to analyze antimicrobial use in companion animals.
• Challenges exist in extracting usable data from raw clinical notes.

Purpose of the Study:

• To develop and apply NLP systems for automated extraction of antimicrobial usage data.
• To assess the completeness of recorded data for evaluating appropriate antimicrobial use in companion animals.
• To understand population-level antibiotic prescribing patterns in Australian companion animal practices.

Main Methods:

• Developed NLP systems using rules-based algorithms and machine learning.
• Automated extraction of clinical indication, antimicrobial agent, dose, and duration.
• Applied methods to over 4.4 million companion animal clinical records from Australia.

Main Results:

• Analyzed consultations involving antimicrobial use across Australian companion animal practices.
• Found that only approximately 40% of records included the reason for prescription, dose, and duration.
• Identified significant data deficiencies in clinical records for assessing antimicrobial use.

Conclusions:

• Automated data extraction using NLP shows promise but is limited by incomplete primary data.
• Essential data for antimicrobial stewardship are frequently missing from clinical records.
• Addressing data recording practices is crucial for effective antimicrobial stewardship analysis.