Estimating Age-Specific Incidence Proportions in the Presence of Non-Response and Death
Abstract
This paper discusses estimation of age-specific incidence proportions of disease from an observational cohort study, with baseline and follow-up interviews, where age-specific occurrence of disease is scheduled to be determined at the follow-up interview, but is missing on some individuals due to death and non-response. We develop a sensitivity analysis methodology to address the issues that the missing data may be informative, so that the age-specific incidence of disease among those individuals without follow-up data is different from those with follow-up data. This work is motivated by the Baltimore City Epidemiological Catchment Area (ECA) Study, in which the disease of interest is first episode of major depression as defined by the Diagnostic Statistical Manual/DSM-IV.