This paper provides new estimates of the impact of the time to death on health care expenditures, when time to death is assumed endogenous. It further proposes estimation methods that take into account the right censoring of time to death. The data consist of twins from Denmark aged over 70 in 1999. The age of death of the mother and the living status of the co-twin are used as instruments for time to death. The results show that IV estimators give estimates higher than those obtained in previous studies when time to death is assumed exogenous. Furthermore, the estimators proposed in this paper provide estimates that are generally lower compared to the ones obtained with the IV estimator. These results indicate that the impact of time to death has been potentially overestimated in previous studies.