Applied Veterinary Informatics: Development of a Semantic and Domain-Specific Method to Construct a Canine Data...

Area of Science:

• Veterinary Medicine
• Biomedical Informatics
• Comparative Pathology

Background:

• Companion animals spontaneously develop diseases mirroring human conditions, offering unique models for pathogenesis research.
• Existing veterinary clinical data is often unstructured, hindering its use in large-scale studies.
• Developing robust data repositories is crucial for advancing veterinary and comparative medical research.

Purpose of the Study:

• To develop a semantic and domain-specific method for constructing a high-quality canine clinical data repository.
• To enable the utilization of routinely collected veterinary data for future research on disease pathogenesis.
• To create a valuable resource for studying spontaneous disease development in companion animals.

Main Methods:

• A two-phase data-cleaning framework combining semantic and domain-specific approaches was developed.
• The method was applied to clinical data from the Matthew J. Ryan Veterinary Hospital (PennVet).
• Data quality and retention were validated using known breed predispositions for specific diseases.

Main Results:

• A canine data repository of 84,405 dogs (2000-2017, 194 breeds) was successfully constructed.
• The two-phase method achieved 99.8% data retention while ensuring data quality.
• The repository validated expected breed associations for mitral valve disease, atrial fibrillation, and osteosarcoma.

Conclusions:

• The described two-phase method is effective for building comprehensive and high-quality veterinary clinical data repositories.
• This approach maximizes data utility from routine veterinary care for research purposes.
• The canine data repository serves as a valuable resource for comparative disease studies and veterinary research.