The development of an early warning system for climate‐sensitive disease risk with a focus on dengue epidemics in Southeast Brazil

Abstract

Previous studies have demonstrated statistically significant associations between disease and climate variations, highlighting the potential for developing climate-based epidemic early warning systems. However, limitations to such studies include failure to allow for non-climatic confounding factors, limited geographical/temporal resolution, or lack of evaluation of predictive validity. Here, we consider such issues in the context of dengue fever in South East Brazil, where dengue epidemics impact heavily on Brazilian public health services. A spatio-temporal generalised linear mixed model (GLMM) is developed, including both climate and non-climate covariates. Overdispersion and unobserved confounding factors are accounted for via a Negative Binomial formulation and inclusion of both spatial and temporal random effects. Model parameters are estimated in a Bayesian framework to allow full posterior predictive distributions for disease risk to be derived in time and space. Detailed probabilistic forecasts can then be issued for any pre-defined 'alert' thresholds, allowing probabilistic early warnings for dengue epidemics to be made. Using this approach with the criterion 'greater than a 50% chance of exceeding 300 cases per 100,000 inhabitants', successful epidemic alerts would have been issued for 81% of the 54 regions that experienced epidemic dengue incidence rates in South East Brazil, during the major 2008 epidemic. Use of seasonal climate forecasts in this model allows predictions to be made several months ahead of an impending epidemic. We argue that the general modelling framework, described here in the context of dengue in Brazil, is potentially valuable in similar applications, both outside of Brazil and for other climate-sensitive diseases.

Climate and Non-Climate Drivers of Dengue Epidemics in Southern Coastal Ecuador

“…30,38 A negative binomial model was used to account for overdispersion found in the dengue count data (extra-Poisson variation), 39,40 where y t is monthly dengue cases, μ t is mean cases, e t is expected cases, and ρ t is the dengue relative risk (Eq. 1).…”

Quantifying the added value of climate information in a spatio-temporal dengue model

“…A negative binomial model was used to account for over-dispersion found in the dengue count data (extra-Poisson variation), where "# is monthly dengue cases, "# is mean cases, "# is expected cases, and "# is the dengue relative risk (Eqn. 1) (Lowe et al, 2013a;Stewart-Ibarra and Lowe, 2013). By including the expected number of cases of dengue as an offset, we estimated the relative risk (SMR) of dengue using a combination of spatio-temporal structures and linear and nonlinear functions of climate.…”

“…This was incorporated via a first order autoregressive latent model, where dengue relative risk in one month is allowed to depend on the relative risk in the previous month. We then included spatial structure using a convolution prior that combined area-specific overdispersion and a neighbourhood dependency structure (see (Besag et al, 1995;Lowe et al, 2013a) for details), which we termed the 'seasonal-spatial' model. This model accounts for temporal dependency from one month to the next and similarities or differences between neighbouring provinces, but no interannual variability.…”

Temperature and population density determine reservoir regions of seasonal persistence in highland malaria

“…[20,21]). We used here a lagged relative risk computed as the log-standardized ratio of observed SOND cases to the global expected seasonal malaria cases.…”

“…We used here a lagged relative risk computed as the log-standardized ratio of observed SOND cases to the global expected seasonal malaria cases. The latter is the population in a kebele multiplied by a global estimate of the average seasonal malaria rate, a value estimated by dividing the total number of cases by the total population across all kebeles [20,21].…”