Abstract:
Dengue infection is still a great burden and has epidemic potential in
northeastern Thailand. A retrospective analysis using various methods was applied to
this problem.
The objective was to determine the temporal patterns, propose forecasting
models, identify the spatial distributions and clusters, and determine the spatial
variations and spatiotemporal dynamics of dengue incidence in northeastern Thailand.
Data series for regional and provincial levels from 1996 to 2005 and for
district level from 1999 to 2005 were analyzed. The STL and Poisson regression
model were used in determining temporal patterns and trends of the disease. The
seasonal ARIMA was applied to propose the forecasting model. The SEB smoothing,
Global Morans I statistic and Moran LISA statistic were used in determining spatial
patterns and clusters of the disease. The GWPR and measures of synchrony were
performed to determine the spatial variations and spatiotemporal dynamics of the
disease.
Data analysis showed that the high peak period of dengue incidence in
northeastern Thailand seemed to occur every two year and last for two or three years.
The peaks occurred in June or July with the amplitude greater than the troughs.
Overall, the trends of dengue incidence have not changed over the years. This study
identified the best fit seasonal ARIMA model for each data series. Using the SEB
smoothing, the significant local clusters and clusters of high rates were shown to
occur mostly in the southwest part of the region at the beginning of the high peak
periods. Positively significant local relationships between population density and
dengue incidence were seen for most districts every year. Statistical synchrony was
observed in almost all paired provinces. The paired correlation of dengue incidence
statistically declined with distance between provinces and distance between pairs
explained 20.6% of the variation in pairwise synchrony. This research will be useful
in helping to predict the dynamics of dengue infection.