• 1

    Minakawa N, Mutero CM, Githure JI, Beier JC, Yan G, 1999. Spatial distribution and habitat characterization of anopheline mosquito larvae in western Kenya. Am J Trop Med Hyg 61 :1010–1016.

    • Search Google Scholar
    • Export Citation
  • 2

    Gimnig JE, Ombok M, Kamau L, Hawley WA, 2001. Characteristics of larval anopheline (Diptera: Culicidae) habitats in western Kenya. J Med Entomol 38 :282–288.

    • Search Google Scholar
    • Export Citation
  • 3

    Robert V, Awono-Ambene HP, Thioulouse J, 1998. Ecology of larval mosquito, with special reference to Anopheles arabiensis (Diptera: Culicidae) in market-garden wells in the urban area of Dakar, Senegal. J Med Entomol 35 :948–955.

    • Search Google Scholar
    • Export Citation
  • 4

    Trape JF, Zoulani A, 1987. Malaria and urbanization in central Africa: the example of Brazzaville. Part II: Results of entomological surveys and epidemiological analysis. Trans R Soc Trop Med Hyg 81 :10–18.

    • Search Google Scholar
    • Export Citation
  • 5

    Khaemba BM, Mutani A, Bett MK, 1994. Studies on the anopheline mosquitoes transmitting malaria in a newly developed highland urban area: a case study of Moi University and its environs. East Afr Med J 71 :159–164.

    • Search Google Scholar
    • Export Citation
  • 6

    Chinery WA, 1984. Effects of ecological changes on the malaria vectors Anopheles funestus and the Anopheles gambiae complex of mosquitoes in Accra, Ghana. J Trop Med Hyg 87 :75–81.

    • Search Google Scholar
    • Export Citation
  • 7

    Barbazan P, Baldet T, Darriet F, Escaffre H, Djoda DH, Hougard JM, 1998. Impact of treatments with Bacillus sphaericus on Anopheles populations and the transmission of malaria in Maroua, a large city in a savannah region of Cameroon. J Am Mosq Control Assoc 14 :33–39.

    • Search Google Scholar
    • Export Citation
  • 8

    Coene J, 1993. Malaria in urban and rural Kinshasa: the entomological input. Med Vet Entomol 7 :127–137.

  • 9

    Gillies MT, Coetzee M, 1987. A Supplement to the Anopheline of Africa South of the Sahara (Afrotropical Region). Johannesburg, South Africa: South African Institute for Medical Research. Publication no. 55.

  • 10

    Gillies MT, DeMeillon B, 1968. The Anopheline of Africa South of the Sahara (Ethiopian Zoogeographical Region). Johannesburg, South Africa: (South African Institute for Medical Research.

  • 11

    Mendis C, Jacobsen JL, Gamage-Mendis A, Bule E, Dgedge M, Thompson R, Cuamba N, Barreto J, Begtrup K, Sinden RE, Hogh B, 2000. Anopheles arabiensis and An. funestus are equally important vectors of malaria in Matola coastal suburb of Maputo, southern Mozambique. Med Vet Entomol 14 :171–180.

    • Search Google Scholar
    • Export Citation
  • 12

    Kenya National Bureau of Statistics, 1999. Executive Report: Summary Statistics. Nairobi, Kenya.

  • 13

    Githeko AK, Service MW, Mbogo CM, Atieli F, Juma FO, 1993. Plasmodium falciparum sporozoite and entomological inoculation rates at the Ahero rice irrigation scheme and the Miwani sugar-belt in western Kenya. Ann Trop Med Parasitol 87 :379–391.

    • Search Google Scholar
    • Export Citation
  • 14

    Petrarca V, Beier JC, 1992. Intraspecific chromosomal polymorphism in the Anopheles gambiae complex as a factor affecting malaria transmission in the Kisumu area of Kenya. Am J Trop Med Hyg 46 :229–237.

    • Search Google Scholar
    • Export Citation
  • 15

    Mbogo CM, Snow RW, Khamala CP, Kabiru EW, Ouma JH, Githure JI, Marsh K, Beier JC, 1995. Relationships between of Plasmodium falciparum transmission by vector populations and the incidence of severe disease at nine sites on the Kenyan coast. Am J Trop Med Hyg 52 :201–206.

