Colombia is an endemic country for dengue fever where the four serotypes of virus dengue (DENV1–4) circulate simultaneously, and all types are responsible for dengue cases in the country. The control strategies are guided by entomological surveillance. However, heterogeneity in aedic indices is not well correlated with the incidence of the disease in cities such as Riohacha, Bello and Villavicencio. As an alternative, molecular detection of dengue virus in mosquitoes has been proposed as a useful tool for epidemiological surveillance and identification of serotypes circulating in field. We conducted a spatiotemporal fieldwork in these cities to capture adult mosquitoes to assess vector infection and explain the differences between Breteau indices and disease incidence. DENV infection in females and DENV serotype identification were evaluated and infection rates (IR) were estimated. The relationship between density, dengue cases and vector index was also estimated with logistic regression modeling and Pearson’s correlation coefficient. The lack of association between aedic indices and dengue incidence is in agreement with the weak associations between the density of the mosquitoes and their infection with DENV in the three cities. However, association was evident between the IR and dengue cases in Villavicencio. Furthermore, we found important negative associations between temperature and lag time from two to six weeks in Riohacha. We conclude that density of mosquitoes is not a good predictor of dengue cases. Instead, IR and temperature might explain better such heterogeneity.