Dimitrov B D, Babayev E S
Academic Unit of Primary Care & Population Sciences, Faculty of Medicine,University of Southampton,Southampton,UK.
Azerbaijan National Academy of Sciences, Baku,Republic of Azerbaijan.
Epidemiol Infect. 2015 Jan;143(1):13-22. doi: 10.1017/S095026881400048X. Epub 2014 Mar 18.
Multicomponent cyclicity in influenza (flu) incidence had been observed in various countries (e.g. periods T = 1, 2-3, 5-6, 8·0, 10·6-11·3, 13, 18-19 years) and its close similarity with cycles in natural environmental phenomena as meteorological factors and heliogeophysical activity (HGA) suggested. This report aimed at verifying previous results on cyclic patterns of flu incidence by exploring whether flu annual cyclicity (seasonality) and trans-year (13 to <24 months) and/or multiannual (long-term, ⩾24 months) cycles might be present. For this purpose, a relatively long monthly flu incidence dataset consisting of absolute numbers of new cases from the Grand Baku area, Azerbaijan, for the years 1976-2000 (300 months) was analysed. The exploration of underlying chronomes or, time structures, was done by linear and nonlinear parametric regression models, autocorrelation, spectral analysis and periodogram regression analysis. We analysed temporal dynamics and described multicomponent cyclicity, determining its statistical significance. The analysis, considering the flu data specifically stratified in three distinct intervals (1976-1990, 1991-1995, 1996-2000), and also combinations thereof, indicated that the main cyclic pattern was a seasonal one, with a period of T = 12 months. Further, a number of multiannual cycles with periods T in the ranges of 26-36, 62-85 or 113-162 months were observed, i.e. average periods of 2·5, 6·1 and 11·5 years, respectively. Indeed, most of these cycles correspond to similar cyclic parameters of HGA and further analyses are warranted to investigate such relationships. In conclusion, our study revealed the presence of multicomponent cyclic dynamics in influenza incidence by using relatively long time-series of monthly data. The specific cyclic patterns of flu incidence in Azerbaijan allows further, more specific modelling and correlations with environmental factors of similar cyclicity, e.g. HGA, to be explored. These results might contribute more widely to a better understanding of influenza dynamics and its aetiology as well as to the derivation of more precise forecasted estimates for planning and prevention purposes.
在各个国家都观察到了流感发病率的多组分周期性（例如周期T = 1、2 - 3、5 - 6、8.0、10.6 - 11.3、13、18 - 19年），并且如所表明的，其与自然环境现象中的周期（如气象因素和日地物理活动（HGA））极为相似。本报告旨在通过探究流感年周期（季节性）以及跨年（13至<24个月）和/或多年（长期，⩾24个月）周期是否可能存在，来验证先前关于流感发病率周期性模式的结果。为此，分析了一个相对较长的月度流感发病率数据集，该数据集由阿塞拜疆巴库大区1976 - 2000年（300个月）新病例的绝对数量组成。通过线性和非线性参数回归模型、自相关、谱分析和周期图回归分析来探索潜在的时间节律或时间结构。我们分析了时间动态并描述了多组分周期性，确定其统计学意义。考虑到流感数据按三个不同时间段（1976 - 1990年、1991 - 1995年、1996 - 2000年）以及它们的组合进行了具体分层的分析表明，主要的周期性模式是季节性的，周期为T = 12个月。此外，还观察到了一些多年周期，其周期T在26 - 36、62 - 85或113 - 162个月范围内，即平均周期分别为2.5、6.1和11.5年。实际上，这些周期中的大多数与HGA的类似周期参数相对应，因此有必要进行进一步分析以研究此类关系。总之，我们的研究通过使用相对较长的月度数据时间序列揭示了流感发病率中多组分周期动态的存在。阿塞拜疆流感发病率的特定周期性模式使得能够进一步进行更具体的建模，并探索与具有类似周期性的环境因素（如HGA）的相关性。这些结果可能更广泛地有助于更好地理解流感动态及其病因，以及为规划和预防目的得出更精确的预测估计。
