Factors associated with the occurrence of Triatoma sordida (Hemiptera: Reduviidae) in rural localities of Central-West Brazil

This study estimates the factors of artificial environments (houses and peridomestic areas) associated with Triatoma sordida occurrence. Manual searches for triatomines were performed in 136 domiciliary units (DUs) in two rural localities of Central-West Brazil. For each DU, 32 structural, 23 biotic and 28 management variables were obtained. Multiple logistic regression analysis was performed in order to identify statistically significant variables associated with occurrence of T. sordida in the study areas. A total of 1,057 specimens (99% in peridomiciles, mainly chicken coops) of T. sordida were collected from 63 DUs (infestation: 47%; density: ~8 specimens/DU; crowding: ~17 specimens/infested DU; colonisation: 81%). Only six (0.6%) out of 945 specimens examined were infected with Trypanosoma cruzi. The final adjusted logistic regression model indicated that the probability of T. sordida occurrence was higher in DU with wooden chicken coops, presence of > 30 animals in wooden corrals, presence of wood piles and presence of food storeroom. The results show the persistence of T. sordida in peridomestic habitats in rural localities of Central-West Brazil. However, the observed low intradomestic colonisation and minimal triatomine infection rates indicate that T. sordida has low potential to sustain high rates of T. cruzi transmission to residents of these localities.

Chagas disease (CD) is endemic in rural populations inhabiting structurally deficient households that favour the colonisation of triatomine bugs. In 2006, the Pan American Health Organization declared that Brazil was free from Trypanosoma cruzi transmission by the domestic vector Triatoma infestans (Dias 2007). However, native species, such as Panstrongylus megistus (Villela et al. 2009), Triatoma sordida (Oliveira & Silva 2007), Triatoma brasiliensis and Triatoma pseudomaculata (Silva et al. 2012), continue to be found in domestic and peridomestic environments. Understanding the factors associated with household infestation by these species may identify new targets for intervention and minimise the risk of T. cruzi vectorial transmission (Cohen & Gürtler 2001).
House structure influences colonisation by triatomine bugs. Proximity of houses to vegetation and the presence of livestock both increase the likelihood of household infestation by triatomines. In addition, peridomestic structures may play an important role in maintaining triatomine populations in close proximity to residences (Campbell-Lendrum et al. 2007, Gurevitz et al. 2011, Dumonteil et al. 2013, Bustamante et al. 2014. The factors associated with the occurrence of triatomine bugs were analysed in northeastern Brazil (Walter et al. 2005). However, these factors may vary in different regions of the country due to human behaviour, local vector ecology and socioeconomic conditions (Vinhaes et al. 2014). The identification of these factors in domestic and peridomestic environments is useful in order to prevent and control these bugs more effectively, thereby reducing the risk of vectorial transmission of CD (Weeks et al. 2013).
Traditionally, the most common synanthropic species captured in the Central-West Region of Brazil has been T. sordida (Pereira et al. 2013). This species occurs primarily in peridomestic environments, particularly chicken coops and therefore exhibit low rates of natural infection by T. cruzi (Forattini et al. 1975, Diotaiuti et al. 1995a, Oliveira & Silva 2007. Although several studies examining ecological and behavioural aspects of T. sordida have been published since the 1970s, the factors that determine its persistence in peridomiciliary environments need to be better understood. The relative importance of biotic, structural and environmental management factors influencing the occurrence of this species in Central-West Brazil has not been systematically evaluated; such research could result in more effective control procedures and lead to more reliable predictions of the likelihood of household infestation. The aim of this study therefore was to estimate the association between structural, biotic and environmental management factors and the presence of T. sordida in households in rural localities of the municipality of Posse, in the Brazilian state of Goiás (GO), Brazil.

MATERIALS AND METHODS
Study area -The study was conducted in rural areas in Posse where CD is the main protozoan infection (Oliveira & Silva 2007). Data from the seroprevalence survey of T. cruzi human infection conducted between [1975][1976][1977][1978][1979][1980] showed an infection prevalence of 7.4% in the state, considered one of the highest rates in Brazil (Silveira et al. 2011). More recently, data collected from the Ministry of Health's Mortality Information System revealed that CD was responsible for 3,589 deaths in GO between 2007-2011. According to Martins-Melo et al. (2012, GO had the highest mortality rates resulting from CD in Brazil, mainly in the northeastern region of the state. Moreover, vector-borne acute cases of CD were confirmed in GO between 2006-2012, one of them in Posse (Vinhaes et al. 2014). However, the last attempt at vector-control in this municipality occurred in 2008.
