In 1998, the New York City Department of Health and the Mayor's Office of Emergency Management began monitoring the volume of ambulance dispatch calls as a surveillance tool for biologic terrorism. We adapted statistical techniques designed to measure excess influenza mortality and applied them to outbreak detection using ambulance dispatch data. Since 1999, we have been performing serial daily regressions to determine the alarm threshold for the current day. In this article, we evaluate this approach by simulating a series of 2,200 daily regressions. In the influenza detection implementation of this model, there were 71 (3.2%) alarms at the 99% level. Of these alarms, 64 (90%) occurred shortly before or during a period of peak influenza in each of six influenza seasons. In the bioterrorism detection implementation of this methodology, after accounting for current influenza activity, there were 24 (1.1%) alarms at the 99% level. Two occurred during a large snowstorm, 1 is unexplained, and 21 occurred shortly before or during a period of peak influenza activity in each of six influenza seasons. Our findings suggest that this surveillance system is sensitive to communitywide respiratory outbreaks with relatively few false alarms. More work needs to be done to evaluate the sensitivity of this approach for detecting nonrespiratory illness and more localized outbreaks.