Optimal Screening for Preclinical Diseases

Certain diseases comprise an initial asymptomatic period during which they can be identified only by a screening test. In many such cases, early detection translates into benefits of more treatment options and potentially better prognosis. In this dissertation, we consider the optimal policy to screen for a preclinical disease while under limited budget. Our objective is to place any given number of screening epochs over an individual's lifetime, such that the probability of identifying the disease while preclinical is maximized. We make mild assumptions about the sojourn times of the individual in the healthy and preclinical states, and we consider the possibility of fallible screening tests. We show that a unique optimal sequence of screening times exist for our model, and that it can be quickly found by any greedy-search algorithm. We further conduct numerical experimentations by which we identify sensitive model inputs. We lastly apply our model to breast cancer screening using practical information and we investigate additional characteristics of this model.