The municipality of Posse is located in northeastern GO (Fig. 1A), approximately 320 km from Brasília, Federal District. It encompasses an area of 1,949.63 km 2 and supports a population of approximately 31,257 individuals. This municipality is located in the Cerrado biome, where the climate is characterised by two well-defined seasons: the rainy season (October-March) and the dry season (April-September). High rates of triatomine infestation were observed in Posse in the 1980s and 1990s (Diotaiuti & Pinto 1991). Moreover, Posse was the last municipality in GO where T. infestans was recorded (Oliveira & Silva 2007) and systematic control programs have been inactive since 2008. Two localities within Posse (Trombas and Periquito) were selected because they had the highest infestation rates (11.6% and 5.5%, respectively) among all localities investigated in 2008.
Triatomine collection and parasite detection -The triatomine survey was conducted in April 2013. After resident permission was obtained, systematic manual triatomine searches in the domiciliary units (DUs) (the house itself plus any peridomiciliary annex buildings and the space between all such structures) were conducted by a team of two trained individuals equipped with gloves, flashlights  and tweezers. The triatomine search required about 1 h per DU. Three visiting attempts on different days were performed when the resident was absent, after which the house was considered uninhabited or abandoned. The intradomicile inspection included all rooms in the DU. The walls, beds and other furniture, the spaces behind posters and frames, clothes baskets, accumulations of wood or bricks and any other sites that could represent a suitable refuge for insects were examined. The peridomiciliary environment was defined as the area surrounding the home, usually comprised of a fenced compound, regardless of the distance from the main house. In this environment, animal shelters (e.g., corrals, chicken coops and pig pens), bricks, wood, tiles and rocks were examined.
All DUs were georeferenced using a Garmin GPS 12 (GARMIN International ® , USA) receiver. The spatial distribution of the DUs was determined after importing the geographical coordinates of each DU (latitude and longitude) into ArcGIS software (v.10.2). Positive and negative DUs in each location were then overlaid with satellite imagery.
Collected insects were stored alive in vials containing filter paper and properly labelled. The insects were separated by sex and nymphal stage and morphologically identified using the taxonomic keys described by Lent and Wygodzinsky (1979). Faeces of the collected triatomines were then examined for the presence of flagellates via direct microscopical observation. Natural infection was determined by examination of fresh faeces obtained by abdominal compression of the triatomine bugs. Parasites were morphologically identified by microscopical observation of Giemsa-stained insect faeces. Finally, standard entomological indicators (e.g., infestation, colonisation, density and crowding) (WHO 1991, Dias & Diotaiuti 1998 were estimated for each locality and habitat. Characterisation of DUs (houses and peridomestic area) -For each DU, a form was filled out describing the features of the houses and of the peridomiciliary environment. The selection of the variables was based on previous studies analysing factors associated with triatomine bug infestation (Walter et al. 2005, 2007, Black et al. 2007, Campbell-Lendrum et al. 2007, Bustamante et al. 2009). In total, 32 structural, 23 biotic and 28 management variables (Table I) were included in our analysis.
Statistical analysis -Descriptive statistics were estimated in relation to all DU variables. Crude and adjusted associations between explanatory variables and the dependent variable (presence or absence of T. sordida in DUs) were assessed using bivariated and multiple logistic regression models, using the PASW Statistics 18 program. A final adjusted model based on multiple logistic regression was estimated using the backwards stepwise method for variable selection, as follows: (i) Pearson correlation matrix was used to identify collinearity among independent variables. Variables with high correlation coefficients (r ≥ 0.8) were considered collinear and only one was selected based on prior knowledge. (ii) Based on bivariated analysis (1 independent variable at a time), some variables were re-categorised to avoid collinearity and/or to improve statistical power by collapsing similar categories. (iii) Also based on bivariated analyses, independent variables whose association with T. sordida occurrence in DUs resulted in p-values < 0.20 were eligible for the multiple logistic regression analysis. (iv) Using the backwards variable selection approach, the least significant variables were excluded (1 by 1) from the model, until all variables (or some of its categories) were statistically associated with T. sordida occurrence at the p < 0.05 level.

RESULTS
In total, 134 DUs were surveyed, composed of 70 (52.2%) in the locality of Trombas and 64 (47.8%) in the locality of Periquito. Some DUs were not surveyed because they were inhabited (24 in Trombas and 15 in Periquito) and/or closed (6 in Trombas and 2 in Periquito).