    • Search Google Scholar
    • Export Citation
  • 16

    Schneider DC, 1994. Quantitative Ecology. New York: Academic Press.

  • 17

    Mendenhall W, Sincich T, 1996. A Second Course in Statistics: Regression Analysis. Englewood Cliffs, NJ: Prentice-Hall, Inc.

  • 18

    Sudman S, 1978. Applied Sampling. New York: Academic Press.

  • 19

    Service MW, 1976. Mosquito Ecology. New York: John Wiley and Sons Publishing.

 
 
 

 

 
 
 

 

 

 

 

 

 

A GEOGRAPHIC SAMPLING STRATEGY FOR STUDYING RELATIONSHIPS BETWEEN HUMAN ACTIVITY AND MALARIA VECTORS IN URBAN AFRICA

View More View Less
  • 1 Department of International Heath and Development, Department of Tropical Medicine, and Department of Civil and Environmental Engineering, Tulane University, New Orleans, Louisiana; Centre for Geographic Medicine Research-Coast, Kenya Medical Research Institute, Kilifi, Kenya; Institute for Science and Public Policy, and School of Civil Engineering and Environmental Science, University of Oklahoma, Norman, Oklahoma; Centre for Vector Biology and Control Research, Kenya Medical Research Institute, Kisian, Kenya; Human Health Division, International Centre of Insect Physiology and Ecology, Nairobi, Kenya

This paper describes a geographic sampling strategy for ecologic studies and describes the relationship between human activities and anopheline larval ecology in urban areas. Kisumu and Malindi, Kenya were mapped using global positioning systems, and a geographic information system was used to overlay a measured grid, which served as a sampling frame. Grid cells were stratified and randomly selected according to levels of planning and drainage. A cross-sectional survey was conducted in April and May 2001 to collect entomologic and human ecologic data. Multivariate regression analysis was used to test the relationship between the abundance of potential larval habitats, and house density, socioeconomic status, and planning and drainage. In Kisumu, 98 aquatic habitats were identified, 65% of which were human made and 39% were positive for anopheline larvae. In Malindi, 91 aquatic habitats were identified, of which, 93% were human made and 65% were harboring anopheline larvae. The regression model explains 82% of the variance associated with the abundance of potential larval habitats in Kisumu. In Malindi, 59% of the variance was explained. As the number of households increased, the number of larval habitats increased correspondingly to a point. Beyond a critical threshold, the density of households appeared to suppress the development of aquatic habitats. The proportion of high-income households and the planning and drainage variables tested insignificant in both locations. The integration of social and biologic sciences will allow local mosquito and malaria control groups an opportunity to assess the risk of encountering potentially infectious mosquitoes in a given area, and concentrate resources accordingly.

INTRODUCTION

The rapid, unplanned urbanization observed in many parts of Africa is changing the context for human population and natural system interactions. To study the complex nature of mosquito-human relationships, it is necessary to understand that human migration, population growth, socioeconomic status and behavior, and the modified environment can potentially affect the dynamics of natural systems. These demographic and physical factors generally determine the degree to which humans are interacting with the natural environment and subsequently influencing ecologic systems. This underscores the importance of understanding how humans modify the environment to affect malaria vector populations and subsequent parasite transmission intensity in urban areas.

In Africa, most studies of anopheline mosquito larval ecology have been done in rural settings.1,2 However, several key studies in Africa have identified factors that contribute to the maintenance of anopheline mosquito populations in urban environments. Robert and others identified the presence of market-garden wells in Dakar, Senegal as important larval sites for Anopheles arabiensis.3 Trappe and Zoulani found ditches, gutters, and tire tracks to be important anopheline larval sites in Brazzaville, Congo.4 In a newly urbanized area of western Kenya, Khaemba and others concluded that An. gambiae preferred human-made, temporary pools of water, such as tire tracks and ditches for ovipositing in the rainy season.5 Although pollution generally inhibits the development of anopheline larvae in urban centers,6,7 and anopheline densities are normally higher in rural areas,8 urban populations may still be at risk of encountering potentially infectious mosquitoes. Anopheles funestus, which typically develops in permanent aquatic habitats,9,10 has already shown it can adapt to a wide range of habitats.11 Poverty, urban farming, deteriorating infrastructures, and overcrowding in sub-Saharan African urban areas contribute to the development of conditions that modify anopheline habitats. This modification either increases or decreases the number of mosquitoes an area can support, which directly affects the number of potentially infectious mosquitoes in an area. If households or communities are unable to effectively protect themselves, the risk of acquiring an infectious bite increases. As such, understanding the effect of human activity on malaria vector population dynamics is important for identifying areas in need of increased source reduction efforts and environmental management.

