Scaling properties and symmetrical patterns in the epidemiology of rotavirus infection
Authors:
Open Access
- 29 October 2003
- journal article
- review article
- Published by The Royal Society in Philosophical Transactions of the Royal Society B: Biological Sciences
- Vol. 358 (1438) , 1625-1641
Abstract
The rich epidemiological database of the incidence of rotavirus, as a cause of severe diarrhoea in young children, coupled with knowledge of the natural history of the infection, can make this virus a paradigm for studies of epidemic dynamics. The cyclic recurrence of childhood rotavirus epidemics in unvaccinated populations provides one of the best documented phenomena in population dynamics. This paper makes use of epidemiological data on rotavirus infection in young children admitted to hospital in Melbourne, Australia from 1977 to 2000. Several mathematical methods were used to characterize the overall dynamics of rotavirus infections as a whole and individually as serotypes G1, G2, G3, G4 and G9. These mathematical methods are as follows: seasonal autoregressive integrated moving-average (SARIMA) models, power spectral density (PSD), higher-order spectral analysis (HOSA) (bispectrum estimation and quadratic phase coupling (QPC)), detrended fluctuation analysis (DFA), wavelet analysis (WA) and a surrogate data analysis technique. Each of these techniques revealed different dynamic aspects of rotavirus epidemiology. In particular, we confirm the existence of an annual, biannual and a quinquennial period but additionally we found other embedded cycles (e.g. ca . 3 years). There seems to be an overall unique geometric and dynamic structure of the data despite the apparent changes in the dynamics of the last years. The inherent dynamics seems to be conserved regardless of the emergence of new serotypes, the re-emergence of old serotypes or the transient disappearance of a particular serotype. More importantly, the dynamics of all serotypes is multiple synchronized so that they behave as a single entity at the epidemic level. Overall, the whole dynamics follow a scale-free power-law fractal scaling behaviour. We found that there are three different scaling regions in the time-series, suggesting that processes influencing the epidemic dynamics of rotavirus over less than 12 months differ from those that operate between 1 and ca . 3 years, as well as those between 3 and ca . 5 years. To discard the possibility that the observed patterns could be due to artefacts, we applied a surrogate data analysis technique which enabled us to discern if only random components or linear features of the incidence of rotavirus contribute to its dynamics. The global dynamics of the epidemic is portrayed by wavelet-based incidence analysis. The resulting wavelet transform of the incidence of rotavirus crisply reveals a repeating pattern over time that looks similar on many scales (a property called self-similarity). Both the self-similar behaviour and the absence of a single characteristic scale of the power-law fractal-like scaling of the incidence of rotavirus infection imply that there is not a universal inherently more virulent serotype to which severe gastroenteritis can uniquely be ascribed.Keywords
This publication has 41 references indexed in Scilit:
- Rotavirus Gastroenteritis in Italian Children: Can Severity of Symptoms Be Related to the Infecting Virus?Clinical Infectious Diseases, 2001
- Epidemic Spreading in Scale-Free NetworksPhysical Review Letters, 2001
- Possible Association of Intussusception With Rotavirus VaccinationPediatrics, 1999
- Power laws governing epidemics in isolated populationsNature, 1996
- Mosaic organization of DNA nucleotidesPhysical Review E, 1994
- Epidemiological model of diarrhoeal diseases and its application in prevention and controlVaccine, 1994
- Finite-size effects on long-range correlations: Implications for analyzing DNA sequencesPhysical Review E, 1993
- Cohort study of rotavirus serotype patterns in symptomatic and asymptomatic infections in Mexican childrenThe Pediatric Infectious Disease Journal, 1993
- ARMA parameter estimation using only output cumulantsIEEE Transactions on Acoustics, Speech, and Signal Processing, 1990
- Estimating the Dimension of a ModelThe Annals of Statistics, 1978