Objectives: This study aimed to explore patterns of health expenditure for in-patient care in the last 3 years of life so as to understand how age and time to death contribute to health-care expenditure.Method: Records of all deaths occurring in Western Australia from 1997 to 2000 inclusive were extracted from the WA mortality register and linked to records from the hospital morbidity data system (HMDS) via the WA Data Linkage System. Inflation adjusted hospital costs were assigned to all in-patient events occurring within 3 years of death from five major causes of death using DRG costing information.Results: Prior to the last 5 months of life the mean cost of hospitalisation was positively associated with age; however, the magnitude of the cost increase in the last 5 months of life was inversely related to age such that the cost in the last month of life was similar across age groups.Conclusion: The finding that increased costs are associated with proximity to death, but that the magnitude of the increase is inversely associated with age, has implications for the ongoing debate about whether proximity to death or age is the dominant driver of health-care costs. The results of this study suggest that models forecasting future health-care expenditure should take into account the interaction of age, time to death and cause of death. In addition, we propose that due to the differences observed across causes of death it may be that a single general population model may not be capable of fully capturing the relationship and that this may be why the debate regarding age and time to death has yet to be resolved in the literature. (c) 2007 Elsevier Ireland Ltd. All rights reserved.