|
Activity Number:
|
318
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Risk Analysis
|
| Abstract - #300513 |
|
Title:
|
Recursive Partition for Early Outbreak Detection
|
|
Author(s):
|
Ross S. Sparks*+ and Chris Okugami
|
|
Companies:
|
CSIRO Mathematical and Information Sciences and CSIRO Mathematical and Information Sciences
|
|
Address:
|
Locked Bag 17, Sydney, International, 1670, Australia
|
|
Keywords:
|
Monitoring ; risk management ; decision trees
|
|
Abstract:
|
Vehicle crash data in New South Wales (Australia) are collected using police reports for crashes with people injured or more than $500 worth in damages. Traditionally monitoring daily counts is used to identify unusual high counts and thus flag problems on the roads. However it is known that crashes start in small clusters. Technology for detecting spatial disease clusters can be found in Raubertas (1989) and Kulldorff (2001). A surveillance system which exploits the clustering nature of crashes by tracking the sources of variation will be demonstrated as an efficient way of monitoring crashes. This strategy is demonstrated to lead to earlier detection than monitoring counts aggregations over all dimensions. This strategy also has the advantage of describing how epidemics move over time in the population.
|
