OBJECTIVES: To develop a stochastic multi-domain micro-simulation model for evaluation of cost-effectiveness and long-term outcome in Alzheimer's disease. METHODS: Key disease indicators (e.g. cognitive function, functional abilities - ADLs and care setting) were simulated over time for individual patients using regression functions derived from longitudinal observational data. Micro-simulation of each individual patient enables incorporating individual variability over time into the model, i.e. the disease progression depends on individual characteristics and previous progression rates. The disease indicators together with patient characteristics were used to predict the need for health care services and ultimately the need for full time care. RESULTS: The model simulated individual patients estimating cognitive function, physical function, resource utilization and care setting for each 6 months period until the event of death. Average disease progression rates and estimated resource use well corresponded to what have been observed in clinical praxis. CONCLUSION: Existing models stratify patients into artificial cohorts using single domains (typically either cognition or care setting) thereby neglecting important explanatory variables and limiting the extent to which individual variability can be modeled. The proposed model provides a dynamic simulation framework completely based on regression functions. This enables inclusion of all relevant disease indicators and incorporation of individual variability into disease progression functions. The proposed model can be used for economic evaluation of any treatment intervention.