Introduction: State-transition Markov models have become the normative approach to health technology assessment (HTA) in the UK. However, the growing sophistication of these methods has promoted the need to increase accessibility and to extend understanding of outputs beyond expert decision analysts. In this context, there are obvious benefits in exploiting the increased graphical and interactive capabilities of computers as well as the growing research literature in information visualisation. The success of visual interactive modelling in discrete event simulation is testimony to the impact that graphical interfaces can bring to model accessibility.

Methods A case study of HTA modelling in dual-chamber heart pacemakers was used to identify areas where visualisation could be usefully implemented. This study uses a Markov model implemented in Microsoft Excel to assess cost utility for a range of scenarios of care. One key requirement was to explore sensitivity across a range of model input parameters and for a number of different patient sub-groups. A ‘sensitivity mixing desk' was implemented enabling interactive control for an array of key model inputs using graphical slider controls. These are varied independently and the effects on the main model outputs observed visually in real time. In addition, a prototype animation was developed which provides a graphical depiction of the dynamics of the Markov model and allows users to control and interactively browse patient flows within the state-transition framework.

Results: Early evaluations of the prototype visualisation tools in this study are promising. Specifically the direct feedback of outputs based on interactive manipulation of model inputs given by the ‘sensitivity mixing desk' is useful in promoting understanding. Users report a strong impression gained of the relative influence of input variables through the use of this application. The Markov animation tool has also proved successfully in demonstrating model structure and dynamics to non-expert users in a direct and accessible graphical format.

Conclusions: There is considerable scope for graphical tools to improve accessibility and promote understanding of Markov models in health technology assessment. The requirement for visualization tools is increasing as the sophistication of model outputs grows and the requirement for understanding extends beyond communities of modelling and decision analytic experts. Sensitivity analysis is a key area where graphical outputs can be of specific importance.