This paper describes a geographic sampling strategy for studying mosquito-human interactions in urban environments. A multivariate regression model was used to test hypotheses and demonstrate the utility of the sampling strategy. A cross-sectional study design provided the framework for quantifying the relationship between human activity and the abundance of potential larval habitats in Kisumu and Malindi, Kenya. Data on the relationships between house densities, levels of planning and drainage, socioeconomic status, and the abundance of aquatic habitat provide a foundation for characterizing and comparing mosquito-human interactions. Although this study focuses on malaria vectors, the sampling strategy described below has broad applications within the social and biologic sciences and can be used in conjunction with multiple research designs in both rural and urban environments, and at various scales.

MATERIALS AND METHODS

Study areas.

Kisumu and Malindi, Kenya were chosen as urban study areas to carry out preliminary geographic and ecologic studies (Figure 1). The existence of laboratory facilities and assembled field teams, community support, and logistical feasibility were the criteria used to select the two urban study areas. Kisumu is the second largest urban center in Kenya (population ≈ 204,000) and is located 10 km south of the equator on Lake Victoria in Nyanza Province.12 Mean daily minimum and maximum temperatures are 18°C and 30°C, respectively, as reported by the Kisumu Meteorological Research Station located 2 km west of town at the Kisumu Airport. Urban Kisumu is made up of residential, commercial, and industrial areas with patches of tall grass savanna, swamp, and undeveloped agricultural land. The altitude is approximately 1,150 meters at the highest point and 1,075 meters at the lowest point. Western Kenya has two distinct rainy seasons, March through May and September through December, with average annual levels of precipitation ranging from 1,000 to 1,500 mm. Kisumu is a major regional transportation center where populations are engaged in formal and informal economic activity. The anopheline vectors of medical importance in this region are An. gambiae, An. arabiensis, and An. funestus.13,14

Malindi is the tenth largest urban center in Kenya (population ≈ 53,000) and is located on the shore of the Indian Ocean in Coast Province 108 km north of Mombasa.12 Mean daily minimum and maximum temperatures are 22°C and 30°C, respectively, with the average relative humidity at 65%, as reported by the Malindi Meteorological Station located 3 km southwest of town at the Malindi Airport. Malindi is comprised of commercial and residential areas, with altitudes ranging from sea level to approximately 50 meters. Coastal Kenya also has two distinct wet seasons, April through June and October through November. Precipitation varies from 75 to 1,200 mm per year throughout the coastal plain. Tourism, fishing, and trading are the major economic activities in this area. The anopheline vectors of medical importance in this region are An. gambiae, An. funestus, An. merus, and An. arabiensis.15 A peri-urban area adjacent to Malindi was also included in this study as a basis for ecologic comparison with urban environments.

Census data (1979, 1989, and 1999), town maps, and district develop plans were obtained from District Development Offices and the Kenya National Bureau of Statistics for each urban study area. The information was reviewed to ascertain environmental and human population characteristics and relative locations of administrative boundaries. The information contained in those documents formed the basis for urban boundary demarcation and subsequent sample frame identification and development.

