As users of long term geriatric services occupy the beds for prolonged periods of time it is important that decision makers understand how clinical and social decisions interact to influence long term care costs. A flow modelling approach enables us to estimate current inpatient activity and to test different care options, thereby optimising decision making. In previous work we developed a two compartment model of patient flows within a geriatric hospital, where patients are initially admitted to an acute or rehabilitative state from which they either are discharged or die or are converted to a long-stay state. Long-stay patients are discharged or die at a slower rate. This initial research discussed the use of a compartmental model to describe flows through the hospital system. We now discuss a three compartment model where the compartments may be described as consisting of acute care, rehabilitation and long-stay care. A Markov model is then used to count and cost the movements of geriatric patients within a hospital system. Such an approach enables health service managers and clinicians to assess performance and evaluate the effect of possible changes to the system. By attaching costs to various parts of the system we may facilitate the evaluation and comparison of different strategies and scenarios. Using the model, we show that a geriatric medical service that improved the acute management of in-patients became more cost-efficient. Hospital planners may thus identify cost-effective options.