|
Abstract:
|
Seasonal infections exhibit marked temporal fluctuations, or cyclic periods of low and high disease incidence. The most commonly used non-parametric approach to characterizing seasonality involves grouping data into categories that reflect time periods (e.g., months, quarters, and seasons) which may reflect annual weather fluctuations. This approach is simple and easy to interpret, but the use of coarse groupings prevents a fully detailed, accurate, and comprehensive analysis of seasonal patterns. Inclusion of finer seasonal categories could lead to over-parameterization of the model. Parametric models, such as a harmonic regression, overcome these problems and allow characterization of essential aspects of seasonality: a point in time when the seasonal curve reaches its maximum (i.e., peak timing), and the amplitude of that peak. In this study, we show the value of using harmonic regression to assess seasonality of infectious diseases. We focus on characterizing peak timing and amplitude, with their respective confidence intervals, using data collected by longitudinal cohort studies, hospitalization records, and surveillance systems to understand underlying reasons of seasonality.
|
