Dynamic disease screening by joint modelling of survival and longitudinal data

Sequential monitoring of dynamic processes is an active research area because of its broad applications in different industries and scientific research projects, including disease screening in medical research. In the literature, it has been shown that dynamic screening system (DySS) is a powerful tool for sequential monitoring of dynamic processes. To detect a disease (e.g. stroke) for a patient, existing DySS methods first estimate the regular longitudinal pattern of certain disease predictors (e.g. blood pressure, cholesterol level) from an in-control (IC) dataset that contains observations of a group of non-diseased people, and then compare the longitudinal pattern of the observed disease predictors of the given patient with the estimated regular longitudinal pattern. A signal of disease occurrence is triggered if their cumulative difference exceeds a certain level, facilitated by a built-in control chart. In practice, a dataset containing longitudinal observations of the disease predictors of both non-diseased and diseased people is often available in advance, from which it is possible to explore the relationship between the disease occurrence and the longitudinal pattern of the disease predictors. This relationship should be helpful for disease screening. In this paper, a new DySS method is suggested based on this idea. Numerical studies confirm that it can improve the existing DySS methods for disease screening.