OUP user menu

Evaluation of a Sampling Methodology for Rapid Assessment of Aedes aegypti Infestation Levels in Iquitos, Peru

A. C. Morrison, H. Astete, F. Chapilliquen, G. Ramirez-Prada, Gloria Diaz, A. Getis, K. Gray, T. W. Scott
DOI: http://dx.doi.org/10.1603/0022-2585-41.3.502 502-510 First published online: 1 May 2004


An epidemic of dengue during 2001 in Northwestern Peru reemphasized the need for efficient, accurate, and economical vector surveillance. Between November 1998 and January 1999, we carried out extensive entomological surveys in two neighborhoods of ≈600 contiguous houses located in the Amazonian city of Iquitos, providing a unique opportunity to evaluate the Aedes aegypti (L.) rapid assessment survey strategy. Based on Pan American Health Organization recommendations, this strategy is used by the Peruvian Ministry of Health (MOH). In our analysis all household locations, including closed and unoccupied houses, were georeferenced and displayed in a geographic information system, which facilitated simulations of MOH surveys based on hypothetical systematic sampling transects. Larval, pupal, and adult mosquito indices were calculated for each simulation (n = 10) and compared with the indices calculated from the complete data set (n = 4). The range of indices calculated from simulations was moderately high, but included actual indices. For example, simulation ranges for house indices (HI, percentage of infested houses from complete survey) were 38-56% (45%), 36–42% (38%), 21–34% (30%), and 13–33% (27%) in four surveys. HI, Breteau index, pupae per hectare, adult index, and adults per hectare were more robust entomological indicators (coefficient of variation [CV]/mean = 0.1–2.9) than the container index, pupae per person, pupae per house, adults per person, and adults per house (CV/mean >20). Our results demonstrate that the MOH's Ae. aegypti risk assessment program provides reasonable estimates of indices based on samples from every house. However, it is critical that future studies investigate the association of these indices with rates of virus transmission to determine whether sampling variability will negatively impact the application of indices in a public health context.

  • Aedes aegypti
  • dengue
  • surveillance
  • geographic information system

No vaccine or chemotherapy is currently available for the prevention or treatment of Dengue fever, a disease caused by a virus that is transmitted by the mosquito Aedes (Stegomyia) aegypti (L.); therefore, prevention and control of the disease are dependent on vector surveillance and control measures. Recent epidemics caused by the introduction of previously unreported dengue virus serotypes (Dengue-2 Asian, Dengue-3, and Dengue-4) in northwestern, southeastern, and northeastern parts of Peru have reemphasized the need for efficient, accurate, and economical entomological surveillance strategies in that country (O.G.E., unpublished data). The Pan American Health Organization (PAHO) recommends the use of systematic or randomized sampling designs using standard sample size calculations (PAHO 1994). In Peru, the Ministry of Health (MOH) has been evaluating a rapid assessment technique that is based on the area to be covered and personnel available, starting blocks are selected at random and a systematic sampling strategy is carried out along a transect running from northwest to southeast along city blocks (Ramirez-Prada et al. 2000). This methodology has numerous logistic advantages including better coverage of a locality than using simple random sampling techniques and more efficient supervision of collectors.

Traditional entomological surveillance techniques are based on a series of indices that were designed to detect the presence or absence of Ae. aegypti larvae (Connor and Monroe 1923, Breteau 1954, Tun-Lin et al. 1995, Focks and Chadee 1997). Those methods assume a strong positive correlation between the presence of larvae and adult females in a household-only adult females transmit virus to humans. There are, however, three important reasons to question the strength of the larvae–adult association. First, because larval mortality can be high, adults may not emerge from a container holding immature mosquitoes. Second, because adults can fly, they can move away and become spatially disassociated from their larval development sites. Third, independent of the surveillance technique (larvae, pupae, or adult collections), citywide surveys carried out using PAHO recommendations or the rapid assessment technique proposed by MOH assume that there is no spatial or geographic structure among infested houses. Alternative entomological surveillance methods such as larval productivity indices (Chan et al. 1971; Bang et al. 1981; Tun-Lin et al. 1995, 1996), and pupal surveys were developed to circumvent the first two concerns. The third requires empirical validation.

