Prognostic tools or risk scores that accurately estimate mortality risk could provide valuable information for elderly nursing home residents, their families, and physicians when establishing goals of care, such as when considering a shift to palliative care. Several Dutch and US studies found male gender, ADL impairment, age, poor intake, and shortness of breath were consistently associated with higher mortality risk both in dementia patients and mixed nursing home populations. Nevertheless, existing prognostic models exhibit only fair performance in identifying those at high risk of dying, which likely impedes their use. Accurately predicting death is challenging for conceptual and practical reasons. It is clearly time for a new research agenda. Future model development might focus on specific conditions (e.g., pneumonia), alternative treatments, outcomes other than mortality, and identifying the best approach to selecting risk cut-off scores to support palliative care decisions. Further study of risk scores might include using them in developing stepwise care planning processes, combining them with physicians' own estimates, and using them as educational tools. Developing a balanced view of the mortality-risk score's usefulness in nursing homes will require further validating prognostic models and testing clinical impact in a variety of settings. © 2011 Elsevier Masson SAS and European Union Geriatric Medicine Society.