This study demonstrates a positive association between DENV-infectious Ae. aegypti
and DENV-infected children living in the same and neighboring houses. Spatiotemporal clustering of DENV infection in children and mosquitoes was detected at a fine scale, consistent with focal aggregation well within a 100-meter radius area. Houses with infectious mosquitoes had an especially high risk (47.4%) of human DENV infection along with elevated measurements of mosquito density; neighboring houses also had elevated risk of human infection. Our results are consistent with the notion that houses with high DENV transmission risk contribute disproportionately to virus amplification and spread. Infections followed a pattern of over-dispersion, which has been reported for other infectious diseases to include indirectly transmitted, mosquito-borne infections 
. At a given point in time, people and mosquitoes in a relatively small portion of houses were responsible for the majority of DENV transmission.
Our results are the first to demonstrate a direct relationship between DENV infection in humans and mosquitoes at very fine spatiotemporal scales in the natural setting. Other researchers have reported heterogeneity of human DENV infection across space and time 
. Many entomological studies have shown the limited flight range and preferential and frequent human feeding behavior of Ae. aegypti
that would be expected to enhance DENV transmission 
. Prior studies of DENV infections in mosquitoes tended to focus on mosquitoes collected in or around houses of people with dengue-like illness 
; and when these studies were done across communities, infected mosquitoes were not explicitly linked to human infection 
. Perhaps because of the difficulty in collecting adult Ae. aegypti
, there has been relatively little research done on mosquito DENV infections in relation to human infection dynamics. Our study expands on this picture by showing that human and mosquito infections are positively associated with each other at small geographic and temporal scales. The strongest association was at the level of the individual house.
We did not directly evaluate the role of human adults in DENV transmission. It is possible that spatiotemporal dynamics of DENV transmission is different in adults and children, perhaps due to age-specific differences in existing immunity, the rate at which they are bitten 
, or in their movement patterns and exposure to daytime-biting Ae. aegypti
. We would not expect our overall conclusions to change, however, because both adults and children would have contributed to our findings whether or not adults were separately evaluated.
Fine scale spatial aggregation of DENV transmission may persist for three weeks or longer. Given the estimated time of infection of RD, ES and PES infections and because all of these categories of infection appeared to show focal aggregation within the ≤100-meter radius of the clusters, the spatial pattern we detected could have been present for greater than three weeks. This pattern, which is similar to what was observed for DENV-infected Ae. aegypti
in households in Mexico 
, may have persisted for a longer period if not truncated by the vector control interventions instituted on day 1 (by the study team) and day 7 (by the MOPH) of the cluster investigations. The lack of focal aggregation among day 15 PCR-positive child contacts supports this notion, although the small number of those infections may have been insufficient for a meaningful analysis.
Our testing method favored identification of PCR-positive mosquitoes that were infectious. The DENV incubation period in mosquitoes from the time that they imbibe an infectious blood meal to the time they become infectious (i.e., extrinsic incubation period) typically lasts for 10–14 days under environmental conditions like those in Kamphaeng Phet, Thailand 
. This implies that DENV-infectious mosquitoes in our study fed on an infected human considerably earlier than the time of cluster initiation and, thus, the transmission chain in houses with infectious mosquitoes had been taking place for some time before the “index” case was detected and the cluster investigation initiated. Consequently, as with infected children in the clusters, focal aggregation of infectious mosquitoes within the clusters may have been going on for two weeks or longer prior to initiation of each cluster investigation. So although “index” cases were used to initiate cluster investigations, they were not necessarily the first infection to occur within the cluster. Again, because vector control measures were instituted on day 1 and 7 and no further entomological collections were performed afterwards, we were not able to determine how long the focal pattern of DENV infection in mosquitoes would have persisted. We speculate that the duration of these focal areas of higher risk is limited more by the availability of susceptible humans than by susceptible mosquitoes. Future studies could investigate the required duration of interventions, which may need to be continued for one month or more.
Significantly more Ae. aegypti
pupae and adult females were collected from houses containing infectious mosquitoes than from those without. In addition, the risk of DENV infection in children was high in houses with infectious mosquitoes and, notably, remained elevated in neighboring houses. The higher entomological indices, however, were detected only in houses that actually contained infectious mosquitoes. These findings indicate that certain individual houses with high DENV transmission risk may disproportionately contribute to virus transmission within neighboring houses, likely due to local human and mosquito movement. Our study did not specifically evaluate when these elevated entomological measurements began or how long they persisted. They could have been present for some time prior to detection. Therefore, even in clusters with high DENV transmission, there may be individual houses that are responsible for the bulk of the transmission risk. Dengue management interventions that fail to include these individual, high-risk houses may have less impact than expected on reducing overall DENV spread. Similarly, surveillance programs that average measurements or indices of risk over a large area may fail to detect individual high-risk houses that disproportionately contribute to persistence and expansion of local transmission 
Fine scale spatiotemporal clustering of human-mosquito DENV transmission supports the hypothesis that DENV spread to more distant locations is driven by human movement 
. Whether DENV is successfully transmitted at those distant locations is likely related to a suite of factors including susceptibility of the local human population, mosquito vector density and infection status, vector competence, degree of human-vector contact, and intrinsic virus factors. Locations with high levels of human movement and potential for high interaction between people and mosquitoes merit additional investigation. These components of transmission may need to be factored into dengue surveillance and control efforts more than is currently being done 
Results from our study have implications for strategies to prevent DENV transmission. Transmission models that address DENV spread and the impact of vaccines alone or in combination with vector control need to account for the spatiotemporal scale and dynamics of DENV transmission. Depending on the questions being asked, these models and the interpretation of surveillance data that feed into them will need to account for the presence of high-risk hotspots of human-vector virus exchange that have a high impact on DENV spread to surrounding areas 
. These efforts should be integrated into an overall multifaceted strategy that takes into account DENV spread by movement of viremic humans among focal areas of concentrated, high levels of transmission.