Validation of text-mining and content analysis techniques using data collected from veterinary practice management...

Area of Science:

• Veterinary Informatics
• Computational Biology
• Data Science

Background:

• Electronic patient records (EPRs) are widely used in human medicine for clinical research and decision support.
• Text mining and content analysis technologies have seen limited application in veterinary medicine.
• Veterinary EPRs contain valuable data for clinical problem investigation and practice-based research.

Purpose of the Study:

• To pilot the use of content analysis and text-mining software for analyzing veterinary electronic patient records.
• To validate this approach for future application in multi-practice research.
• To assess the accuracy of text mining tools for information retrieval from veterinary EPRs.

Main Methods:

• Utilized content analysis (Prosuite) and text mining (WordStat) software.
• Employed Keyword in Context (KWIC) and Keyword Retrieval (KR) text mining tools.
• Interrogated both a bespoke test dataset and a real-world veterinary clinic dataset.

Main Results:

• KWIC analysis demonstrated high accuracy across both datasets (sensitivity 85.3-100%, specificity 99.1-99.7%).
• Keyword Retrieval (KR) analysis, using machine learning, performed slightly better than KWIC on the clinic dataset.
• This study is the first to validate content analysis and text mining for information retrieval in veterinary EPRs for research.

Conclusions:

• Content analysis and text mining are effective tools for synthesizing and analyzing veterinary electronic patient records.
• The validated approach supports future large-scale, practice-based research in veterinary epidemiology.
• Further research is needed to explore applications for complex diseases and larger populations.