Purpose: Evidence-based justifications of personal and policy decisions depend on transparency, validation, reproducibility and maintainability. An open format graph formalism is sufficiently general to represent and illustrate all types of decision models and provides a level of abstraction that allows the model specific information to be separated from the implementation details. The subsequent clarity helps minimize modeling errors, while the ability to evaluate models using multiple software applications helps to validate the corresponding algorithms. Complete documentation of models and their underlying data sets can be achieved with an interoperable methodology that incorporates ubiquitous spreadsheet and web/browser software.

   Methods: Decision models and their corresponding data sets were represented in a set of interoperable, XML-compliant formats. The data sets, including derivations and links to sources, were represented in OpenOffice/Excel compatible spreadsheets. Additional metadata allow these spreadsheets to be transformed to a dynamic web/browser presentation and evaluation framework.

   Results: A series of Markov models to simulate the cost-effectiveness outcomes of alternative contraceptive strategies was developed for application to various populations in both the UK and US. As new evidence data were published, the datasets were revised and the interface components were dynamically updated. The spreadsheets were then automatically transformed to a web/browser presentation interface and deployed to pharmaceutical marketing agents to explore the cost-effectiveness of a product for various managed-care populations and cost structures. A variety of analyses were performed and corresponding results displayed, by transforming the required model components and customized data sets to an open-format input package for evaluation with two different evaluation engines.

   Conclusions: A set of open standards for the modular representation of decision analytic models and the complete documentation of their underlying data sets provide interoperability in a transparent and maintainable fashion. This methodology can be extended to represent a wide spectrum of model types and evaluation software applications. We would like to encourage all decision modelers to adopt such a common open standard and propose that we convene a panel for their review, refinement and endorsement.