Until the development of geographic information system (GIS), examination of Ae. aegypti surveys data for spatial structure was impractical. Within a GIS environment, geographically referenced point data sets can be analyzed by considering the distance between each point and all other points so that any spatial structure of that data can be described and controlled (Gatrell et al. 1996, Getis 1999). Alternatively, GIS provides a tool for simulations of systematic geographically based sampling strategies if complete or "census" data sets are available.

The purpose of this study was to evaluate the PAHO rapid assessment sampling strategy in two neighborhoods in Iquitos, Peru. Dengue virus transmission has been endemic in Iquitos since 1990. Our surveys were exhaustive. Our final objective was to identify spatial structure in measures of entomological risk, evaluate different sampling designs, and evaluate the relative robustness of larval, pupal, and adult Ae. aegypti indices. Herein, we specifically evaluate the rapid assessment sampling strategy and discuss its implication for entomological surveillance and dengue risk assessment in Peru.

Materials and Methods

Study Area

The area chosen for this study consists of two neighborhoods in the Amazonian city of Iquitos, Peru (73.2° W, 3.7° S, altitude 120 m above sea level). Iquitos has been described in detail in previous studies in the city (Hayes et al. 1996, Watts et al. 1999).

The two neighborhoods where we carried out entomological surveys were Maynas (MY) and Tupac Amaru (TA) have been described in detail for a related study (Getis 2003; Fig. 1). The sites were chosen because of previously observed differences in Ae. aegypti indices and dengue seroprevalence rates (Scott and Morrison 2003); MY represents a neighborhood of higher risk of dengue transmission.

Fig. 1

Map of study locations in the neighborhoods of Tupac Amaru (bottom right)and Maynas (left)within the city of Iquitos, Peru (top right). Large maps show coverage during entomological surveys carried out during first sample period from November to December 1998.

Study Design

Entomological surveys were carried out exactly as described by Getis (1999). Briefly, 600 contiguous houses were surveyed on 20 city blocks in MY and on 14 city blocks in TA twice in consecutive surveys first from mid-November to mid-December 1998 and second from mid-December 1998 to mid-January 1999. Five two-person entomology collection teams were provided a map of a block to be surveyed with a designated start house. The interior and exterior of each household were surveyed in sequence daily along the block from the start house between 0700 and 1300. On alternating days each neighborhood (MY and TA) was surveyed. Unoccupied or closed houses and houses where residents did not provide permission for the survey, businesses, offices, and schools were not sampled. Collecting teams were rotated among blocks each day in an effort to limit temporal and collector biases. Each day, an attempt was made to inspect houses that were closed or refused access. This was done before continuing surveys of unsampled households. Access to closed houses was sought a minimum of three times.

Entomological Surveys

Our survey methodology was based on techniques suggested by Focks et al. (1993) so that traditional larval indices (PAHO 1994), and pupal indices (Focks and Chadee 1997) could be obtained for each of the four surveys. In addition, adult mosquito collections were made using a backpack aspirator (John W. Hock Company, Gainesville, FL). Briefly, after asking permission to survey the household, one member of the team administered a small demographic survey designed to determine the number of occupants, dimensions of the property, house construction materials, method of cooking, water use patterns, type of sewage disposal, and insecticide use. Simultaneously, the other team member began collecting adult mosquitoes by using a backpack aspirator (John W. Hock Company). Aspiration collections were attempted in all rooms of the house—when permitted—passing the aspirator over walls, under furniture, and inside closets and other likely adult mosquito resting sites (Scott et al. 2000). Aspiration collections were similarly attempted outside the house from outside walls, under eaves, vegetation, and outdoor stored materials.

