FRAMEWORKS TO SELECT THE APPROPRIATE MODELLING APPROACH FOR DECISION-ANALYTIC ECONOMIC EVALUATIONS—DO DIFFERENT MODELLING APPROACHES LEAD TO DISPARATE RESULTS?

Purpose:

   There is a growing range of alternative modelling approaches available to conduct health economic evaluations. However, the decision on how to select a modelling approach is often unclear. This study aims to investigate the frameworks proposed in the selection of the appropriate modelling approach(es) and to examine the empirical studies that have compared the performance of different modelling approaches.

Method:

   A comprehensive literature review was conducted to identify studies from both peer-reviewed databases and grey literature that presented: (i) theoretical frameworks on how to select an appropriate modelling approach; and/or (ii) empirical studies that have directly compared different modelling approaches (i.e. models constructed with the same underlying data set), in the field of health economics. Controlled vocabulary terms and keywords were used to construct the search strategy with the search including publications up to January 2014.

Result:

   Of the 3344 unique studies identified, only eight frameworks met the inclusion criteria. Each framework highlighted a different set of selection criteria which should be considered when choosing a modelling approach. Both technical and practical features were identified as being central to this decision with the most frequently-mentioned technical elements being ‘population resolution’ (i.e. cohort vs. individual) and ‘interactivity’ (i.e. independent vs. dependent; static vs. dynamic). Seven empirical studies were identified that provide a limited assessment on th selection criteria. Consistent findings were observed in the empirical studies across a heterogeneous range of diseases. Studies that have assessed ‘population resolution’ suggest that both cohort and individualistic models produce similar results with the key consideration being a validity-feasibility trade-off. In assessing interactions amongst individuals, studies consistently demonstrated that dynamic models produce disparate results compared to static models, especially when indirect effects are large.

Conclusion:

   Although frameworks can make the choice of selecting a modelling approach more systematic, there is presently no universally-accepted, evidence-based framework. Empirical evidence can help guide the development of future frameworks. Further research is thus required to enhance our understanding on how modelling approaches differ in terms of their performances and results. This is vital given that the modelling approach selected can impact the validity of the underlying economic model and may have downstream implications in terms of a model’s efficiency, transparency and relevance to stakeholders.