Presentation (CTI)

OBJECTIVES: Health Economics models are primarily built in MS Excel or R so far. Although these software have shown their power, flexibility and wide availability, their use requires a learning curve, each model is re-built from scratch (unless one uses templates to be created or packages), and their structure does not facilitate quick prototyping and the exploration of alternate model structures.
METHODS: This research presents the concept of visual programming, a way to program that lets health economists create models by manipulating program elements graphically rather than by specifying them textually. Each graphical element can represent model concepts at different levels: simple programming elements (“if”, “switch”, ⋯), simple modelling concepts (unit costs, market share projections, adverse events, e.g.), complex modelling concepts (like tree node, health state, sensitivity analyses) or even encapsulate a whole Markov trace or model. Some graphical elements can also represent functions modulating the normal behaviour of other elements (like discounting, allowing vial sharing or not, ⋯) or manage data input and output (including charting and report generation). In visual programming, elements are linked (and unlinked) by dragging connectors (with a mouse in an editor). These connectors ensure information is transferred between graphical elements and determine the execution order. This versatility allows faster prototyping due to its visual nature, allowing modellers to build and test ideas quickly. Ultimately, elements and connectors can be automatically translated into classical modelling languages (Excel or R, for instance), and these implementations could be of higher quality (several rounds of specialised review) and well documented.
RESULTS: This research presents the concept of visual programming of health economics models.
CONCLUSIONS: This could allow for faster model prototyping and the easy exploration of alternate model structures while preserving the possibility of transforming them into more classical languages. A practical implementation will follow this research.