After recording demographic information, including the number of occupants, the other team member examined all potential Ae. aegypti development sites for water, larvae, and pupae both inside and outside the house. At the initiation of the study 15 container categories (rain gutters, tree holes, coconut shells, bottles, cans, drums, metal pots, plastic containers, ceramic jars, tires, flower vases, elevated tanks, low tanks, wells, and miscellaneous containers) were defined. During the course of the study, new categories were recognized (toilet tank, toilet connector depression, pet dish, plastic bags, puddle, sewer, holes, sink, pool, lid, and plastic tubing). All pupae and a sample of larvae found were placed in a twist-top plastic bag and labeled with a house and container code. Larvae, pupae, and adults were transported the same day to the field laboratory in Iquitos for processing. Containers were not actively destroyed nor treated by surveyors and larvae and pupae collected were not returned to containers. Total collection time for a house (larvae, pupae, and adult collections) varied with the size and complexity of the property (average 7 min inside and 5 min outside; range 2–45 total minutes).

In our field laboratory, larvae were identified as Ae. aegypti by the relative size of the sifon and their movement compared with the other most commonly found Culex species (Consoli and de Oliveira 1994). Limatus spp. larvae were differentiated by the characteristics on the 8th tergite (Consoli and de Oliveira 1994). All larval samples were cross-checked with the entomology collection sheets provided by the field team. Pupae were counted and placed in plastic emergence vials (no. 9 dram with no. 3 one-half dram vial as lid, Thorton Plastics, Salt Lake City, UT), ≤30 per vial, and labeled with the house and container code. Each subsequent day, emerged adults were aspirated with a mechanical aspirator, transferred to small plastic cups (6 cm in diameter by 8 cm in length) and placed in a −20°C freezer. After 30 min to 1 h, adults were removed from the freezer and identified to species, counted by sex, and data were recorded on the entomology collection sheet. Adults captured by aspiration were placed in a −20°C freezer for a minimum of 30 min, identified to species and number was recorded for the following groups: Ae. aegypti, Culex spp. (all non-melanoconium group); Mansonia spp., Cx. melaniconium, and other species.

Computer Simulations and Analysis of Data

We developed a GIS, by using ARC/INFO and ARC/VIEW software, for the city of Iquitos (Getis 2003). All entomological and demographic data collected from the four surveys were related to geographic coordinates via a unique house code. Households were then selected manually within Arc View to form the simulation data set described below.

The following entomological indices were calculated for each for the four surveys: house index (HI, number of houses with one or more containers positive for immature Ae. aegypti divided by the number of houses sampled multiplied by 100); container index (CI, number of water-holding containers with immature Ae. aegypti divided by the number of water-holding containers inspected multiplied by 100); Breteau index (BI, number of water-holding containers with immature Ae. aegypti per every 100 households inspected); pupae per hectare (pu/ha, number of pupae per hectare inspected); pupae per person (pu/per, number of pupae collected over the total number of inhabitants of the households inspected); pupae per house (pu/hse, number of pupae per household inspected); adult index (AI, number of houses positive for adult Ae. aegypti divided by the number of houses inspected multiplied by 100); adults per hectare (ad/ha, number of adult Ae. aegypti mosquitoes collected by backpack aspiration per hectare inspected); adults per person (ad/per, number of adult Ae. aegypti mosquitoes collected by backpack aspiration over the total number of human habitants from the households inspected); and adults per house (ad/hse, number of adult Ae. aegypti mosquitoes collected by backpack aspiration per household inspected). The total number of hectares sampled was determined by summing the area of the lot sizes of the households surveyed and the total number of persons by summing the number of occupants in each surveyed household from the demographic survey.

Demographic information (lot sizes and human occupants) and entomological data for each survey were linked to a unique household code within our GIS. Entomological indices were calculated for each complete survey, which we refer to as the "census" data sets, for comparison with output from computer simulations. Households that were not surveyed during census surveys, because access was denied, they were closed, schools, businesses, churches, unoccupied, or vacant lots, were identified in the GIS so that those houses would not be selected in simulations of MOH surveys.

