Surgical procedure prediction using medical ontological information.

Area of Science:

• Healthcare Informatics
• Medical Data Analysis
• Surgical Workflow Optimization

Background:

• Accurate prediction of surgical procedure duration is crucial for efficient operating room scheduling.
• Traditional methods often lack precision due to the complexity of surgical procedures.

Purpose of the Study:

• To enhance surgical procedure duration predictions using medical ontological information.
• To evaluate the impact of ontological data on scheduling efficiency.

Main Methods:

• Developed two methods integrating medical information: one using Systematised Nomenclature of Medicine Clinical Terms (SNOMED CT) and another using text fragments.
• Incorporated procedure relationships into a regression model for duration prediction.
• Applied methods to New Zealand healthcare data and compared prediction accuracy.

Main Results:

• Both methods demonstrated improved prediction accuracy compared to traditional categorical encoding.
• Continuous ranked probability scores for procedure duration prediction improved from 18.4 min to 17.1 min.
• Significant improvements were observed for infrequently performed procedures (25.3 min to 21.3 min).

Conclusions:

• Medical ontological information offers a significant improvement over traditional models for predicting surgical procedure durations.
• Enhanced duration prediction accuracy positively impacts operating room scheduling efficiency, as shown in simulations.