Lai Poh-Chin, Kwong Kim-Hung, Wong Ho-Ting
Geospat Health. 2013 Nov;8(1):183-92. doi: 10.4081/gh.2013.65.
This study describes the development of a spatio-temporal disease model based on the episodes of severe acute respiratory syndrome (SARS) that took place in Hong Kong in 2003. In contrast to conventional, deterministic modelling approaches, the model described here is predominantly spatial. It incorporates stochastic processing of environmental and social variables that interact in space and time to affect the patterns of disease transmission in a community. The model was validated through a comparative assessment between actual and modelled distribution of diseased locations. Our study shows that the inclusion of location-specific characteristics satisfactorily replicates the spatial dynamics of an infectious disease. The Pearson's correlation coefficients for five trials based on 3-day aggregation of disease counts for 1-3, 4-6 and 7-9 day forecasts were 0.57- 0.95, 0.54-0.86 and 0.57-0.82, respectively, while the correlation based on 5-day aggregation for the 1-5 day forecast was 0.55- 0.94 and 0.58-0.81 for the 6-10 day forecast. The significant and strong relationship between actual results and forecast is encouraging for the potential development of an early warning system for detecting this type of disease outbreaks.
本研究描述了基于2003年香港严重急性呼吸系统综合症（SARS）疫情构建的时空疾病模型。与传统的确定性建模方法不同，这里描述的模型主要是空间性的。它纳入了环境和社会变量的随机处理，这些变量在空间和时间上相互作用，以影响社区内疾病传播模式。该模型通过对患病地点的实际分布和模型分布进行比较评估来验证。我们的研究表明，纳入特定地点特征能够令人满意地重现传染病的空间动态。基于疾病计数3天汇总的5次试验，对于1 - 3天、4 - 6天和7 - 9天预测的皮尔逊相关系数分别为0.57 - 0.95、0.54 - 0.86和0.57 - 0.82，而基于5天汇总的1 - 5天预测的相关系数为0.55 - 0.94，6 - 10天预测的相关系数为0.58 - 0.81。实际结果与预测之间显著且强烈的关系，对于开发检测此类疾病爆发的预警系统的潜力而言是令人鼓舞的。
