Inverse probability weighted generalised estimating equations for longitudinal data

Abstract: Longitudinal study designs, in which variables of interest are observed at multiple time points in a study population, are frequently used in clinical research. Missing data are common in these types of studies. Moreover, in studies investigating a population where the mortality rate is high, data can be truncated by death. To handle missing or truncated data, it is important to investigate the underlying causes to be able to mitigate potentially biased results. By simulating data with different underlying missingness mechanisms, several estimands were investigated in an elderly study population with distinct handling of missing data and truncation by death. Unweighted and weighted generalized estimating equations (IPWGEE) were used to estimate the lung function decline by age. The results suggest that the IPWGEE method is robust when applied to different estimands based on the simulated data set.