We simulated 10 MOH surveys for each of the four entomological surveys actually carried out by selecting houses in Arc View by using the following strategy. Each subsample was based on systematic sampling transects beginning on three randomly selected blocks that three collectors (15–25 houses each) carried out to obtain the PAHO recommended sample size of ≈60 houses (10% sample). Figure 2 shows the houses (black dots) selected for nine simulated MOH surveys in MY. The shaded blocks were selected using a random number table. Starting on the northwest corner of the three randomly selected blocks, simulated surveyors would sample every third house, when possible along the west and south sides of the block. After reaching the southeast corner of that block, the surveyor would continue the route on the block immediately southeast of his or her position. If the eastern limit of the study area was reached, the direction of the transect would change to a northeast corner to southwest corner direction; alternatively, if the southern limit of the study area was reached the transect would then follow a southeast to northwest direction. Each surveyor would continue sampling until he or she had sampled 15–25 houses and had finished a block along the respective transect routes. Simulated transect routes considered house status (e.g., closed, business, not occupied, denied access), selecting the house directly before or after the house designated by systematic sampling of every third house. Entomological indices were calculated for each of the 40 simulated data sets. The mean, standard deviation, and coefficient of variation (CV) were calculated using the 10 simulations for each simulated survey at each site: MY and TA. Indices with low coefficients of variation compared with the mean were considered to be reliable and less affected by sampling variation (Tun-Lin et al. 1996).

Fig. 2

Example of nine rapid assessment transects sampled in simulations of the first Maynas survey. Shaded blocks were selected and random and black dots illustrate sampled houses.


Entomological surveys were carried out in 92–96% of the houses within the study areas in TA and MY during the two survey periods (Table 1). Coverage was highest during the first survey period in both TA (96%[562/585] versus 93%[545/585]) and MY (94%[582/617]) versus 92%[567/617]). Approximately 10% of lots in each study area were not sampled because they contained an unoccupied house, school, business, church, or were vacant. Neither the coverage rate of households that could be surveyed (surveyed houses plus closed or houses where access was denied) nor the proportion of lots not sampled was significantly different among the study areas or survey dates.

View this table:
Table 1

Entomological indices calculated from census data sets were significantly higher in the MY study area (Tables 2 and 3). For example, during the first survey period, the house index was 28.8% (142/562) in TA and 44.7% (260/582) in MY (χ2 = 30.8, df = 1, P < 0.0001). Similarly, during the second survey period, the house indices were 22.7% (124/545) and 38.1% (216/567) in TA and MY, respectively (χ2 = 30.8, df = 1, P < 0.0001). This trend was consistent for all indices (Tables 2 and 3). Larval indices (HI, CI, and BI) were 1.5–1.9× higher in MY than TA. Pupal indices showed the greatest difference in magnitude ranging from 2.2 to 2.8× higher in the MY than TA study area. Adult indices reflected an intermediate magnitude of difference between the two study area; 1.5–2.2 times higher in MY.

View this table:
Table 2
View this table:
Table 3

In both sites, indices decreased between the first and second surveys. For example, the house index decreased from 28.8 to 22.7% in TA (χ2 = 0.95, df = 1, P = 0.3276) and from 44.7 to 38.1% in MY (χ2 = 5.1, df = 1, P = 0.0236); the decrease was statistically significant only in MY. This trend was consistent among larval and pupal indices (Tables 2 and 3), but the magnitude of the difference (1.1–1.4 times higher in first survey) was less than the differences between the two study areas. Adult indices, however, showed a different pattern between the two survey periods. In MY, all estimates of adult indices were slightly lower between the first and second surveys (Table 3). The pattern in TA was similar in that estimates showed little change between the two periods, but the adult index increased between the first and second surveys (Table 2). These differences were not statistically significant.

