Quickly detecting an unexpected pathogen can save many lives. In cases of bioterrorism or naturally occurring epidemics, accurate diagnoses may not be made until much of the population has already been jeopardized. The goal of syndromic surveillance is to detect early anomalies that emerge from patient data in a given population area and to note disease patterns before more individuals begin to experience definitive symptoms. We developed a syndromic surveillance approach for generating advance warnings of potential wide-spread diseases as well as identifying demographic attributes that are predictive of the diseases. We describe the Causal Reasoning Engine (CRE), a multipurpose decision support system for diagnosing causes from observed symptoms and predictors. The CRE uses Bayesian inference and machine learning methods and deploys an intuitive explanation-based framework for causal modeling. We also present a diagnostic decision support tool based on the CRE that allows emergency responders to analyze and interrogate findings.