Base-maps for each study area were prepared using ArcView 3.2® (Environmental Systems Research Institute, Redlands, CA), a geographic information system (GIS). Data points were collected using a Trimble (Sunnyvale, CA) global positioning system (GPS). The base maps illustrate all major and minor roadways and waterways. Latitude, longitude, and elevation data were collected for each point, and information relative to land-use type and specific levels of drainage were noted. Municipal water sources, major market centers, and other significant landmarks were also identified and georeferenced. Although both Kisumu District and Malindi District cover a large area of land, not all sub-locations within the respective districts are considered urban in nature. As a result, and for purposes of this study only, urban Kisumu and urban Malindi were considered to include only those sub-locations, or parts of a sub-location that have high population densities or were contiguous to residential areas but were in fact industrial or commercial in nature. Outlying sub-locations that did not fit the selection criteria were not considered further. Ground-truthing was done to validate the accuracy of the base-maps and to check sub-location inclusion criteria.

Intuitively, the size of a study area raises the issue of how to obtain accurate estimates for key entomologic and human ecologic variables that support systematic hypothesis testing.16 A properly designed stratified probability sample makes it possible to infer the characteristics of a large study area based on a relatively small sample of sites selected from the study area.17,18 As such, the goal was to develop a random sampling scheme that captured both the biologic and human ecological effect of mosquito-human interaction with minimal bias and maximum heterogeneity with respect to the distribution and types of aquatic habitats and households. Unless otherwise noted, details of the sampling scheme are the same for both Kisumu and Malindi.

The initial step involved the segmentation of the study areas. The GIS was used to overlay a series of 270 meter × 270 meter grid cells, corresponding to a 9 pixel × 9 pixel LANDSAT (National Aeronautics and Space Administration Goddard Space Flight Center, Greenbelt, MD) Thematic Mapper remote sensing satellite imagery, on each study area (Figure 2). Each grid cell was assigned a unique identification number. In Kisumu, 317 grid cells fell within the urban study area. In Malindi, 244 grid cells fell within the urban study area and 164 grid cells fell within the peri-urban study area. These numbers constituted the sampling frame, from which the stratified sample of grid cells was selected.

The stratification process involved assessing the level of planning and drainage in each grid cell and assigning a value: 1) planned, well drained; 2) planned, poorly drained; 3) unplanned, well drained; 4) unplanned, poorly drained; and 5) peri-urban. Because the strata contained different grid cell frequencies, probability proportionate-to-size sampling was used to select grid cells. District development plans, existing town maps, GIS base maps, and ground-truthing were used to determine specific levels of planning. House spacing, presence or absence of engineered drainage systems, types and patterns of roads, and community water sources were examined to determine whether an area was planned or unplanned. Well-drained areas versus poorly drained areas were determined by the presence or absence of functional engineered drainage systems and topographic features within each grid cell. The distribution of grid cells per stratum and the corresponding proportion for each city are listed in Table 1.

A systematic sample with a random start was used to select individual grid cells within each stratum. The sampling interval was calculated for each stratum using the formula I = F/S, where I is the sampling interval for each stratum, F is the total number of grid cells in each stratum, and S is the desired sample size for each stratum, proportionate to the actual number of grid cells listed in each stratum. In cases where the interval was not a whole number, the outcome was rounded to the nearest integer. Grid cells were listed in consecutive order for each stratum. A random number generator was used to select the first grid cell. Every Ith grid cell in the respective strata was then selected. This insured that the probability of selection was equal for each grid cell. The number of grid cells selected per stratum for each urban study area is listed in Table 1. The total number of grid cells selected per study area (n = 20 in Kisumu and n = 28 in Malindi) was a function of time and logistic feasibility. All selected grid cells were geo-referenced and located in the field using GIS base maps, existing landmarks, and a compass. The 270 meter × 270 meter grid cells served as a sampling frame for the entomologic and human ecologic components of the project. The distribution of the randomly selected grid cells used to collect data in this study is shown in Figure 3.

Household counts and data collection.

A questionnaire was developed, pre-tested, and administered in both study areas to characterize and compare socioeconomic conditions across strata. Households administered the socioeconomic questionnaire were selected from within the randomly selected grid cells. Approximately 100 household questionnaires were completed in each stratum for both Kisumu and Malindi. Ten houses were identified from each selected grid cell by locating the centroid and using random direction to serve as a house selection guide. In strata containing fewer than 10 randomly selected grid cells, the 100 households to be surveyed were divided equally among the total number of selected grid cells falling within the stratum. In grid cells where houses were either absent or inaccessible, questionnaires were administered to households in adjacent grid cells of the same stratification type. A unique identifier was assigned to each adjacent box falling under the same stratification type. A random number was generated to select the adjacent box. Houses in adjacent boxes were selected for inclusion using the same random direction method described earlier.