The types of containers observed was remarkably similar across sites and surveys, with plastic containers (including buckets, bowls, basins, wash tubs, and storage containers) and metal pots comprising between 74 and 78% and 7 and 12% of wet containers, respectively. Drums (55-gallon; 3.7%), miscellaneous (2.7%), low tanks (2.1%), and tires (1.6%) were the next most abundant wet container types. Containers that were more difficult to sample such as wells, high tanks, pools, and sewers were relatively rare accounted for <0.4% of all wet containers combined. Rain gutters were examined but not found to contain standing water. Overall, plastic containers accounted for 33% (range, 29–36%), drums for 14% (range, 12–16%), tires and miscellaneous containers for 10% (range, 7–16%) each, and low tanks for 9% (range, 3–12%) of all Ae. Aegypti-positive containers. Low tanks, however, were the most productive container type for pupae, accounting for 33% of all pupae collected, followed by plastic containers (21%), drums (14%), miscellaneous (12%), tires (9%), metal pots (4%), and cans (3%). We observed Ae. aegypti pupae in four of 33 wells, sampling with buckets until no more could be removed.

Computer simulations based on sample size of ≈60 houses (10%) by using the rapid assessment technique revealed a range of index estimates for each of the 10 simulations for each of the four surveys. The mean, standard deviation, range of observed indices, CV, and relationship of coefficient of variation to the mean are summarized for TA in Table 2 and MY in Table 3. The mean of the 10 simulated surveys compared with the index calculated for the census data sets were similar for all indices in all four surveys, but the range of estimates among the individual simulations was wide. For example, mean HIs in the first MY survey estimated by census and simulation were 45 and 48, respectively. The range of HI values was 38–56.

The coefficient of variation's were lowest for the larval indices ranging from 5 to 12 in MY and 20–42 in TA. Coefficient of variation's for adult indices were intermediate ranging from 16 to 40 and from 17 to 75 in MY and TA, respectively. Pupal indices showed the most sample variation with coefficient of variation between 45 and 67 in MY and 39 and 48 in TA. The coefficient of variation's were higher in the TA surveys where infestation rates were significantly lower than in MY.

Index stability was evaluated by comparing the coefficient of variation to the mean of the simulation results. HI, Breteau index, pupae per hectare, adult index, and adults per hectare were more robust entomological indicators (CV/mean = 0.1–2.9) than the container index, pupae per person, pupae per house, adults per person, and adults per house (CV/mean >20).


Ae. aegypti surveillance relies heavily on larval surveys, in large part because surveillance and control grew out of an eradication paradigm that promoted complete, thorough, and repeated coverage of infested areas (Reiter and Gubler 1997). In 1994, however, the Pan American Health Organization declared eradication of this species an unattainable goal and promoted Ae. aegypti control, which they defined as the "cost-effective utilization of limited resources to reduce vector populations to levels at which they are no longer of significant public health importance" (PAHO 1994). Unfortunately, scant empirical data exist defining entomological thresholds for dengue transmission. Experience with yellow fever and recent computer simulation estimates indicate that entomological thresholds for dengue are low (Focks et al. 1993, 2000; Focks and Chadee 1997; Reiter and Gubler 1997). Empirically defined thresholds will require prospective, longitudinal studies in which investigators simultaneous monitor the relationship between dengue virus transmission in a human cohort and Ae. aegypti population densities. Meaningful analysis of both human cohort data and Ae. aegypti surveillance data collected within routine public health programs must consider the following three questions (Scott and Morrison 2003). First, can individual households be considered independent sampling units or is there some type of local spatial structure of the mosquito population? That is, are neighboring houses more likely to be infested than houses further apart or from different neighborhoods. Second, what is the most informative and reliable measure of entomological risk? Third, what is the most appropriate survey design for sampling houses within a city? Our study contributed to an improved understanding for each of these issues and directly evaluated a sampling design that has significant logistic advantages over strategies commonly used (PAHO 1994). Our GIS database made it possible for us to simulate and test a transect-based systematic sampling design.

