Censoring is an unignorable issue when analyzing survival data and/or medical cost data. Medical costs may be viewed as a type of survival data-in that they accrue over time until an endpoint such as death-or a 'mark' variable. Since Lin et al. (1997) and Mushlin et al. (1998) published landmark papers on this topic, censored cost data have been extensively studied. In this tutorial, we explain how to estimate mean cost and cost-effectiveness ratios, along with three examples under two different data scenarios: when only total cost data (so one observation per person) or longitudinal data (or cost history) are available. We also provide an updated literature review. SAS codes in the supplement could be useful to practitioners and data analysts.
