Association rule mining is used to find association relationships in data. Our work describes the use of association rule discovery as a basis for creating an early warning bio-terror attack system. The system establishes a baseline of “normal” behavior by mining historical emergency response (911) data. Using probabilistic models, we generate spatial and temporal statistics to correlate incident frequency and location in order to identify if a variation in future incidents carries an outbreak signature consistent with the effects of a biological warfare attack. Using three years of real emergency response data for experimentation, this work is focused on the activities relating to the processing and generation of detection rules. Preliminary results indicate that the system can provide reasonable detection rules but there is also more work to address inherent issues of both emergency response and biological warfare such as data quality during incident reporting and population mobility as it relates to outbreaks.