Word search performance for diagnoses of equine surgical colics in free-text electronic patient records.

This research involves improving the search performance for the diagnosis of equine surgical colic cases in free-text electronic patient records. Through identification and elimination of search limitations, development of programmable recommendations, and provision of guidelines for effective search strategies, the researchers aim to enhance the efficiency and reliability of information retrieval from the Veterinary Medical Teaching Hospital’s clinical information system.

Objectives of the Study

• The study aimed to identify the limits of search sensitivity and positive predictive value (PPV) for free-text surgical diagnoses in electronic patient records at the Veterinary Medical Teaching Hospital (VMTH) of the University of California, Davis.
• The project sought to build procedural or programmable recommendations to eliminate these identified search limitations
• The researchers aimed to provide effective search strategy guidelines for users utilizing the VMTH’s clinical information system for aggregate searches.

Methodology Used

• All recorded horses that were submitted to the VMTH for gastrointestinal disorders that needed surgical intervention in 1995 were identified using billing procedure codes and saved in the electronic patient record.
• The researchers reviewed patient records and surgical reports to identify causes of these gastrointestinal disorders and the variances in their nomenclature.
• Keyword searches were performed for four distinct gastrointestinal disorders. The search performance was evaluated by estimating the search sensitivity and PPV.

Findings of the Study

• The results showed that search sensitivity ranged from 33% to 98%, and PPV ranged between 2% and 74%.
• Identifying that nomenclature inconsistencies led to insufficient search results, the study provided a procedural recommendation to enhance search performance— a more uniform naming of gastrointestinal disorders.
• Such uniformity in naming would liberate users from having to predict all possible word combinations that could appear in the relevant documents, thereby reducing irrelevant document retrieval and increasing search efficiency.