A potential drawback of the rapid assessment approach is that random selection of starting blocks can result in overlapping routes among collecting groups. To avoid this pitfall, one of our coinvestigators (G.D.), who had carried out rapid assessment surveys previously, was consulted about what route alterations would be made under field conditions. In addition, we were able to account for the location of houses that were unoccupied, closed, denied access, business, schools, churches, and vacant lots. In our simulations, we walked the block just as a Ministry of Heath worker would, surveying the house before or after. Although, there was variation in index estimates among the 10 simulations for each survey, the mean values were very close to those calculated with the census data sets, and the coefficients of variations were relatively low. Overall, results from our simulations behaved as would be expected under an assumption of independence, as such standard sample size calculations were appropriate. Furthermore, additional simulations (data not shown), in which entire blocks (every third house) were surveyed rather than transects and the entire area was surveyed (every 10th house), generated index mean estimates and coefficient of variation's nearly identical to those generated using the rapid assessment transect. Thus, no geographic biases were observed using the rapid assessment transect technique. We conclude that households infested with Ae. aegypti are randomly distributed within the sampled neighborhoods. This deduction is supported by spatial statistical analysis (K-function and G*) by using the same data set and in which we found no clustering of larvae or pupae beyond individual households and only limited clustering of adult Ae. aegypti to a distance of ≈30 m (Getis 2003). Furthermore, numerous researchers have reported spatial and temporal clusters of clinically ill dengue patients in the same household or adjacent houses (Halstead et al. 1969, Chan 1985, Waterman et al. 1985, Gubler 1992). Using a georeferenced data set of dengue cases during a 1991 epidemic in the town of Florida, Puerto Rico, dengue cases clustered significantly within households but not beyond (Morrison et al. 1998). Our results demonstrate that individual households are the appropriate spatial unit for entomological surveys and the rapid assessment technique is an efficient and appropriate study design. The transect approach provides wide coverage of the study area, as long as routes are adjusted to avoid duplication and facilitate supervision of entomological collectors.

Ae. aegypti surveys with complete coverage of a study area are rare, often are considered operationally unnecessary, and are almost always beyond the financial resources of local public health budgets (PAHO 1994). Thus, our data set provides a unique opportunity to evaluate standard as well as novel measures of entomological risk. Indices based on adult female mosquito abundance should be the most appropriate measure of entomological risk because female mosquitoes are the only life stage that transmit virus. In at least one previous study, adult Ae. aegypti abundance was correlated with diagnosed dengue cases (Rodriguez-Figueroa et al. 1995). Sampling methods for adults, however, are more biased than for other life stages (Reiter and Nathan 2001) and the value of larval indices has been challenged because their relationship with adult densities is questionable (Reiter and Gubler 1997, Focks et al. 2000). Pupal indices are being considered as alternatives to traditional larval indices (Focks et al. 1993, 2000; Focks and Chadee 1997). Pupal indices are attractive for three reasons. First, it is theoretically possible to make absolute counts of their abundance, something that cannot be done for flying and difficult-to-capture adults. Second, pupal mortality is low. The magnitude of the pupal population should, therefore, be directly and relatively easily correlated with adult densities. Third, because the pupa is the life stage that directly precedes the virus-transmitting adult, pupae should be a more direct measure of transmission risk than larvae, which are a developmental step removed from adults.

Coefficients of variation from our study were lowest for larval indices and were comparable with those reported for a similar simulation experiment carried out in Australia (Tun-Lin et al. 1996). In the Australian simulation, data pooled from four small towns (758 houses) was sampled at random 30 times for 50 houses. That analysis represented a 7% rather than the 10% sample in our surveys. Australian house and Breteau indices were similar to that found in our Tupac Amaru surveys

Quotients of coefficients of variation in relation to means suggest that the HI and BI are more robust than CI and generally less sensitive to sample variation than adult or pupal indices. As pointed out by Tun-Lin et al. (1996), however, the robustness of larval indices may be a reflection of their relative lack of variation. Pupal indices had the widest range of values and had the highest coefficients of variation. Significantly, the coefficient of variation/mean depended on the denominator of the index calculated. Pupae per hectare was consistently the most robust index (CV/mean 0.08–0.53), whereas pupae per person and house were the least robust. Patterns associated with adult indices were similar to those for pupae; the adult index and adults per hectare were highly stable and adults per person and adults per household were highly unstable. In fact, using persons in the denominator resulted in coefficient of variation/mean values ranging from 164 to 1,052. The number of household occupants was obtained at the time of the survey and may be subject to reporting bias.

