扫描统计量的理论及其在空间流行病学中的应用

Scan statistic theory and its application in spatial epidemiology

介绍扫描统计量的基本概念、计算方法及在空间流行病学中的用途。以近年来杭州市监测点的心血管疾病急性发作资料为实例，说明回顾性时空重排扫描统计量（采用SaTScan7．0．3统计分析软件）在疾病监测中评价其时空聚集性的应用。结果表明，监测点建德的乾潭镇监测人群心血管疾病急性发作在1997年1月1日至2007年2月28日具有聚集性（P=0．001），富阳的鹿山街道、灵桥镇监测人群心血管疾病的急性发作在1997年1月1日至1999年2月28日具有聚集性（P=0．003），富阳的大源镇、新义镇和受降镇监测人群心血管疾病的急性发作在2001年3月1日至2004年2月29日具有聚集性（P=0．004）和杭州市城站街道、紫阳街道、湖滨街道、清波街道、小营街道、望江街道、潮鸣街道、长庆街道、武林街道、天水街道、文晖街道、石桥街道监测人群心血管疾病的急性发作在2004年3月1日至2006年2月28日具有聚集性（P=0．005）。回顾性时空重排扫描统计量是一种能分析疾病的时空聚集性的工具，扫描统计量是空间流行病学中可以用来评价或预测疾病聚集性的一种行之有效的方法。

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.