Most houses had full concrete walls (59%), ceramic tile (68%) and cement floors (84%). In the peridomiciles, the presence of wooden chicken coops (89%) and structures covered with asbestos tiles (46%) were very common. Most houses had a wire fence (52%) and peridomiciliary piles of material (85%) (e.g., wood, tile and bricks). The DUs features according to the variables related to structural, biotic and environmental management factors in each locality are present in Supplementary Tables I-III. In total, 1,059 triatomine specimens were collected, consisting primarily of T. sordida (99%). Only two stage V nymphs of Triatoma costalimai were captured in the peridomiciliary environment. The infestation rate of DUs by T. sordida was 47%, while 81% of DUs were colonised by this species (Table II). Although adult and stage V nymphs of T. sordida predominated, specimens of all developmental stages were captured (Table III).
T. sordida was found to be widely distributed within the research areas ( Fig. 1A, B). Periquito, closest to the urban centre of Posse, had a lower frequency of infested DUs, whereas the distribution of T. sordida in the localities showed that the DUs with the largest numbers of captured insects were located in Trombas (Fig. 1B), when compared to Periquito, where most of the DUs had densities lower than 20 individuals (Fig. 1C).
Only six specimens of T. sordida were infected with trypanosomatids in Trombas, corresponding to an infection rate of 0.6% (Table III). These trypanosomatids were morphologically similar to T. cruzi. The infected specimens were found in piles of tiles and in chicken coops.
In total, 692 ecotopes were surveyed, with chicken coops predominant (21.8%). The infestation rate by T. sordida was higher in chicken coops, where most of the insects were captured, resulting in higher density and crowding values. However, this species has been detected in other ecotopes, including barns, tool storage sheds, corrals, pigpens and other animal-shelter structures. The colonisation rate of T. sordida was higher in corrals and barns. Only four uninfected adult T. sordida were captured inside houses (Table IV).
The occurrence of T. sordida in DUs was associated with the presence of animals (cats, dogs and chickens), brick, wood, tile and rock piles surrounding the house, structure of the chicken coops and management variables such as distance of chicken coop from the house and cleanliness of peridomestic annexes (Supplementary Tables IV-VI).
Among the 34 variables selected to be included in the multiple logistic regression model, five remained statistically associated with T. sordida occurrence in the final model (Table V). The results indicate that the probability of occurrence of T. sordida is four times higher in Trombas and, in DUs with wooden chicken coops ( Fig. 2A, B), corrals holding > 30 animals, DUs with wooden piles in the peridomestic area ( Fig. 2C) and those with food storage structures.

DISCUSSION
This study estimated factors associated with T. sordida occurrence in rural localities in the Central-West Region of Brazil. The results showed that factors related to structural characteristics (e.g., wooden chicken coops and pens, wood piles around houses and food storages) and biotic (e.g., number of animals in a corral) are the factors that best explain the occurrence of this triatomine species.
The results obtained in this study revealed high peridomiciliary infestation rates by T. sordida, especially in chicken coops. Furthermore, no nymphs were captured inside the houses, indicating a small likelihood of house colonisation by T. sordida in the study areas. The dominance of this species in peridomiciles associated with the  . Even in some areas where there was a predominance of T. infestans in the 1970s (e.g., north of MG), after chemical control, T. sordida became the predominant species following control programs for T. infestans (Diotaiuti et al. 1995a). The high rate of occurrence of T. sordida in chicken coops is consistent with the known ornithophilic habits of this species and to the high availability of this food source in rural areas of Brazil (Forattini et al. 1982, Pires et al. 1999.
Only two nymphs of T. costalimai were collected during field research, with one specimen being found in a chicken coop and the other in a tool storage shed. Despite the infrequency of this species in this study, the identification of these nymphs may indicate colonisation of these ecotopes. It is important to note that the DU where the nymphs were collected was located near rock outcroppings, common habitats for this species (Mello 1981). T. costalimai has also been found in the peridomiciliary areas of other municipalities in GO (Oliveira & Silva 2007, Machiner et al. 2012. Dispersion of T. sordida can occur both passively (e.g., by transporting firewood to domiciles or by birds carrying nymphs during flight) and actively by flying adults (Forattini et al. 1971). According to Forattini et al. (1975), T. sordida is highly active compared to other triatomine species. These ecological studies, coupled with feeding and defecation dynamics studies (Diotaiuti et al. 1995b), suggested that T. sordida would play an important role in the transmission of T. cruzi. Noireau et al. (1997), on the other hand, demonstrated that the probability of T. cruzi transmission to humans by domiciliary T. sordida in Bolivia was low. Moreover, vectorial transmission of T. cruzi was strongly reduced after T. infestans control measures were implemented in Brazil, even with the permanence of T. sordida in rural areas (Silveira & Dias 2011). However, T. sordida can maintain T. cruzi cycles in peridomestic environments (Diotaiuti et al. 1995c). T. sordida was implicated as a potential vector of T. cruzi to humans in an oral outbreak of acute DC in BA. A colony of T. sor- dida was detected in the kitchen area of the house, where residents stored food; 50% of the triatomines were positive for T. cruzi and precipitin tests indicated association with birds, opossums, rodents and humans. Additionally, an entomological survey performed around the house revealed that 40% of the T. sordida specimens collected were infected with T. cruzi (Dias et al. 2008). These results indicate the importance of maintaining monitoring and control programs in areas infested with T. sordida.