The socioeconomic variable used in this analysis was equal to the proportion of households classified as high-income per grid cell. Scaled ownership data was used to classify households as high, medium, or low-income, based on item ownership reported on the questionnaire. Households were considered high-income if a household reported owning a car or satellite dish, plus a radio, bicycle, or television. Proportions were estimated by counting the number of high-income households in each selected grid cell, plus the number of high income households in any adjacent grid cells of the same stratification type sampled, divided by the total number of households sampled in the respective grid cells.

House density data were obtained by counting the total number of occupied households contained within each 270 meter × 270 meter selected grid cell. Two independent observers performed household counts. In grid cells where the two counts yielded different results, two additional counts were performed and the average of the four counts was used.

Entomologic sampling.

All accessible aquatic habitats were located, counted, and characterized within each selected grid cell during April and May of 2001. All aquatic habitats were visited to avoid the bias associated with collecting samples only at sites most likely to contain anopheline larvae. Aquatic habitats included in this analysis were identified as bodies of water harboring at least one anopheline larvae over the two month period, and bodies of water with no anopheline larvae present but within 20 meters of an active larval site and similar with respect to habitat type. In this analysis, multiple water-filled containers in close proximity (i.e., trash dumps, tire piles, etc.) were considered to be one aquatic habitat. Likewise, artificial water storage containers existing in isolation of other containers were considered to be one aquatic habitat, respectively, in this analysis. Standard dipping methods were used to collect mosquito larvae at each site.19 Larvae were collected, preserved in alcohol, and transported to the laboratory for further identification. Data on habitat type and environmental attributes of the area were also recorded at each aquatic habitat. Although many aquatic habitats yielded species other than Anopheles, they were not considered further in this analysis. The complete results from the entomological component of the project will be reported in separate papers for Kisumu and Malindi, respectively.

Data analysis.

Pearson’s correlation coefficient (r) was calculated to test for associations between the number of potential anopheline larval sites and the number of households, and the proportion of high-wealth households, per grid cell. A multivariate regression analysis was also conducted to determine the nature of any relationship that may exist between the abundance of potential anopheline larval sites present per square area (dependent variable) and house density, socioeconomic status, and the level of planning and drainage (independent variables) in both Kisumu (n = 20) and Malindi (n = 28). A graphic review of a scatterplot suggests that the number of larval sites increases as the number of households increase, but only to a point, and then decreases as the number of households continues to increase (Figure 4). Thus, establishing the degree of curvature associated with increasing house densities, while controlling for planning and drainage, may contribute significant information to the model. A second-order quadratic term was introduced into the equation to account for any curvilinearity that may exist in the house density data.

The research hypothesis is that house density, the proportion of high-income households, and an area’s level of planning and drainage are significant factors affecting anopheline larval site abundance in an urban setting. Specifically, we tested whether larval site abundance varies in relationship to house density, decreases with increasing proportions of high-income households, and varies accordingly as a function of planning and drainage in the grid cell. The regression model used to test this hypothesis is (E)Y = βo + β1X1 + β2X12 + β3X2 + β4X3 + β5X4 + β6X5 + β7X6 + E, where (E)Y is the predicted number of potential larval sites per 270 meter × 270 meter grid cell; βo is the Y intercept, which is constant; X1 is the number of occupied households per 270 meter × 270 meter per grid cell; X12 is the number of occupied households per 270 meter × 270 meter grid cell squared; X2 is equal to the proportion of households classified as high income in each grid cell; X3 is equal to 1 if area is planned and poorly drained and 0 if not; X4 is equal to 1 if area is unplanned and well-drained and 0 if not; X5 is equal to 1 if area is unplanned and poorly drained and 0 if not; X6 is equal to 1 if area is per-iurban and 0 if not; and E is the error term, which is assumed to be normally distributed with a mean value of zero. The planned and well-drained stratum is the reference category for both Kisumu and Malindi. In Kisumu, no peri-urban stratum was identified; thus, X6 was omitted from the regression. Dataset construction and data analysis was carried out in Arc-View 3.2®, Microsoft (Bellevue, WA) Excel®, and SPSS version 10 (SPSS, Inc., Chicago, IL) software programs.

