(394f) Effect of Screening Compliance on Optimal CRC Screening Policies | AIChE

(394f) Effect of Screening Compliance on Optimal CRC Screening Policies


Young, D. - Presenter, Auburn University
Cremaschi, S., Auburn University
Effect of Screening Compliance on Optimal CRC Screening Policies

Session: Chemical Engineers and Public Health

David Young, Selen Cremaschi
Department of Chemical Engineering, Auburn University, Auburn, AL, USA


It was estimated1 that in 2010 costs associated with colorectal cancer (CRC) care totaled around $14 billion for the U.S. alone. CRC is also the third most common and the second most deadly form of cancer worldwide according to the World Health Organization2. Deaths associated with CRC are closely linked to how far the cancer has progressed within the individual, with late stage 5-year survival rate for CRC being around 12 %, whereas early stage 5-year survival rate being upwards of 90 %. Therefore, early detection of CRC and its precursors is an effective mitigation technique to reduce the overall impact and the financial burden of CRC on society. Early detection is achieved for CRC through screening an asymptomatic individual using a specified screening test(s) throughout an individual’s life. What screening test(s) to perform and at what age(s) to perform these tests are referred to as a screening strategy.

The impact of different screening strategies on total CRC costs and life years gained have been studied for general populations by what-if analysis utilizing CRC microsimulations. Microsimulations are computer models that simulate numerous entities and allow researchers to study the interactions and impacts that a decision or policy has on those entities3. In the case of a CRC microsimulation, the entities are individuals, and the decisions are screening strategies to apply to a population of those individuals. Each individual has the progression of CRC described within their lifetime within a section of the simulation known as the natural history portion. This portion models the adenoma-carcinoma sequence allowing for some of the individuals to develop CRC. The screening strategy is implemented within the screening portion of the simulation. This portion incorporates screening compliance for the individuals, and it then applies the desired screening to the individual, if that individual is compliant4. Knudsen et al.5, Goede et al.6, and Aronsson et al.7 are recent studies that analyze the impact of screening strategies on CRC costs and life years gained. In these studies, only one implementation of screening compliance model was considered, e.g., flat compliance rate with an individual either always or never compliant. The results of their sensitivity analysis revealed that the benefits of screening are noticeably affected by varying compliance. Furthermore, one compliance model may not be adequate for representing the compliance behavior of all individuals in a population. To the best of our knowledge, there are no studies that systematically compare the impacts of different screening compliance models on the cost-effectiveness of and life years gained due to CRC screening strategies.

In this study, we coupled a derivative free optimization (DFO) technique with a microsimulation model of CRC to study the effects of different screening compliance models on the optimal CRC screening strategy, and cost-effectiveness and life-years gained due to that strategy. Our previous work8 showed that the coupled DFO-CRC microsimulation framework enables the identification of the optimal CRC screening strategy for cost-effectiveness. Derivative free optimization is a field of optimization in which the optimization algorithm does not know the functional form of the objective function for the problem it is solving, treating it as a black-box model. The algorithm can obtain the value of the objective function for any valid set of decision variable values by directly evaluating the black-box model. The algorithm then uses specific search strategies to determine the next set of decision variable values, iteratively trying new sets to converge to the optimal objective value. In our framework, we have a CRC microsimulation as our black-box model, with the decision variable as the screening strategy to be evaluated. The output, or objective, of our black-box model, is the difference in the total cost associated with screening and treating CRC for a population that is screened compared to the same population if they never partook in a screening regime. The objective also incorporates the value of the quality of life years gained through screening by putting a monetary value of $100,000 per quality adjusted life year gained through screening, a value used as a threshold for evaluating the effectiveness of various CRC screening strategies9. This framework is used to analyze and compare four different compliance models in terms of their effects on the cost-effectiveness and life years gained: 1) perfect compliance, i.e. every person follows the recommended strategy, 2) flat compliance, i.e., an individual is either perfectly compliant or never compliant with the recommended strategy, 3) sometimes, always or never compliant, where the sometimes compliant individuals follow a strict annual compliance, where the ages at which the strategy recommends screening are the only ages the individuals are allowed to be screened, 4) sometimes, always or never compliant, where the sometimes compliant individuals follow a loose annual compliance, where screening can occur at any age between the starting and ending ages, but must be at least the set number of years apart. Within each of these models, the compliance rate is based on the test type of the recommended screening strategy. This talk will give a brief overview of the coupled DFO-CRC microsimulation framework and discuss in detail the changes in the optimal screening strategies and the benefits to the society under different screening compliance models.


  1. Mariotto, A. B., Robin Yabroff, K., Shao, Y., Feuer, E. J. & Brown, M. L. Projections of the cost of cancer care in the United States: 2010-2020. J. Natl. Cancer Inst. 103, 117–128 (2011).
  2. World Heath Organization. Cancer. (2019). Available at: https://www.who.int/news-room/fact-sheets/detail/cancer. (Accessed: 27th February 2019)
  3. Orcutt, G. H. A new type of socio-economic system. Rev. Econ. Stat. 116–123 (1957).
  4. Rutter, C. M. & Savarino, J. E. An evidence-based microsimulation model for colorectal cancer: Validation and application. Cancer Epidemiol. Biomarkers Prev. 19, 1992–2002 (2010).
  5. Knudsen, A. B. et al. Cost-effectiveness of computed tomographic colonography screening for colorectal cancer in the medicare population. J. Natl. Cancer Inst. 102, 1238–1252 (2010).
  6. Goede, S. L. et al. Cost-effectiveness of one versus two sample faecal immunochemical testing for colorectal cancer screening. Gut 62, 727–734 (2013).
  7. Aronsson, M., Carlsson, P., Levin, L., Hager, J. & Hultcrantz, R. Cost-effectiveness of high-sensitivity faecal immunochemical test and colonoscopy screening for colorectal cancer. Br. J. Surg. 104, 1078–1086 (2017).
  8. Young, D. & Cremaschi, S. A Simulation-based Optimization Approach to Develop Personalized Colorectal Cancer Screening Strategies. in Computer Aided Chemical Engineering 44, 2125–2130 (Elsevier, 2018).
  9. van Hees, F. et al. The Appropriateness of More Intensive Colonoscopy Screening than Recommended in Medicare Beneficiaries: A Modeling Study. JAMA Intern. Med. 174, 1568–1576 (2014).