Probabilistic / stochastic sensitivity analysis

Probabilistic sensitivity analysis (PSA) is a technique used in economic modelling to quantify the level of confidence in the output of the analysis, in relation to uncertainty in the model inputs. There is usually uncertainty associated with input parameter values of an economic model, which may have been derived from clinical trials, observational studies or, in some cases, expert opinion. In the base case analysis, the point estimate of each input parameter value is used. In the probabilistic analysis, these parameters are represented as distributions around the point estimate, which can be summarised using a few parameters (such as mean and standard deviation for a normal distribution).
Different distributions are generally appropriate for different types of variable, backed up by supporting evidence from source studies where possible. For example, measures of effect, such as hazard ratios or relative risk reductions, may be represented by a normal distribution and survival curves by a Weibull distribution. In a PSA, a set of input parameter values is drawn by random sampling from each distribution, and the model is ‘run’ to generate outputs (cost and health outcome), which are stored. This is repeated many times (typically 1,000 to 10,000), resulting in a distribution of outputs that can be graphed on a cost-effectiveness plane and analysed. A key output of a PSA is the proportion of results that fall favourably (i.e. considered cost effective) in relation to a given cost-effectiveness threshold. This may be represented using a cost-effectiveness acceptability curve. It is preferable to generate the primary results of an economic evaluation using the mean outcomes from the PSA, rather than the deterministic results. This is because many models contain non-linear inputs and/or contain non-linear interactions between parameters.