Using Space-Time Scan Statistic To Detect Pertussis and Shigellosis Outbreaks

BACKGROUND:   Early detection of disease outbreaks is important for timely implementation of disease prevention and control measures.  Including patients’ demographic and disease information,  communicable disease data in New York State (NYS), excluding New York City, is stored in the NYS Department of Health Communicable Disease Electronic Surveillance System (CDESS).   This data can be used to identify potential disease outbreaks by detecting spatial-temporal case clusters with scan statistic.

METHODS:   The SatScan space-time permutation model was applied to pertussis and shigellosis cases reported in CDESS, using patients’ residential zip code as the geographic level and case investigation date as the event date.  The maximum cluster size was set as a circle with radius <= 20 kilometer (km) that could span up to15 days.  With cases reported since 1/1/2011, twenty-four separate space-time analyses were performed prospectively - two for each month of 2012. The SatScan analyses were called by a SAS program which also produces reports of detected clusters in table and map. The outbreak status of reported cases in CDESS was used to evaluate accuracy of the clusters detected by the scan analyses.  For the purpose of this study, a cluster is considered as a confirmed outbreak if it contains one or more outbreak cases.

RESULTS:   There were 2,719 pertussis and 802 shigellosis cases reported in 2012.  At the p< 0.05 significance level, the scan statistic detected 47 pertussis and 34 shigellosis clusters, of which 37 and 21were confirmed as outbreaks, yielded positive predictive value (PPV) of 79% and 62% respectively. The median case count was 6 with medians spatial size of 11.6 km and 7 days in duration for pertussis clusters and the medians were 8 cases, 9.2 km, and 10 days respectively for shigellosis clusters.

CONCLUSIONS: As the space-time permutation scan statistic only requires disease counts, event date and disease location, the method can be easily implemented for detecting disease outbreaks using data routinely collected from disease surveillance systems. The current study showed that scan statistic is a useful tool for detecting pertussis and shigellosis clusters with reasonable PPVs for outbreaks. This method also returns important information to assist outbreak investigations, such as geographic location and time-span of the potential outbreaks.  Since the scan statistic result can be greatly affected by disease patterns such as case frequency, incubation period, transmission pattern, etc., it is important to set up scan parameters properly according the disease of interest to achieve optimal results.