The present study confirmed that anopheline species composition can vary with different sampling methods. Only four species (An. philippinensis, An. minimus s.l., An. baimai, and An. annularis) were collected using human bait (HB), the most vital method for discriminating anthropophilic species. Anopheles baimai and An. annularis were dominant only in HB collections. Light traps (LTs) captured all species and can be considered the most neutral method for collecting mosquitoes. However, the differences between indoor or outdoor LT catches indicate that attention must be paid to where LTs are set. Six species were collected using pyrethrum spray (PS); however, more than 93% were An. vagus. Mosquito species composition differed between LT and PS collection methods, and only An. vagus and An. subpictus were collected more often by PS. These results verified the effectiveness of HB, which only collects anthropophilic species, and demonstrated that the application of PS is not particularly useful for collecting anthropophilic species.
The prevalence of malaria is influenced by host preferences of Anopheles species, but little information is available on such preferences in Bangladesh . In nature, the expression of host preference (selection of host) by a mosquito may depend on several extrinsic or intrinsic factors [33, 34]. In the present study, the HBI values of mosquitoes collected indoors and outdoors were 6.96% and 11.73%, respectively. The higher HBI in outdoor-collected mosquitoes suggests two possibilities: people may be bitten more frequently outdoors, or indoor-biting mosquitoes do not remain inside and instead exit at night. Both possibilities may indicate that mosquitoes are becoming exophagic or that exophagic traits are favored by insecticidal pressure of bed nets. Anopheles baimai, an efficient malaria vector in Bangladesh, usually rests and bites humans outdoors, even though it is highly anthropophilic [6, 35, 36]. The higher HBI in outdoor-collected anophelines along with the higher outdoor-biting tendency of malaria vectors in Bangladesh [6, 35] indicate that bed nets or other devices used indoors may not provide enough protection from vectors.
Anopheles species incriminated as malaria vectors exhibit preferences for humans . Ramsay et al. reported a substantial preference for human blood by An. minimus, which was supported in part by Toumanoff and Hu  in Vietnam. However, this species (collected outdoors) exhibited a low HBI in the Philippines  and India . We found that both bovine (B
= 0.54) and human (B
= 0.39) hosts were preferred by An. minimus s.l. in Kumari, Bangladesh.
Anopheles maculatus has been reported as an important malaria vector in the Malay Peninsula but is considered less important in Bangladesh , Assam, Borneo [37, 41], and the Philippines . Wharton  reported that the Malayan An. maculatus feeds uniformly on animals, although it exhibits a slight preference for humans even when animals are accessible. We found that both humans (B
= 0.23) and bovines (B
= 0.34) were the preferred hosts of this species. Similar results were reported in Hong Kong , whereas An. maculatus appears to feed largely on bovine hosts in the Philippines . Thus, An. maculatus feeds on either bovine or human hosts, whichever is more abundant or accessible.
A higher value of the HBI in an Anopheles species indicates that it can function as a malaria vector . In our study, higher values of HBI were found in An. baimai (80%), An. minimus s.l. (43.64%), and An. annularis (37.50%). These three species have been recognized as malaria vectors in Bangladesh, and the former two species play a main role in transmitting malaria in hilly and forested areas [6, 35, 36]. Very low to negligible values of HBI were found in An. kochi, An. nigerrimus, An. aitkenii, An. karwari, the An. hyrcanus group, An. umbrosus, and An. barbirostris. Recently, Alam et al. reported a high malaria infection rate for An. karwari, An. barbirostris, An. nigerrimus, and An. subpictus in Bangladesh. In contrast, these species are generally zoophilic and prefer to feed on bovine blood . In the present study, we found low HBI in An. karwari, An. barbirostris, and An. nigerrimus. These species are considered to be non-vectors, with the exception of An. subpictus, on the Indian subcontinent , which is in agreement with our observed HBI values for these species. Some non-vector species may be overestimated as malaria vectors due to the methods used . For example, using the entire mosquito body in ELISA can lead to over evaluation of vectors because Plasmodium species in human blood within mosquitoes are more likely to be detected. It is essential to check for Plasmodium within the upper parts of the mosquito body (thorax and head), and an even more reliable method is dissection because blood is often dispersed throughout the mosquito body and not only in the abdomen.
The densities of potential hosts in the study area must be measured to obtain a better understanding of mosquito host preferences . The selection index (B
) enables the evaluation of mosquito host preferences with consideration of environmental conditions. A strongly anthropophilic mosquito species would only use humans as preferred hosts, whereas an opportunistic species would prefer more than one host. The value of B
for a particular mosquito species quantifies the intensity of preference, and the number of preferred hosts is a measure of the opportunistic behavior. Anopheles minimus s.l., An. annularis, An. maculatus s.l., and An. pallidus showed opportunistic blood-feeding behavior, indicating that they chose either human or bovine hosts depending on conditions. The other species exhibited clear preferences for bovine hosts. Therefore, the selection index (B
) results demonstrated that most of the mosquito species preferred bovines and goats as their hosts but not humans. An accurate host population estimation would be rather difficult for estimating the forage ratio. However, forage ratios are more powerful indicators for examining mosquito host blood-feeding preferences compared to other specialized indices . These ratios may be used to compare the feeding preferences of various mosquito species in different areas.
Anopheles annularis and An. vagus are zoophilic, exophilic and exophagic in nature; however, they have been considered malaria vectors in India . These two species were identified as vectors during epidemics in the floodplain areas of Bangladesh. They may have been implicated perhaps due to the low availability of nonhuman mammalian hosts [9, 10]. Using gel-diffusion methods, a blood-meal analysis of several mosquito species in Dhaka, including three anophelines, indicated that An. vagus and An. barbirostris were highly zoophilic , which is in accordance with our results.
We identified a considerable proportion (5.61%) of multiple blood meals, a majority of which had been taken from humans and goats. We hypothesize that this combination was particularly frequent because goat shelters are, in many cases, located close to human sleeping rooms. The highest proportion of mixed blood meals occurred in An. vagus, which was collected indoors using PS. Multiple blood-feeding probably occurred due to disturbances or climatic factors . Multiple blood-feeding is common during a single gonotrophic cycle among mosquitoes, and its epidemiological importance is controversial [47, 48]. The loss of sporozoites by biting nonhuman animals during mixed feeding could be important in malaria control.