How to Predict Drug Expenditure: A Markov Model Approach with Risk Classes

Although pharmaceutical expenditures have been rising for decades, the question of their drivers remains unclear, and long-term projections of pharmaceutical spending are still scarce. We use a Markov approach considering different cost-risk groups to show the possible range of future drug spending in Germany and illustrate the influence of various determinants on pharmaceutical expenditure.

We compute different medium and long-term projections of pharmaceutical expenditure in Germany up to 2060 and compare extrapolations with constant shares, time-to-death scenarios, and Markov modeling based on transition probabilities. Our modeling is based on data from a large statutory sickness fund covering around four million insureds. We divide the population into six risk groups according to their share of total pharmaceutical expenditures, determine their cost growth rates, survival and transition probabilities, and compute different scenarios related to changes in life expectancy or spending trends in different cost-risk groups.

If the spending trends in the high-cost groups continue, per-capita expenditure will increase by over 40% until 2040. By 2060, pharmaceutical expenditures could more than double, even if these groups would not benefit from rising life expectancy. By contrast, the isolated effect of demographic change would "only" lead to a long-term increase of around 15%.

The long-term development of pharmaceutical spending in Germany will depend mainly on future expenditure and life expectancy trends of particularly high-cost patients. Thus, appropriate pricing of new expensive pharmaceuticals is essential for the sustainability of the German healthcare system.