Objective
Public health expenditures generally grow at a faster pace than GDP and their development therefore often warrants special budgetary considerations. Accurate projections of health expenditures are particularly challenging during times of economic or public health shocks when the contribution of the different expenditure drivers can shift significantly. In Austria, expenditure caps are utilized to prevent uncontrolled health expenditure growth, but setting these caps requires a good understanding of expenditure drivers.

Methods
We decompose historic public health expenditure by volume and price effects, including the two demographic (volume) effects of population growth and ageing. For this purpose we utilize health expenditure profiles by age and gender. These decompositions are then applied to project future health expenditures and we test the model fit for a past reference period compared to two other approaches of projections which are based on trend continuation and policy-favored GDP linking.

Results
We find sizable shifts in health expenditure drivers from the reference period 2013 - 2019 to our projection period of 2020 - 2028. Demographic effects caused an average increase of health expenditure by 1,3% in the reference period and somewhat decrease to 1,07% in the projection period. The reduction stems from lower population growth, while the ageing effect remains almost constant (around 0,55%). In relative terms, however, the demographic effects shift from being responsible for 31% of overall health expenditure growth to only 17%, whereas the contribution of consumer price inflation increases from 36% to 60%.
We further show that explicit consideration of demographic factors in the models improves the fit of the projections. Nevertheless, macroeconomic shocks, such as the global financial crisis 2008 - 2009, provide particular challenges for projection calibrations.

Discussion
Demographic effects are a relatively consistent and important long-term driver of health expenditure. Their explicit incorporation in models can significantly improve model fit and support a better understanding of other drivers. However, economic and public health shocks can cause significant short-term increases with potential long-term level shifts. These less predictable drivers are primarily visible through price effects.