Page 90 - Htain Manual

Sensitivity analysis

Sensitivity analysis is a way to address the extent of uncertainty in the results and

outcomes of an economic evaluation arising due to underlying assumptions (both in model
structure and methods), precision of the parameter estimates and generalizability.

Methodological uncertainty pertains to disparities in the choice of methods used in an

economic evaluation related to assessment of costs, health consequences, quality of life,

perspective, etc. Similarly, uncertainties arising due to underlying assumptions and scientific

arguments made while designing and interpreting the structure of a decision analytic model
are defined as model structure uncertainty. Parameter uncertainty refers to ambiguity related

to the true value of model parameters such as relative risk, survival rates, transition

probabilities, etc. Generalizability relates to extent to which the results of a cost effectiveness
analysis can be applied to different settings in terms of different geographical regions,

population groups, etc.

There are two basic types of sensitivity analysis approaches i.e., Deterministic

sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA). Deterministic sensitivity
analysis is carried out to assess that how the results of an economic evaluation are sensitive

to variation or change in values of certain parameters. For example, a parameter value is

changed from upper to lower bounds to see how sensitive the ICER value is with respect to
the change in specific parameters.

Generally there are three main types of DSA used in health economic evaluations i.e.,

one way or univariate, two-way or bivariate and multi-way or multivariate sensitivity analysis.

The traditional univariate sensitivity analysis tries to examine the change in an ICER by varying
the value of one variable at a time and holding all other parameters constant. Similarly, in

two-way and multi-way analysis the value of two or more parameters are changed

simultaneously and its impact on ICER is seen.

A sub-type of one-way sensitivity analysis is threshold analysis. Under this, the value

of a certain parameter is varied over a range and a level is determined, below or above which

the conclusion of the study changes. For example, a threshold analysis can be undertaken to

78