Systems Modeling of Lung Cancer Screening Programs to Improve Quality

Thesis (Master's)--University of Washington, 2016-08

The main aim of this thesis is to study the factors that affect the quality of lung cancer screening with special focus on community based lung cancer screening programs. The variation in false positives and other outcomes found across screening centers and across different healthcare elements within a screening center motivates the need to study this problem. Conceptual modeling and simulation modeling are the systems modeling tools employed to study this problem. This conceptual modeling portion of the thesis, deduces some qualitative insights on the importance of the role of lung cancer screening program coordinators and database management systems used in screening programs. The simulation modeling portion of the thesis utilizes a Monte Carlo simulation extended from the conceptual model. The scope of the model was restricted to the factors of importance identified by an advisory board from participating institutions. Analysis of the model develops quantitative insights on the impact of these identified factors and processes on the overall quality outcomes of the screening program as indicated by the false positive rate, early detection rate, radiation induced harms and quit rates in the smoking cessation program. In addition to the typical factors such as nodule detection sensitivity and nodule length variation, the simulation model observes the effect of recall bias in smoking history and shared decision making visits on the quality outcomes of lung cancer screening. It is concluded that, following a nodule management system like LUNGRADS can help achieve the best balance of the quality outcomes, establishing peer evaluation committees that reduce the variation in nodule length could significantly improve quality and that there should be increased focus on helping candidates quit smoking earlier on in the screening process. Though recall bias in smoking history affects false positive rate in a statistically significant way, the effect on the process is quantitatively small when compared to other factors like nodule detection sensitivity and nodule length variation.