Costs and effectiveness of screening strategies for colorectal cancer can be predicted using mi­crosimulation models, such as MISCAN-Colon. Cost-effectiveness analyses use these outcomes to recommend efficient strategies. These studies usually consider a small number of strategies, mostly using only a single screening test and fixed intervals between interventions. By consid­ering more strategies, efficiency can possibly be improved. However, the number of possible strategies is high and the microsimulation models are computationally expensive. Thus, an effi­cient algorithm is needed to identify efficient strategies. This thesis compares the performance of four multi-objective evolutionary algorithms (NSGA-11, SPEA2, PESA-II and IBEA) on an enumerated test case. First, each algorithm was tuned to perform well on this test case. Per­formance was then assessed using three unary (c-Performance, Inverted Generational Distance and Hypervolume) and two binary (Binary Hypervolume and Coverage) multi-objective perfor­mance measures. Statistical analysis showed that all measures indicate that NSGA-11 performs best on this problem. SPEA2 performed slightly worse, followed by IBEA and finally PESA-II. Inverted Generational Distance and Hypervolume were the most powerful measures, as they were able to find significant differences between each pair of algorithms. Finally, NSGA-11 was used to identify efficient strategies for a real case based on the United States scenario. Effec­tiveness could be improved by 2-22%, depending on the budget. Costs could be reduced by 8-201%, depending on the desired effectiveness. However, the identified strategies are probably too complex to be implemented in practice.

Additional Metadata
Keywords : colorectal cancer, screening, cost-effectiveness, microsimulation, multi-objective optimization, multi-objective evolutionary algorithms
Thesis Advisor Birbil, S.I.
Persistent URL
Series Econometrie
Dunnewind, N. (2020, February 14). Finding cost-effective colorectal cancer screening strategies using multi-objective evolutionary algorithms and the MISCAN-Colon microsimulation model. Econometrie. Retrieved from