The spatiotemporal dynamics of infectious diseases in endemic settings often are poorly characterized. In settings where infections are continuously occurring, it is difficult to elucidate transmission chains and the movement of pathogens in space. The field of phylogeography presents a potential solution. By combining the spatiotemporal location of individuals who become infected with the genetic sequence of the infecting pathogen, it is possible to track individual viral lineages as they move through a region. This information can be extremely useful for mounting responses and for understanding the mechanisms that create and maintain pathogen diversity. This doctoral dissertation research project will use geocoded home addresses and infecting serotypes of 6,659 patients who were diagnosed with one of the four serotypes of dengue fever at a Bangkok hospital between 1995 and 2010. In addition, the genome of the infecting virus will be sequenced for a subset of cases from 2006. The doctoral student will analyze spatiotemporal dependence between dengue cases using adapted space-time statistics to understand how far, both in distance and space, cases from the same viral lineage tend to occur from each other. Furthermore, the clustering behavior of cases across time lags will be characterized to describe the changing patterns of spatial dependence at small spatial scales through time that could be induced through serotype-specific community immunity effects.
Dengue fever is a potentially life-threatening mosquito-borne viral disease that causes at least 36,000,000 symptomatic cases every year across the globe. By understanding the spatiotemporal clustering of individuals infected with viruses from the same lineage, this project will provide insight into what may be driving the dispersal of the disease. Furthermore, understanding the spatial and temporal extent at which transmission-related cases occur will facilitate the identification of populations at risk of infection and will provide guidance regarding the spatial distances at which to implement insecticide spraying of neighboring homes upon detection of an index case. This project will provide an opportunity to understand the potential drivers of disease ecology in a large endemic urban setting that have become the key location for dengue infections. The approaches developed in this project will be generalizable to other spatiotemporal point patterns for which there are heterogeneities in the labels attached to points (such as genotype or species) or in dynamic systems where there are changing patterns in underlying spatiotemporal dependence. As a Doctoral Dissertation Research Improvement award, this award also will provide support to enable a promising student to establish a strong independent research career.
Dengue is a mosquito-transmitted virus that causes significant morbidity and mortality in many parts of global tropical and sub-tropical regions. Intervention measures are currently limited to mosquito control efforts, such as the spraying of insecticides. These activities are resource intensive. Understanding where other cases of dengue are likely to occur upon detection of an infected individual is therefore crucial to the efficient deployment of such interventions. Characterizing the spatial clustering between cases of dengue is complicated by a number of factors. Firstly, like many other diseases, individuals often do not develop symptoms when they become infected by dengue virus or do not get sick enough to visit a healthcare practitioner. It is estimated that fewer than five percent of dengue infections result in healthcare visits. There may also be spatial differences in the tendency for sick individuals to visit healthcare providers. Finally, many places are endemic for dengue resulting in many unrelated transmission chains overlapping in space and time. In this project we developed novel methodologies that use information on the infecting pathogen to characterize the spatial and temporal dependence between cases even in settings where these complicating factors exist. There are four broad groups of dengue viruses called serotypes that often co-circulate. It is possible to identify the dengue serotype an individual has been infected with. As an infectious individual infected with one serotype can only infect subsequent individuals (via mosquitoes) with the same serotype, we can compare the spatial separation between individuals infected with the same serotype with those infected with any serotype to estimate the spatial dependence between potentially transmitted-related individuals adjusting for the underlying tendency of any case to be detected by the surveillance system. We built a range of individual based models that simulated disease transmission processes with four different serotypes representing dengue transmission. We found that we could recover the true underlying spatial dependence between cases, even when only one percent of cases were observed and we imposed significant spatial and temporal biases in the observation process. Having demonstrated the robustness of our approach, we applied it to dengue case data from a Bangkok hospital. We used the geocoded home location of over 14,000 dengue cases, hospitalized between 1994 and 2006. We found that significant spatial dependence was observed between dengue cases occurring within the same month at distances up to 1km, suggesting dengue transmission results in a large footprint of case distribution at any time point. This also suggests that the scales of co-location may be too large for spatially targeted interventions such as insecticide spraying to be effective. As there may be several different transmission chains of the same serotype circulating at any time, we are currently exploring the consistency of our results when we use even greater detail on the infecting pathogen: the full genetic sequence of the virus. By comparing the sequence of the virus infecting individuals, we can identify pairs of individuals that were part of the same transmission chain. We have sequenced the viral genome from 200 of the dengue cases from a single serotype and will be estimating the consistency of our findings to this greater level of detail. Our approaches are relevant across disease and non-disease systems. To facilitate its use, we have developed and will shortly release a freely available R package that will allow researchers and students to employ these methods across disciplines.
