Tailoring screening to address patient heterogeneity?
In its most basic form, the traditional convolution disease natural history model assumes that at its onset, the disease is asymptomatic and progresses to a symptomatic phase. Extensions of the basic model allowed investigators to incorporate stage progressions such as from a noninvasive to an invasive disease; or to refine the asymptomatic phase to allow for cases that would never progress to symptomatic disease. These models have been broadly utilized to study the natural history of breast cancer, prostate cancer, among others and to develop screening recommendations. In this talk, we revisit these models and consider Bayesian model extensions to further accommodate patient heterogeneity, in particular, considering adherence to screening recommendations. We present results from simulation studies as well as a case-study and examine the impact of patient heterogeneity on setting screening policies.