Uncertainty- The lack of perfect knowledge of the parameter value- and variability -heterogeneity of the population that is a consequence of the physical system and irreducible by additional measurements- are usually presented together as sources of variation of the model parameters in the economic evaluation (henceforth EE) of health care programs. OBJECTIVES: In this presentation, we illustrate how second-order models may be highly relevant for EE and we show that separation of uncertainty and variability is not only necessary to estimate bounds of risk but may also quantitatively affect the outcome of the EE. METHODS: An application to the analysis of cost data is provided by using three different methods. First, we make a deterministic estimate, without using probability distributions. Secondly, we derive probability distributions for the model parameters without separating uncertainty and variability. Thirdly, we do the same with making this distinction. RESULTS: The separation of uncertainty and variability is implemented in cost data using Monte Carlo simulations by first sampling once from the uncertainty distributions of the parameters and then sampling from the variability distributions of the parameters. CONCLUSIONS: The different results illustrate the relevance of using probability distributions for different model parameters, and the need to separate uncertainty and variability in the EE of health care programs.