Timing of surgery for enterovesical fistula in Crohn's disease: decision analysis using a time-dependent compartment model.

OBJECTIVES

Previous decision analyses of inflam matory bowel diseases (IBD) have used decision trees and Markov chains. Occasionally IBD patients present with medical problems that are difficult or even impossible to phrase in terms of such established decision tools. This article aims to introduce modeling by a time-dependent compartment mode and demonstrate its feasibility for decision analysis in IBD METHODS: A Crohn's disease patient presented with a pelvic abscess and an enterovesical fistula. Being hesitant to operate in an acutely inflamed area, the surgeon recommended that the patient continue antibiotic therapy until the abscess had re solved. The gastroenterologist argued that the patient had already been treated with antibiotics for a prolonged time period and expressed concern that the patient's overall diminished health status would deteriorate by further delay of surgery. The occurrence of fistula, abscess, urinary tract infection, antibiotic therapy, surgical operation, and health-related quality of life were modeled as separate compartments, with time-dependent relationships among them. The simulation was carried out on an Excel spreadsheet.

RESULTS

In the model, the surgeon's predictions were associated with rapid resolution of the pelvic abscess under antibiotic therapy and improvement of the patient's health status. The gastroenterologist's predictions resulted in a smaller decline in abscess size and further deterioration of the patient's health while waiting for a definitive treatment. The disagreement between surgery and gastroenterology arose from predicting different time courses for the individual disease events, in essence, from assigning different time constants to the time-dependent influences of the disease model.

CONCLUSIONS

The compartment model provides a simple and generally applicable method to assess time dependent-changes of a complex disease. The present analysis also serves to illustrate the usefulness of such models in simulating disease behavior.