Shifting host choices of the vector of Chagas disease, Triatoma infestans, in relation to the availability of hosts in houses in north-west Argentina.
Data from a 4-year study on the blood meal sources of 1964 specimens of T. infestans collected from human sleeping places in 3 rural villages in north-west Argentina were analysed using individual houses as the basic sampling unit. Logistic multiple regression analysis was used to test whether the proportion of domiciliary T. infestans that fed on humans, dogs, chickens or cats was related to the number of these hosts in the house and to vector density. On average, dogs were 2.6× or 2.3× more likely to be selected than humans, relative to availability, in summer 1992 (at 23 houses) or spring summer 1988-89 (at 11 houses), respectively, but 0.8× less likely to be selected than humans, relative to their availability, in winter 1988 (at 12 houses). In the hot season, chickens were 4.8× more likely to be selected than humans, relative to their assumed availability. From data collected during the hot season, the proportion of bugs that fed on humans, among all bugs with any identified blood source (the human blood index, HBI), or on humans only (HBIO), decreased significantly as both the number of dogs sharing human sleeping places (room-mate dogs) and the proportion of bugs that fed on chickens increased. The proportion of bugs that fed partly on dogs, or on dogs alone, increased significantly with increasing numbers of room-mate dogs and decreased significantly with an increasing proportion of bugs feeding on chickens. From data collected during the winter, when fewer cases of feeding on chickens were detected and most people slept indoors, both HBI and HBIO increased significantly with the number of humans and decreased with the number of room-mate dogs. The effects of vector density on HBI or HBIO were investigated by multiple logistic regression analysis. In summer 1992, after a backward elimination procedure, both HBI and HBIO were significantly and negatively related to the density of domiciliary T. infestans per house, the number of room-mate dogs and the proportion of the bugs that fed on chickens. A marginally significant and negative relationship between HBI and the log-density of bugs was verified in 2 published datasets. Analysis of 13 published host-feeding patterns of domiciliary T. infestans, and of another 27 datasets from 11 other triatomine species collected inside or around bedroom areas, showed that HBI was inversely and significantly related to the proportion of bugs that fed on chickens (in both groups of vector species) or on dogs (for T. infestans). This study is the first to show that: (i) the likelihood of feeding partly on humans, or on humans only, significantly decreased in the presence of dogs and chickens in bedroom areas; (ii) the HBI was also negatively related to the density of bugs per house after controlling for the effects of chickens and room-mate dogs; and (iii) the likelihood of feeding on dogs increased significantly with more room-mate dogs and decreased with an increasing proportion of bugs feeding on chickens. Some epidemiological consequences of these relationships for transmission, control, zooprophylaxis and modelling are discussed.