RESULTS

In Kisumu, 98 aquatic habitats were identified from within the 20 grid cells. Of the habitats identified, 65% were human made and 39% were harboring Anopheles larvae. Of the 375 anopheline larvae collected, 320 were as identified as An. arabiensis and 25 were identified as An. gambiae. The remaining anopheline larvae were not identifiable. Human-made habitat types consisted mainly of broken pipes, roadside ditches and potholes, and temporary pools of water along unpaved roads and paths within and around family compounds. Natural habitats included swamp, streams, and ponds. In Malindi, 91 aquatic habitats were identified from within the 28 grid cells. Sixty-five percent of those sites were positive for anopheline larvae, while 93% of the sites were human made. Of the 94 anopheline larvae collected, 83 were identified as An. gambiae, 4 as An. arabiensis, and 1 as An. merus. The remaining anopheline larvae were not identifiable. Important types of human-made larval sites in Malindi included drained or abandoned swimming pools, tire tracks, shallow garden wells, and temporary pools of water along unpaved roads and paths within and around family compounds. Natural habitats consisted mainly of swamp, streams, and temporary puddles.

The proportion of total aquatic habitats identified varied across strata in both Kisumu and Malindi. In Kisumu, the unplanned poorly drained stratum contained 70% of the total aquatic habitat identified, while the planned well-drained stratum contained 16% of the total (n = 98). In Malindi, 38% of the total aquatic habitats identified were contained in the unplanned poorly drained stratum, 21% were found in the peri-urban stratum, and 16% of the total were contained in the planned well-drained stratum (n = 91). In this analysis, the peri-urban stratum identified in Malindi appears to be an ecologic extension of the unplanned poorly drained stratum, with similar environmental and physical attributes. In Malindi, 40% (n = 88) of all Anopheles larvae were collected from the peri-urban and unplanned poorly drained grid cells. In Kisumu, anopheline larvae were collected from all strata, although the unplanned, poorly drained stratum had the highest proportion of aquatic habitats positive for anopheline larvae.

The selected socioeconomic variables collected during this project are listed in Table 2. In Malindi, results from the household survey suggest that household socioeconomic status decreases across the strata. The proportion of Malindi households within the high, medium, and low-income groups was 19.4%, 9.8%, and 70.8%, respectively, with 38% reporting high income in the planned well-drained stratum, 33% in the planned poorly drained stratum, 16.3% in the unplanned well-drained stratum, 5% in the unplanned poorly drained stratum, and 4% in the peri-urban stratum.

In Kisumu, the proportion of households within the high, medium, and low-income groups was 13.1%, 6.3%, and 80.7%, respectively. There was no clear socioeconomic division across strata, with 24.3% reporting high income in the planned well-drained stratum, 11% in the planned poorly drained stratum, 21.4% in the unplanned well-drained stratum, and 4.2% in the unplanned poorly drained stratum. However, variables included in Table 2 suggest that a socioeconomic division exits between planned areas only. Moreover, data on household ownership indicates that those households in the planned strata report having more luxury items on average than those households sampled in the unplanned strata.

In Kisumu, a significant negative correlation was observed between the abundance of potential larval sites and the proportion of households reporting high-income status per grid cell (r = −0.554, degrees of freedom [df] = 18, P = 0.011). Thus, as the proportion of households reporting high-income increases, the number of potential larval habitats decreases. In Malindi, no correlation was detected (r = −0.08, df = 26, P = 0.685) between the abundance of potential larval sites and the proportion of high-income households per grid cell. A significant positive correlation was observed between the abundance of potential larval habitats and the total number of occupied households per selected grid cell in both Kisumu (r = 0.653, df = 18, P < 0.001) and Malindi (r = 0.533, df = 26, P < 0.001).

