Development of a simulation model of colorectal cancer

Colorectal cancer (CRC) is deadly if not found early. Any protocols developed for screening and surveillance and any policy decisions regarding the availability of CRC resources should consider the nature of the disease and its impact over time on costs and quality-adjusted life years in a population. Simulation models can provide a flexible representation needed for such analysis; however, the development of a credible simulation model of the natural history of CRC is hindered by limited data and incomplete knowledge. To accommodate the extensive modeling and remodeling required to produce a credible model, we created an object-oriented simulation platform driven by a model-independent database within the .NET environment. The object-oriented structure not only encapsulated the needs of a simulation replication but created an extensible framework for specialization of the CRC components. This robust framework allowed development to focus modeling on the CRC events and their event relationships, conveniently facilitating extensive revision during model construction. As a second-generation CRC modeling activity, this model development benefited from prior experience with data sources and modeling difficulties. A graphical user interface makes the model accessible by displaying existing scenarios, showing input variables and their values, and permitting the creation of new scenarios and changes to its input. Output from the simulation is captured in familiar tabbed worksheets and stored in the database. The eventual CRC model was conceptualized through a series of assumptions that conformed to beliefs and data regarding the natural history of CRC. Throughout the development cycle, extensive verification and validation calibrated the model. The result is a simulation model that characterizes the natural history of CRC with sufficient accuracy to provide an effective means of evaluating numerous issues regarding the burden of this disease on individuals and society. Generalizations from this study are offered regarding the use of discrete-event simulation in disease modeling and medical decision making.