Modelling the unobservable: the role of model calibration methods in a lung cancer setting

Abstract: In order to evaluate any policy aimed at improving the early diagnosis of a disease a good understanding of the natural history of that disease in its pre-diagnostic, and therefore pre-treatment, state is needed. Ethical and diagnostic difficulties make it all but impossible to create observational datasets suitable to observe patient transitions in these states. In this seminar I will present the role of model calibration methods to inform estimates of the transitions in such a natural history model. These methods will be applied to the evaluation of a policy aimed at improving the signs and symptoms of lung cancer and considered against the use of expert elicitation alone.