Spatial Disease Surveillance by Detection of Anomalous Clusters

Daniel B. Neill, Carnegie Mellon University

Tuesday, February 21
10:15am - SENSQ 5317
Students meet the speaker at 11:30 a.m.

Refreshments at 10:00am

Hosted by Milos Hauskrecht

Abstract

The social and economic costs of disease outbreaks can often be significantly reduced by early detection. We have developed statistical and computational methods for the early, automatic detection of emerging outbreaks, by monitoring public health data (hospital visits and medication sales) for spatial clusters of disease. In this talk, I will present the "generalized spatial scan," a flexible, model-based statistical framework for accurate and computationally efficient cluster detection in diverse application domains. I will describe two of the main technological advances that have made this framework feasible and useful for cluster detection in massive real-world datasets: the "expectation-based statistic," a new spatial statistical method that combines time series analysis with anomaly detection to accurately detect emerging clusters, and the "fast spatial scan," a very fast, scalable spatial algorithm that enables us to detect clusters hundreds or thousands of times faster than previous approaches. One major contribution of this work is the development of a system for nationwide disease surveillance, currently used in daily practice by several state and local health departments. Every day, this system receives data from over 20,000 stores and hospitals nationwide, automatically detects emerging clusters of disease, and reports these results to public health officials via the Web. Through retrospective case studies and semi-synthetic testing, we have demonstrated that the system can detect disease outbreaks significantly earlier and more accurately than previous bio-surveillance methods. I am currently extending our cluster detection methods to a variety of other problems in medicine, public health, and homeland security, including tumor detection, spatial analysis of water quality data, and detection of terrorist groups.

Biography of Speaker

Daniel B. Neill is currently a Ph.D. student in the School of Computer Science at Carnegie Mellon University. His research interests include machine learning, data mining, algorithms, and game theory. He is particularly interested in developing statistical and computational methods for discovery of relevant patterns in massive real-world datasets. He received his B.S.E. from Duke University in 2001, M.Phil. from Cambridge University in 2002, and M.S. from Carnegie Mellon in 2004.