Deterministic Model

A model in which all input parameters take single, fixed values, producing one point estimate of cost and effect, unlike a probabilistic model.

Source: Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. Oxford University Press; 2006. doi:10.1093/oso/9780198526629.001.0001.

A deterministic model assigns each input one fixed value, so running it produces exactly one estimate of cost and effect, with no spread around it. Because nothing in the calculation is random, repeating the run gives the same answer every time. Any sense of how uncertain that answer is must come from changing the inputs by hand, one or several at a time, rather than from the model itself. This single-valued nature is what distinguishes it from a probabilistic model. Briggs and colleagues (2006) contrast the two.

Source: Briggs et al. 2006

A deterministic model handles uncertainty not within the model itself but through sensitivity analysis, in which input parameters are varied to see how the results change. One-way sensitivity analysis varies one parameter at a time, and scenario analysis varies several together, showing how sensitive the conclusion is to particular assumptions. Because the base model uses fixed values, it gives a single estimate, and this separate exploration of how inputs affect results is how a deterministic approach addresses uncertainty.

Source: Briggs, Claxton & Sculpher 2006

A deterministic model uses fixed values for its parameters and produces a single point estimate, whereas a probabilistic model represents parameters by probability distributions and, through simulation, produces a distribution of results reflecting the combined parameter uncertainty. Deterministic sensitivity analysis varies inputs by hand, while probabilistic analysis samples them together. The probabilistic approach captures overall uncertainty and correlations more fully, but the deterministic model is simpler and provides the base-case estimate around which uncertainty is explored.

Source: Briggs, Claxton & Sculpher 2006

A deterministic model is simple, transparent, and quick to run, giving a clear base-case result that is easy to compute, check, and communicate. Deterministic sensitivity analysis makes the influence of individual parameters explicit, showing which assumptions matter most. Because it does not require specifying distributions or running simulations, it is less demanding than a probabilistic model. These features make the deterministic approach a useful starting point and a clear way to present base-case results and one-way sensitivity.

Source: Briggs, Claxton & Sculpher 2006

A deterministic model does not capture the combined effect of uncertainty in all parameters at once, since it uses fixed values and varies inputs only a few at a time, so it cannot show the overall uncertainty in the result or the probability that one option is cost-effective. It also may miss the effect of correlations between parameters. For a full account of decision uncertainty, a probabilistic model is needed, so the deterministic approach is often complemented by probabilistic analysis.

Source: Briggs, Claxton & Sculpher 2006