Dengue in Bangladesh: assessment of the influence of climate and under-reporting in national incidence

Abstract

Dengue occurs in many tropical countries, despite substantial effort to control the Aedes mosquitoes that transmit the virus. The majority of the burden occurs in the South-East Asian Region of the World Health Organization. Bangladesh is a lower-middle income country located in South Asia, with strong seasonal weather variation, heavy monsoon rainfall, and high population density. Dengue has been endemic in Bangladesh since an epidemic in 2000. The aim of my research was to investigate the influence of climate on dengue transmission in Bangladesh over the period January, 2000 - December, 2009. To achieve this aim, I conducted a series of studies integrating epidemiological and socio-environmental factors into a unified statistical modelling framework to better understand transmission dynamics.

In a narrative review (Chapter 3), I discuss the emergence and establishment of dengue along with the possibility of future epidemics of severe dengue. Introduction of a dengue virus strain from neighbouring Thailand likely caused the first epidemic in 2000. Cessation of dichlorodiphenyltrichloroethane (DDT) spraying, climatic, socio-demographic, and lifestyle factors also contributed to epidemic transmission and endemic establishment of the virus. However, there has been a decline in reported case numbers following the largest epidemic in 2002, albeit with relatively greater case numbers in alternate years. This occurred despite the absence of significant additional control measures and no changes in the surveillance system having been introduced during the study period. The observed decline from 2002 may be an artefact of the national hospital-based passive surveillance system even though a real decline in incidence could plausibly have occurred due to increased prevalence of immunity, greater public awareness, and reduced mosquito breeding sites.

From a temporal negative binomial generalised linear model (Chapter 4), developed using monthly dengue cases in Dhaka from January, 2000 - December, 2009, I identify that mean monthly temperature (coefficient estimate: 6.07; 95% confidence interval: 3.38, 8.67) and diurnal temperature range (coefficient estimate: 15.57; 95% confidence interval: 8.03, 22.85) influence dengue transmission, with significant interaction between the two (coefficient estimate: -0.56; 95% confidence interval: -0.81, -0.29), at a lag of one month in Dhaka, the capital city of Bangladesh where the highest number of cases were reported during the study period. In addition to mean monthly rainfall in the previous two months, dengue incidence is associated with sea surface temperature anomalies in the current and previous months through concomitant anomalies in the annual rainfall cycle. Population density is also significantly associated with increased dengue incidence in Dhaka.

Chapter 5 reports an investigation into non-linear dengue–climate associations using the same dataset as used for the previous model in Chapter 4. A Bayesian semi-parametric thin-plate spline approach estimates that the optimal mean monthly temperature for dengue transmission in Dhaka is 29oC and that average monthly rainfall above 15mm decreases transmission. This study also reveals that between 2000 and 2009 only 2.8% (95% Bayesian credible interval 2.7- 2.8) of cases estimated to have occurred in Dhaka were reported through passive case detection.

A Bayesian spatio-temporal model (Chapter 6), formulated using monthly dengue cases reported across the country from January, 2000 - December, 2009, identifies that the majority of dengue cases occur in southern Bangladesh with the highest in Dhaka (located almost in the middle of the country), accounting for 93.0% of estimated total cases across the country from 2000-2009. Around 61.0% of Bangladeshi districts are identified as affected with dengue virus during the high transmission season of August and September, contrasting with national surveillance data suggesting that only 42.0% of districts are affected.

My thesis provides a better understanding of the dengue-climate relationships that will enable more accurate predictions of the likely impacts of changing climate on dengue risk. Knowledge about the extent of under-reporting will facilitate precise estimation of dengue burden which is vital to assess the risk of severe epidemics. These will help public health professionals to design interventions to strengthen the country’s capacity for prevention of severe dengue epidemics.