Logistical considerations aside, based on the results from our simulations we recommend pupal or adult counts per hectare as the best estimate of immediate risk for dengue virus transmission. Those indices would be most effected by changes in "key producing" containers that contribute a disproportionate amount of the pupae or adults captured during a survey (Tun-Lin et al. 1995). It is encouraging however, that the HI and BI performed well and did reflect historical differences in dengue transmission rates between the two neighborhoods. We found the collection of pupae and backpack aspirator collections to be practical in Iquitos in a research context. For routine Ae. aegypti surveillance activities, the feasibility of pupal and adult collections remains untested. Pupal collections were facilitated in Iquitos, where removing pupae, although time-consuming, was feasible because containers were usually accessible and of manageable volume. In addition, we did not observe wet rain-gutters (they tend to be flushed out daily) or subterranean containers where removal of pupae would be limited by access. There remains however, a need for alternative methods to collect adult mosquitoes in a standardized manner; for example, sticky ovitraps are showing promise (Bangs et al. 2001, Ritchie et al. 2003). The relative merits of the different indices in predicting pupal production patterns in different container types and dengue virus transmission will be the subject of future publications.

As we expected, indices were more stable when infestation rates increased; for example, in MY over TA. As infestation rates decreased, however, their predictive value similarly decreased. The latter is important because it indicates that entomological surveillance will be decreasingly cost-effective as mosquito control becomes increasingly effective and mosquito densities are reduced. An epidemiologically important consequence of this effect could be unstable dengue transmission and an unanticipated increase in disease. Long-term mosquito population reduction, without eradication, will lower human herd immunity, lower minimum entomological thresholds, and increase the potential for explosive epidemic virus transmission. Before conclusive evaluation of entomological surveillance techniques can be carried out the empirical relationship between entomological indices and virus transmission must be determined directly. Clear public health goals need to be defined. For example, if the goal is to prevent introduction of a new virus serotype or prevent transmission, entomological thresholds may be so low that entomological surveillance and Ae. aegypti eradication become nearly synonymous. If, however, entomological surveillance is used for assessment of relative risk between cities or regions of cities, density estimates can be used to assess risk and predict the relative merits of different intervention strategies. For purposes of investigating the dynamics of dengue transmission, our results point out the need to assess risk of human infection at the household level at frequent time intervals. In that context, the rapid assessment sampling design is an appropriate approach and is certainly justified by its logistical advantages.


We thank the residents of Tupac Amaru and Maynas in Iquitos, Peru, for allowing us to survey homes. We are grateful to the following individuals for carrying out the entomological surveys: Jimmy Maikohol Castillo Pizango, Abner Enrique Varsallo Lachi, Angel Puertas Lozano, Victor Elespuru Hidalgo, Manuel Golbert Ruiz Rioja, Rusbel Inapi Tamani, Guillermo Inapi Huaman, Fernando Chota Ruiz, Nestor Jose Nonato Lancha, Juan Luis Sifuentes Rios, and Edson Pilco Mermao. We thank Karla Block and Rosa Burga for logistical support in Iquitos and Doug Watts, Kevin Russell, Ellen Andersen, and Michael Zyzak for the support from NMRCD in Lima. We thank Carlos Calampa and Ruben Naupay from the Direcion Regional de Salud de Loreto for support of this project. This study received financial support from National Institutes of Health grant (AI)-42332.


  • An earlier version of this manuscript occurred in the Web-based Revista Peruana de Epidemiologia, a nonpeer-reviewed organ of the Peruvian Department of Health. In agreement with the editors of the J Med Entomol and the Rev Peruana Epid and with the consent of all authors, this manuscript is published in J Med Entomol to ensure wider circulation.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial reuse, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions{at}oup.com

References Cited

View Abstract