Trottier Helen, Philippe Pierre, Roy Roch
Department of Social & Preventive Medecine, University of Montréal, Montréal, Canada.
Emerg Themes Epidemiol. 2006 Aug 8;3:9. doi: 10.1186/1742-7622-3-9.
The goal of this paper is to analyze the stochastic dynamics of childhood infectious disease time series. We present an univariate time series analysis of pertussis, mumps, measles and rubella based on Box-Jenkins or AutoRegressive Integrated Moving Average (ARIMA) modeling. The method, which enables the dependency structure embedded in time series data to be modeled, has potential research applications in studies of infectious disease dynamics. Canadian chronological series of pertussis, mumps, measles and rubella, before and after mass vaccination, are analyzed to characterize the statistical structure of these diseases. Despite the fact that these infectious diseases are biologically different, it is found that they are all represented by simple models with the same basic statistical structure. Aside from seasonal effects, the number of new cases is given by the incidence in the previous period and by periodically recurrent random factors. It is also shown that mass vaccination does not change this stochastic dependency. We conclude that the Box-Jenkins methodology does identify the collective pattern of the dynamics, but not the specifics of the diseases at the biological individual level.
本文的目标是分析儿童传染病时间序列的随机动态。我们基于Box-Jenkins或自回归积分移动平均（ARIMA）建模方法，对百日咳、腮腺炎、麻疹和风疹进行单变量时间序列分析。该方法能够对时间序列数据中蕴含的依存结构进行建模，在传染病动态研究中具有潜在的研究应用价值。我们分析了加拿大在大规模疫苗接种前后的百日咳、腮腺炎、麻疹和风疹按时间顺序排列的序列，以刻画这些疾病的统计结构。尽管这些传染病在生物学上存在差异，但研究发现它们都可以用具有相同基本统计结构的简单模型来表示。除了季节性影响外，新发病例数由上一时期的发病率和周期性反复出现的随机因素决定。研究还表明，大规模疫苗接种并没有改变这种随机依存关系。我们得出结论，Box-Jenkins方法确实能够识别动态的总体模式，但无法识别生物个体层面疾病的具体特征。