Structures made of wood predominated in the peridomiciles in the study area. The results of the logistic regression showed that the presence of a wooden chicken coop adjacent to the house is a factor associated with T. sordida occurrence in the studied areas. One possible explanation is that, in the wild, this species is often found under tree bark and in hollow trees of the Cerrado biome (Forattini et al. 1971, Diotaiuti et al. 1993. The presence of wood piles as an associated factor for the occurrence of other triatomine species, such as T. pseudomaculata (Walter et al. 2005), Triatoma pallidipennis (Cohen et al. 2006) and Triatoma longipennis (Walter et al. 2007), has also been described, indicating the importance of proper environmental management of these breeding ecotopes for triatomine control. The association between storerooms and occurrence of Rhodnius prolixus (Campbell-Lendrum et al. 2007), T. infestans (Gurevitz et al. 2011) and Triatoma dimidiata (Bustamante et al. 2014) has also been recorded, suggesting that storerooms provide additional places for triatomine refuge. Food storage also provides resources for other animals, such as synanthropic rodents, which could act as blood sources for triatomines. In Brazil, the association between triatomines and corrals had been described for T. pseudomaculata in BA (Walter et al. 2005, Resgo et al. 2006).
Our results suggest that changes in the construction of chicken coops may decrease the occurrence of T. sordida. Improving peridomestic structures, removal of piles of material and cleaning the peridomicile also reduce food sources and hiding places for triatomines (Lucero et al. 2013). The present study recommends the replacement of wood with wire for chicken coop construction and increasing the distance between houses and the coops, in addition to more frequent cleaning of coops. Longitudinal studies that measure the rate of infestation of chicken coops before and after these actions would be useful for examining the effectiveness of this type of environmental management. Studies have shown that not only is environmental management effective in maintaining low triatomine infestation, but it also improves the effectiveness of insecticide spraying because it reduces the attractiveness of such ecotopes as sites for triatomine reproduction (Gorla et al. 2013, Lucero et al. 2013, Stevens et al. 2013. The improvement of chicken coops appears to be a promising strategy, considering the difficulties of controlling T. sordida in peridomestic environments using insecticides. According to Diotaiuti et al. (1998), spraying does not completely eliminate the bugs due to the low residual activity of the insecticide in peridomiciles.
Although the structural, biotic and environmental management features were similar between the two localities studied (Supplementary Tables), other variables that were not sampled could explain the differences between the entomological indicators of these locations. The higher infestation rates observed in Trombas could be a result of its greater distance from the administrative centre of Posse (which could hinder control activities) because it contains the lowest proportion of households that have undergone housing improvements (Supplementary Tables) and because of the history of land use in these locations (e.g., deforestation and agriculture). Insecticide  control performed by residents of Periquito may also explain the differences in DU infestation between the two locales. Future studies that include more locations could better assess differences in the occurrence of T. sordida and expand the inferences obtained in this study. Limitations of the present study included the relatively small number of localities sampled, the high number of DUs that were not sampled (47 houses were closed or uninhabited) and the method of sample collection. Manual searching is the standard method used for the detection of triatomine bugs, but its limited sensitivity can often result in underestimations of the values of entomo-logical indicators and detection error/bias may occur at low bug densities. Repeated surveys using baited traps may increase the detection sensitivity within houses (de Arias et al. 2012, Abad-Franch et al. 2014.
The results of this study suggest that T. sordida has low potential for sustaining high rates of T. cruzi transmission to residents of the studied areas, as in other areas with rural characteristics similar to those in Trombas and Periquito. However, due to the high infestation rates of the peridomestic ecotopes and to the observed associated factors, we recommend that routine and efficient entomological monitoring programs that are capable of detecting changes in the behaviour of the species be maintained, in order to identify any incipient household colonisation by T. sordida or other native species. leia Diotaiuti, Pedro Tauil, Marcos T Obara, Walter Souza Santos and Andrey José de Andrade, for reviewing the paper.       only variables that showed significant association are presented. CI: confidence interval; OR: odds ratio.