Climate-based descriptive models of dengue fever: The 2002 epidemic in Colima, Mexico

Dengue is a public health problem on the rise in many tropical regions and affects approximately 100 million people every year worldwide. In this paper, the authors retrospectively assess the association between five climatological variables and dengue incidence using data from the 2002 dengue epidemic in Colima, Mexico. Pluvial precipitation (mm), evaporation (mm), and mean, maximum, and minimum temperatures (degrees C) were obtained from local meteorological stations. The highest correlation of dengue incidence with maximum temperature was found at a lag of one month, and the highest correlation for evaporation was found at a lag of three months. A multiple-linear-regression model that includes all the climatological variables was correlated with 94 percent of the observed variance. Two simpler linear models with variables significant at the 99 percent confidence level were correlated with 88 percent (Precipitation + Evaporation) and 79 percent (Precipitation + Maximum Temperature) of the observed variance.

References:

• G. Chowell, P. Diaz-Duenas, J.C. Miller, P.W. Fenimore, J.M. Hyman, C. Castillo-Chavez. Estimation of the reproduction number of dengue fever from spatial epidemic data. Math. Biosci. 2007 Aug;208(2):571-89 (pdf)
• G. Chowell, F. Sanchez. Climate-based descriptive models of dengue fever. J. Env. Health 68 (10) Jun, 60-3 (2006) (pdf)
• G. Chowell , C.A. Torre, C. Munyaco-Escate, L. Suárez -Ognio, R. López-Cruz, J.M. Hyman, C. Castillo-Chavez. Spatial and temporal dynamics of dengue fever in Peru: 1994-2006.  Epidemiology and Infection, (to appear 2008).

Figure 1. Lagged cross-correlation of the number of dengue cases with A) precipitation, B) mean temperature, C)minimum temperature, D)maximum temperature, and E) evaporation. While precipitation, temperature, and minimum temperature have their highest correlation with dengue incidence without a lag period(0 lag), maximum temperature and evaporation were most higly correlated with dengue incidence at lags of three months, respectively.

Figure 2. The epidemic data are indicated by circles, the solid line represents the full model fit with the five climatological variables included, the dashed line shows the model fit with precipitation and evaporation included(both significant at the 99% level), and teh dash-dot line shows the model fit with precipitation and maximum temperature as predictors(both significant at the 99% level).