OBJECTIVES Alzheimer’s dementia (AD) poses a significant challenge to healthcare systems around the world.  Whilst treatment options are currently limited, with no effective disease modifying treatment, new advances in diagnosis and management of AD and promising advances in health technologies have the potential to significantly impact on the burden of the disease.  However, alongside treatment advances it is important to improve the evaluation framework if we are to capture the potential benefits to people with AD.  Methods to model disease progression over time, for use in comparative and cost-effectiveness analyses (CEA), is a priority area for further research. The objective in this research is to develop a new framework for modeling AD progression over time using the three main symptom domains of cognitive function, behaviour and mood, and functioning. METHODS Development of a descriptive system, comprising a set of health states for AD, using the three symptom domains.  Statistical modeling of disease progression through states over time, using US data from the National Alzheimer’s Coordinating Center (NACC) (n=3009). The model is tested in a decision-analytic context, using time to progression and a cost-per-QALY framework. RESULTS A 20-state disease progression pathway has been developed using multi-variate health states described using the three symptom domains. Transition probabilities and hazard rates have been estimated to model progression over time through the multi-variate descriptive system.  In a baseline model over a 5-year timeframe, using mild-to-moderate AD starting states, 78% of people progressed to health states considered severe on at least one of the symptoms (46% severe for cognition).  In a HTA context simulating a treatment with a modest effect the modeling framework predicted significant QALY differences between control and treatment over 5-years. CONCLUSIONS This new modeling framework shows promise and presents a broader opportunity to capture the impacts of treatment over time using a range of symptom domains.