Scan statistic theory and its application in spatial epidemiology / 中华流行病学杂志

ABSTRACT

To introduce the basic concept of scan statistic, its computation method and application in the area of spatial epidemiology. Retrospective space-time permutation statistics for evaluating the clustering of disease monitoring program is illustrated, using data on recent acute onset of cardiovascular disease in Hangzhou, China. Calculations were performed with SaTScan Version 7.0.3. With 999 Monte Carlo replications, the program took 5 seconds to run on a 100-MHz Pentium PC. The geographical surveillance program on acute onset clusters of cardiovascular disease, data which showed statistical significance, would include a) from January 1, 1997 to February 28, 2007 in Qiantan township, Jiande county (P = 0.001); b) highly significant between January 1, 1997 and February 28, 1999 for Lushan street, Lingqiao township in Fuyang county (P = 0.003); c) between March 1, 2001 and February 29, 2004 for Dayuan town, Xinyi town, Shouxiang town in Fuyang (P = 0.004); d) between March 1, 2004 and Feb 28, 2006 for Chengzhan street, Ziyang street, Hubin street, Qinbo street, Xiaoying street, Wangjiang street, Chaoming street, Changqing street, Wulin street, Tianshui street, Wenhui street and Shiqiao street in Hangzhou (P = 0.005), respectively. The retrospective space-time permutation statistics seems useful as a screening tool for identifying the cluster of disease. Scan statistics are practical and effective method for deciding which cluster alarms would merit further investigation and which clusters are probably chance occurrences in the study of spatial epidemiology.