Model Overview
The CRC model was developed to examine the impact of population screening strategies and compliance on the development and consequences of CRC. The model was developed by the University of North Carolina at Chapel Hill (UNC) and North Carolina State University (NC State) using AnyLogic®simulation software, which is built on the Java® programming language.
The model is intended to be used as a “virtual world” in which to simulate the effects of alternate scenarios about population demographics, disease determinants, clinical interventions, or policy on CRC screening, incidence, treatment, and mortality within states over time to inform state-level, system-level, and community-level analyses. The model can simulate realistic cohorts (e.g., for comparative effectiveness research) or the entire state population (e.g., to support endoscopy capacity planning). While the model reflects best available evidence and substantial local data, many of its components remain uncertain. Therefore, the model should be considered a representation of our current understanding of the determinants of CRC disparities across the state, and should be tested and updated as new data become available.
Model Description
The model includes three sets of assumptions about demography, natural history, and screening as inputs. The demographic inputs include the age- and race-specific populations of state residents, and their geographic distribution, income, and insurance status. The natural history inputs include parameters that determine the development and progression of colorectal polyps, CRC incidence, and mortality from CRC and other causes. Inputs for screening and testing include compliance with recommended screening, diagnostic testing, and surveillance, as well as preference for the testing modality received. Outputs include CRC cases by age, race, income, and stage at diagnosis, compliance with screening, CRC deaths, and costs incurred by CRC screening and treatment.
Using the model, one can see how outputs of cancer incidence and mortality are affected by varying the inputs of the model. For example, “screening as usual” uses the historical trends for CRC incidence and mortality, along with current, observational data on preferences for testing modality as well as compliance with recommended screening to generate the number of CRC cases, CRC deaths, and costs associated with testing and treatment under present policies. The “screening as usual” scenario can be compared to a scenario for which compliance with routine screening is increased by use of a population-level intervention. The number of persons up-to-date with screening, CRC cases, CRC deaths, and costs associated with testing, treatment, and the intervention are generated from the modified model. The intervention can be compared to screening as usual with respect to the life-years up-to-date with screening, life-years gained, and life-years gained per dollar spent on screening (i.e., the cost-effectiveness of the intervention).
Demography Determines the CRC-Free Lifespan
The demography part of the model generates a life history for each simulated individual according to the population structure and death rates. The individuals in the population are followed for their entire lifetime. We collect statistics related to CRC screening on all individuals who are between the ages of 50 and 75. The population is based on U.S. Census data from the year 2009 and the American Community Survey completed in 2009 for a state. A simulated life history includes the date of death from causes other than CRC.
Natural History May Modify the CRC-Free Lifespan
An individual’s natural history is defined as the health outcomes that would occur in the absence of CRC screening. This includes the development and progression of polyps and cancer, the clinical discovery of cancer through the emergence of symptoms, and death from CRC or other causes. The expected lifespan is based on race- and sex-specific life tables from the U.S. Census.
In the model, simulated individuals may have multiple lesions develop in their lifetime. Lesions represent the emergence of a polyp or preclinical cancer. The natural history parameters determine whether polyps turn into CRC. Once a polyp has developed into clinical CRC, the survival experience of a person is generated according to the survival parameters.
The natural history of CRC for each simulated person may modify his or her CRC-free lifespan. If a person dies from CRC before he or she would die from other causes, the lifespan and age at death are adjusted accordingly.
Screening May Modify the CRC Natural History Lifespan
The screening part of the model may further adjust the life history from the natural history part of the model. The screening and natural history parts of the model run simultaneously so that screening for CRC may detect polyps or cancers at any stage in the natural history. Polyps that are detected during screening will be removed and lesions will be biopsied for clinical diagnosis and treatment. As a consequence, a person’s life history may be modified by diagnosis and treatment of precancerous lesions or CRC. The effect of screening, if any, is reflected in the aggregated gains in life-years as a result of screening.
The model can be used to test the effects of various interventions on life-years and costs by increasing an individual’s probability of being screened for CRC. The intervention part of the model has no effect on the course of diagnostic or surveillance testing.