**5. Directly interpretable entomological metrics of mosquito behaviours, insecticide susceptibility, intervention effectiveness, and intervention impact**

can be modified according to need by countries that are particularly large, small or diverse. In addition to routine surveys of established surveillance sites, *ad hoc* spot checks and focus investigations are also recommended as ways to further improve vector surveillance and control [53]. Additionally, much more intensive, finer-scale surveillance of vector population dynamics is required wherever pro-active mosquito abatement methods, specifically larvicide application or space spraying, are deployed. These vertically-managed methods for delivering insecticides across large areas need to be repeated on a regular basis, often as frequently as every week. In order for mosquito population density measurements to be useful for monitoring purposes, they need be collected at spatial scales fine enough to identify

**Figure 4.** A suggested flow diagram for cyclical collection and assessment of programmatic surveillance data to inform malaria vector control on national scales. Formal annual review cycles may be supplemented by *ad hoc* review meetings

at short notice, whenever surprizing or alarming trajectories are observed through ongoing data monitoring.

410 Towards Malaria Elimination - A Leap Forward

Entomological monitoring to inform routine programme implementation needs to yield measurements that can be directly and informatively interpreted, so that those collecting the data in the field can readily use and quality control it. As discussed in the section that follows, simple summary metrics of mosquito behaviour and insecticide susceptibility do have limitations that need to be addressed with simulation models, but nevertheless need to have decision-making value in their own right.

Appropriate graphical tools are particularly important for helping programme staff to accurately interpret data. For example, many entomologists directly interpret the results of indoor and outdoor human landing catches without weighting these biting rate measurements in proportion to estimates of where people spend various times of the night (**Figure 5A** and **B**). This common misinterpretation is even endorsed by the latest WHO guidelines [53], which recommend numerical expression in the form of an *endophagy index*, comprising the mean indoor biting rate divided by the sum of the mean indoor and outdoor biting rates. This approach usually grossly overestimates the outdoor fraction of transmission exposure because participants in human landing catches behave in a deliberately misrepresentative manner, spending an average of half their time indoors and half outdoors across all times of the night. In the vast

Entomological Surveillance as a Cornerstone of Malaria Elimination: A Critical Appraisal

http://dx.doi.org/10.5772/intechopen.78007

413

**Figure 5.** Two examples of how raw mosquito and human behaviour data must be combined with simple analytical models to allow visualization and quantification of where and when human exposure actually occurs as a behavioural

interaction [6, 47, 61, 62].

Despite their limitations, existing simple insecticide bioassays provide an excellent example. Some insecticides can induce delayed but nevertheless invaluable mortality among mosquitoes that are classified as highly resistant based on the 24-hour holding period traditionally used in standard susceptibility assays [59]. Nevertheless, complete and rapid mortality within a day of exposure can only be a good thing and favours the selection of an insecticide verified to do so. Once interventions like LLINs or IRS have been deployed, it is always encouraging if they can be verified to exhibit durable insecticidal efficacy in the field, using well-established cone or wire ball assays with fully-susceptible insectary-reared mosquitoes.

For measuring impact, reduced biting densities and sporozoite prevalence rates can be directly interpreted as indicators of intervention success. Also, vector population rebounds can be identified by directly examining simple graphs of longitudinal trends in density and infection prevalence (**Figure 2**). Any such suspected intervention failure should trigger careful examination of all the above vector behaviour and insecticide susceptibility metrics, as well as indicators of effective vector control products and delivery processes in the field.

On the behavioural front, high estimates for the proportion of human exposure to mosquito bites occurring indoors is always an encouraging indicator that LLINs should at least provide strong personal protection [6]. They may also achieve vector population control if they are also susceptible to the insecticidal active ingredients and obtain a large proportion of blood meals from humans [1, 6–8]. Once high LLIN use has been achieved, high proportions of residual transmission may occur outdoors, and the vector may become more reliant upon livestock as a source of blood, indicating that spatial insecticide emanators or veterinary endectocides may be considered as possible supplementary interventions [7]. While surveys of bloodmeal sources among samples of engorged mosquitoes are always biased to overrepresent the indoor-resting and human-feeding fraction of the vector population [36, 37], very high estimates of the human blood index are nevertheless a strong indicator of both vectorial capacity and vulnerability to attack with human-centred approaches [1, 6–8, 38]. Conversely, where large proportions of blood meals are found to originate from livestock, this is an encouraging indicator that veterinary formulations of endectocides could be useful as a supplementary vector control tool [6–8].

