A Bivariate Mover-Stayer Model for Interval-Censored Recurrent Event Data: Application to Joint Damage in Rheumatology

Interval-censored recurrent event data can exhibit considerable between subject heterogeneity in the event rate, and frequently there is a proportion of individuals experiencing no events. We consider a likelihood based analysis of bivariate interval-censored recurrent event data in which a random effect is used to accommodate heterogeneity in the rate of events, and bivariate mover-stayer indicators are introduced to accommodate a non susceptible sub-population for each type of event. The model facilitates the estimation of the joint distribution of two mover-stayer indicators and hence the association in the susceptibility for two types of joint damage assessments in patients with psoriatic arthritis.