Measures of lost years of life

In survival analysis much emphasis is on the hazard function and hazard ratios are frequently used as summary measures when comparing the survival distributions among various groups. The main reason why expected values are used less frequently is of course the presence of right-censoring that often prevents inference for the tail of the survival time distribution. As an alternative parameter, the restricted mean life time may be studied. We will first discuss how (cause-specific) measures of life years lost can be defined via the restricted mean life time. Second, we will summarize a substantial number of suggestions that has been put forward to quantify life years lost for patients suffering from a given disease compared with the general population and we will give some recommendations. Data from Danish patients with bipolar disorder will be used as illustration.