Perhaps the most important reason for entomological surveillance data to be readily and directly interpretable is so that data interrogation begins with the front-line staff who collect it in the field. The closer to the point of collection that data is examined and interpreted, the sooner it is acted upon and the sooner it is queried for completion and correctness. Even within our specialized research groups, we have recently achieved huge improvements in entomological data quality simply by having it entered by the people who collected it on the day it was recorded. Entomological surveillance indicators that can be directly and intuitively interpreted in the field are much easier to quality control and quality assure, especially through decentralized data collection platforms.

Appropriate graphical tools are particularly important for helping programme staff to accurately interpret data. For example, many entomologists directly interpret the results of indoor and outdoor human landing catches without weighting these biting rate measurements in proportion to estimates of where people spend various times of the night (**Figure 5A** and **B**). This common misinterpretation is even endorsed by the latest WHO guidelines [53], which recommend numerical expression in the form of an *endophagy index*, comprising the mean indoor biting rate divided by the sum of the mean indoor and outdoor biting rates. This approach usually grossly overestimates the outdoor fraction of transmission exposure because participants in human landing catches behave in a deliberately misrepresentative manner, spending an average of half their time indoors and half outdoors across all times of the night. In the vast

data in the field can readily use and quality control it. As discussed in the section that follows, simple summary metrics of mosquito behaviour and insecticide susceptibility do have limitations that need to be addressed with simulation models, but nevertheless need to have

Despite their limitations, existing simple insecticide bioassays provide an excellent example. Some insecticides can induce delayed but nevertheless invaluable mortality among mosquitoes that are classified as highly resistant based on the 24-hour holding period traditionally used in standard susceptibility assays [59]. Nevertheless, complete and rapid mortality within a day of exposure can only be a good thing and favours the selection of an insecticide verified to do so. Once interventions like LLINs or IRS have been deployed, it is always encouraging if they can be verified to exhibit durable insecticidal efficacy in the field, using well-established

For measuring impact, reduced biting densities and sporozoite prevalence rates can be directly interpreted as indicators of intervention success. Also, vector population rebounds can be identified by directly examining simple graphs of longitudinal trends in density and infection prevalence (**Figure 2**). Any such suspected intervention failure should trigger careful examination of all the above vector behaviour and insecticide susceptibility metrics, as well as indicators of effective vector control products and delivery processes in the field.

On the behavioural front, high estimates for the proportion of human exposure to mosquito bites occurring indoors is always an encouraging indicator that LLINs should at least provide strong personal protection [6]. They may also achieve vector population control if they are also susceptible to the insecticidal active ingredients and obtain a large proportion of blood meals from humans [1, 6–8]. Once high LLIN use has been achieved, high proportions of residual transmission may occur outdoors, and the vector may become more reliant upon livestock as a source of blood, indicating that spatial insecticide emanators or veterinary endectocides may be considered as possible supplementary interventions [7]. While surveys of bloodmeal sources among samples of engorged mosquitoes are always biased to overrepresent the indoor-resting and human-feeding fraction of the vector population [36, 37], very high estimates of the human blood index are nevertheless a strong indicator of both vectorial capacity and vulnerability to attack with human-centred approaches [1, 6–8, 38]. Conversely, where large proportions of blood meals are found to originate from livestock, this is an encouraging indicator that veterinary formulations of endectocides could be useful as a

Perhaps the most important reason for entomological surveillance data to be readily and directly interpretable is so that data interrogation begins with the front-line staff who collect it in the field. The closer to the point of collection that data is examined and interpreted, the sooner it is acted upon and the sooner it is queried for completion and correctness. Even within our specialized research groups, we have recently achieved huge improvements in entomological data quality simply by having it entered by the people who collected it on the day it was recorded. Entomological surveillance indicators that can be directly and intuitively interpreted in the field are much easier to quality control and quality assure, especially

cone or wire ball assays with fully-susceptible insectary-reared mosquitoes.

decision-making value in their own right.

412 Towards Malaria Elimination - A Leap Forward

supplementary vector control tool [6–8].

through decentralized data collection platforms.

**Figure 5.** Two examples of how raw mosquito and human behaviour data must be combined with simple analytical models to allow visualization and quantification of where and when human exposure actually occurs as a behavioural interaction [6, 47, 61, 62].

majority of human populations, most people sleep indoors at night for security reasons, so very little of the biting activity measured outdoors is relevant to normal human exposure patterns [60]. Taking the major African malaria vector *An. gambiae* as an example, the traditional narrative describing it as *endophilic* is inaccurate, because their biting rates indoors and outdoors are usually similar and they have no strong or consistent preference for attacking people in either location [50]. It is the timing of biting activity that caused most historical exposure to occur indoors. This human-specialized mosquito species usually exhibits biting activity peaks that occur in the middle of the night when most people are asleep indoors (**Figure 5A**), and therefore vulnerable to attack unless protected with an LLIN. It is therefore more accurate to say that these vectors are highly *nocturnal*, feeding mostly at times when humans exhibit strong *endophilic* tendencies [35], and that is why most human exposure occurred indoors at night before the scale up of LLIN use (**Figure 5C**).

