PCN23: DISEASE MODELING: DEVELOPING THE INFRASTRUCTURE FOR A COMPREHENSIVE, MULTI-NATIONAL, CLINICAL AND ECONOMIC BREAST CANCER TREATMENT MODEL

Abstract

OBJECTIVES: To develop the cost data infrastructure to support a comprehensive, multi-national breast cancer treatment decision-analysis model. The specifications required a user-friendly interactive interface for over 70 comparators composed of nearly 350 cost components used in 24 unique decision trees. The model required an ability to vary components readily and add new treatments and cost components to multiple trees. METHODS: Since standard decision-analysis software doesn't permit categorization of variables or application of the same variable to multiple decision trees: (1) trees were programmed in Visual Basic for the interactive interface, and (2) cost data were loaded into a Microsoft ACCESS database linked to the trees. Because of this structure, it was possible to categorize cost data as: 1) Drug Acquisition and Administration, 2) Adverse Events/Complications, 3) Concomitant Medications, 4) Hospitalizations, and 5) Monitoring Costs. For each country in the model, a separate database was developed with country-specific costs obtained from standardized databases, government sources, published literature, and a provider survey. RESULTS: This model was developed for six countries—US, U.K., Germany, Japan, France, and Italy—and included clinical and economic variables related to the diagnosis, treatment, and outcomes of breast cancer. The structure permits dynamic analyses via varying cost and probability scenarios that reflect country-specific treatment practices and international variations. Each country's cost database applies to four distinct decision trees representing different stages of breast cancer. The costs can be easily summarized by category and modified so that multiple cost components in multiple trees can be varied with one edit. New cost components can be added to each country's database and linked to the trees. CONCLUSION: When constructing large models (such as disease models) with several treatments having common cost components in multiple decision trees, using a categorized cost database linked to the treatment pathways will generate a user-friendly model with easily-varied cost inputs.