OBJECTIVES: Economic evaluation(EE) incorporate some degree of uncertainty and variability that arises in a number of ways. Uncertainty represents lack of perfect knowledge on the part of the analyst and may be reduced by further measurement and variability represents heterogeneity or diversity in a population that is irreducible by additional measurements (Spanish-guidelines proposal). This paper tries to shed light on the need to separate uncertainty and variability in the EE. METHODS: We propose the Probabilistic Sensitivity Analysis (PSA) as an efficient methodology to treat uncertainty associated to the model “inputs”. In PSA, a single variable (or subset of variables) is allowed to vary within its specified probability distribution, and repeat-run sampling-based simulations are performed to produce a weighted distribution of output estimates. It is proposed a bayesian estimation of the results of a target parameter [θ|Data]=[Data|θ]*[θ]/[Data] subsequently to PSA as an improvement of the method. We propose calculating the Bayesian interval of probability (BIP) [θ|a,b] of the costs associated with treatment during the PSA calculations(it has been assumed that [θ|a,b]≈Beta(a,b)[UNKNOWN NODETYPE 9]), defined as those that have an interval probability "high" to contain the parameter; equivalent to frequentist confidence interval P(θmin≤θ≤θmax)=1-α[UNKNOWN NODETYPE 9], using Markov Chains Monte-Carlo but measured as a probability not as confidence (α based). RESULTS: We have studied different scripts using WinBugs and FirstBayes packages for calculating of the estimated costs BIP in a PSA, simulating highly skewed distributions of costs. The separation of uncertainty and variability can affect the study results and policy-making decisions in a non-negligible manner and the best methodology to treat the uncertainty is PSA. CONCLUSIONS: Furthermore this paper is a brief introduction to the decision models, their relation to Bayesian decision theory, and the tools typically used to describe the uncertainties involved presenting an improvement in the PSA using a BIP of the estimated parameters as a robust method.