Detecting space-time clusters of dengue fever in Panama after adjusting for vector surveillance data

Abstract

Long term surveillance of vectors and arboviruses is an integral aspect of disease prevention and control systems in countries affected by increasing risk. Yet, little effort has been made to adjust space-time risk estimation by integrating disease case counts with vector surveillance data, which may result in inaccurate risk projection when several vector species are present, and little is known about their likely role in local transmission. Here, we integrate 13 years of dengue case surveillance and associated Aedes occurrence data across 462 localities in 63 districts to estimate the risk of infection in the Republic of Panama. Our space-time modelling approach detected the presence of five clusters, which varied by duration, relative risk, and spatial extent after incorporating vector species as covariates. Dengue prevalence (n = 49,910) was predicted by the presence of resident Aedes aegypti alone, while all other covariates exhibited insignificant statistical relationships with it, including the presence and absence of invasive Aedes albopictus. Furthermore, the Ae. aegypti model contained the highest number of districts with more dengue cases than would be expected given baseline population levels. This implies that arbovirus case surveillance coupled with entomological surveillance can affect cluster detection and risk estimation, improving efforts to understand outbreak dynamics at national scales. Author SummaryDengue cases have increased in tropical regions worldwide owing to climate change, urbanization, and globalization facilitating the spread of Aedes mosquito vectors. National surveillance programs monitor trends in dengue fever and inform the public about epidemiological scenarios where outbreak preventive actions are most needed. Yet, most estimations of dengue risk so far derive only from disease case data, ignoring Aedes occurrence as a key aspect of dengue transmission dynamic. Here we illustrate how incorporating vector presence and absence as a model covariate can considerably alter the characteristics of space-time cluster estimations of dengue cases. We further show that Ae. aegypti has likely been a greater driver of dengue infection in high risk districts of Panama than Ae. albopictus, and provide a discussion of possible public health implications of both spatial and non-spatial model outcomes.