Presenting measured behavioural interactions between mosquitoes and humans in such intuitive graphical formats is important for enabling accurate interpretation, and can be facilitated with user-friendly spreadsheet templates or automated visualization options in surveillance

Entomological Surveillance as a Cornerstone of Malaria Elimination: A Critical Appraisal

http://dx.doi.org/10.5772/intechopen.78007

415

The use of models to look at entire life histories of mosquitoes was central to the very earliest work of Ross [64], and to the ethos of *epidemiological entomology* defined by Garrett-Jones half a century ago [65]. While direct interpretation of simple indicators should provide the essential core of evidence used to inform programmatic decisions, astute application of analytical models to examine the life histories of mosquito populations can also yield important insights

For example, the slow-acting toxicity of pyrethroids to mosquitoes that are clearly resistant to this insecticide class was only recently identified as being central to the sustained impacts of LLINs [59]. While most African *Anopheles* populations are now sufficiently resistant against pyrethroids to survive immediately after exposure, they do suffer increased mortality over the longer term, essentially all of which occurs within the 10 days required for the parasite to complete sporogonic development [59]. As a result, while pyrethroid resistance clearly does compromise the impacts of LLINs [66], it falls far short of abrogating them entirely, so they

Also, fitting process-explicit models of mosquito population dynamics to vector density trends may yield some insights that cannot be obtained by direct interpretation. Such mechanistic modelling analyses have been successfully applied to field data to identify negative density-dependence of mosquito reproduction, which make vector populations more robust to control than would otherwise be expected [67]. Similar models have been fitted to the population trajectories of self-propagating populations in large cages, which were experimentally exposed to different vector control measures and combinations thereof. These biologically-informative analyses quantified impacts on specific target parameters like survival and fecundity, helped confirm that near-extinction of these small populations was achieved, and revealed a surprising mode of action for one of these emerging technologies (Ng'habi et al., Unpublished). Such approaches could be readily extended to data from routine population dynamics monitoring, allowing the complementarities, synergies and redundancies achieved by combinations of vector control measures to be understood at an unprecedented

Relatively simple deterministic models have also been used to illustrate how insecticide resistance traits and intervention avoidance behaviours can interact synergistically, allowing resilient mosquito populations to persist despite widespread LLIN use without necessitating any major adaptations of their preferred feeding times [68]. By foraging cautiously and repeatedly inside houses, to maximize their feeding opportunities while minimizing

**6. Life history analyses to identify otherwise non-obvious vector** 

data dashboards.

level of detail.

**control challenges and opportunities**

that would not otherwise be obvious.

remain an invaluable tool for malaria vector control [59].

Contrasting with vector populations like those of *Anopheles farauti* in the Solomon Islands, where humans are mostly exposed to outdoors, the most important feature of this behavioural interaction is again the timing of host-seeking activity. By feeding predominantly in the evening, this species can readily attack humans indoors or outdoors while they are still awake and active so bed net use is impractical (**Figure 5B** and **D**). Again, biting densities are similar indoors and outdoors (**Figure 5B**), so it is inaccurate to describe this vector as *exophagic* in the strict sense, and much more important to emphasize that so much exposure occurs outdoors because it is *crepuscular*, with feeding activity that peaks at dusk when people are awake, active and cannot use bed nets.

The overall exposure distribution estimates represented by the areas under the curves in **Figure 5C** and **D**, can then be combined with direct field estimates for the proportion of bloodmeals obtained from humans to visualize the maximum limit of *biological coverage* [63] achievable with human-targeted measures like LLINs as simple box graphs (**Figure 6**).

**Figure 6.** Box diagrams illustrating how the two different vectors described in **Figure 5** differ in terms of their overall behavioural vulnerability to population suppression with long-lasting insecticidal nets, expressed as the maximum achievable *biological coverage* of blood resources used by the mosquitoes [6].

Presenting measured behavioural interactions between mosquitoes and humans in such intuitive graphical formats is important for enabling accurate interpretation, and can be facilitated with user-friendly spreadsheet templates or automated visualization options in surveillance data dashboards.
