This paper outlines a methodology for analyzing the representational support for knowledge-based decision-modeling in a broad domain. A relevant set of inference patterns and knowledge types are identified. By comparing the analysis results to existing representations, some insights are gained into a design approach for integrating categorical and uncertain knowledge in a context sensitive manner.