Results from the multivariate regression analysis indicate that the number of households per selected grid cell is a significant factor affecting the abundance of potential larval sites in both Kisumu and Malindi independently, after controlling for the areas level of planning, drainage, and socioeconomic status (Table 3). An F value of 9.97 (df = 13, P < 0.001) in Kisumu and 4.17 (df = 20, P = 0.006) in Malindi indicates that the regression model is significant in both locations, with 82% of the variance in Kisumu and 59% of the variance associated with potential anopheline larval site abundance in Malindi explained. The standard error of the estimate is 1.46 in Kisumu and 1.92 in Malindi.

Although the proportion of high-income households per grid cell did not test significant in either location, the sign of the beta coefficient was in the expected direction, indicating an inverse relationship with the abundance of potential larval habitats in both Kisumu and Malindi (Table 3). One possible explanation is that the previously defined planning and drainage strata are proximate determinants of the areas level of socioeconomic status. A partial F test in Kisumu (F = 0.127, df = 1,13, P > 0.1) and Malindi (F = 0.073, df = 1,20, P > 0.1) indicates that the high-income variable is not a statistically significant addition to the model after accounting for the variability in index values associated with house density, planning, and drainage.

Planning and drainage variables also tested insignificant in this model for both Kisumu and Malindi (Table 3). However, in Kisumu, the signs of the beta coefficients were in the expected directions, with the well-drained strata having a negative affect, and the poorly drained strata having a positive affect on the abundance of potential larval habitats per grid cell. In Malindi, the unplanned well-drained stratum has a negative affect on the abundance of potential larval habitats. The unplanned, poorly drained; planned, poorly drained; and peri-urban strata also have negative coefficients, which is opposite to what one might expect. The planned, well-drained stratum appears to have a positive affect on the dependent variable, which is also opposite of what one might expect. No interaction was detected between the independent variables in either location.

An example of the estimation equation with parameter values from Table 3 is given for Kisumu and Malindi to show the relationship between first-order and second-order terms across strata. In Kisumu, (E)Y = −0.806 + 0.06362(50) − 0.0001393(2500) − 1.804(0.25) + 2.297(1) + E, where (E)Y is equal to 3.87 potential larval sites when the number of occupied households in an unplanned poorly drained grid cell is 50, and the proportion of high-income households is 25%. In Malindi, (E)Y = 2.412 + 0.06473(50) − 0.0001743(2500) − 0.981(0.25) − 3.831(1) + E, where (E)Y is equal to 1.14 potential larval sites when the number of occupied households in an unplanned well-drained grid cell is 50, and the proportion of high-income households is 25%.

DISCUSSION

In both Kisumu and Malindi, most larval sites were human made, with the highest numbers of aquatic habitats observed in the unplanned, poorly drained stratum in Kisumu, and the unplanned, poorly drained and peri-urban strata in Malindi. In Kisumu, nearly all larvae collected were An. arabiensis, while in Malindi, most larvae collected were An. gambiae s.s., yet the types of habitats harboring anopheline larvae were quite similar in both cities. Although this observation is interesting and may be extremely important with respect to the epidemiology of malaria parasite transmission, and it is known that the distribution of specific larvae among habitat is a function of adult mosquito productivity in the area, the cross-sectional approach used in this study coupled with limited adult mosquito sampling precludes the comprehensive identification of factors responsible for this phenomena. As such, a key assumption we make in this analysis is that the abundance and distribution of potential larval sites, as measured here, is a proximate determinant of the number of adult mosquito vectors in the area, although we recognize that a plethora of factors interact on multiple scales to affect the actual abundance and distribution of adult anopheline mosquitoes.

The number of houses per selected grid cell was greatest in the unplanned strata for both cities. This suggests that infrastructure development, sanitation service development, and housing development are important factors affecting the development of potential anopheline larval habitats. Moreover, the unplanned, poorly drained strata were the poorest with the least education and access to piped water and sewage systems, compared with the planned strata for both study areas. This suggests that the socioeconomic status of the area, or a household’s level of wealth, may also be important determinants of larval site abundance in urban environments. Again, because our goal was to stratify in such a way as to capture both the entomologic and human ecologic affect of mosquito-human interaction, our assessment of the level of planning and drainage for each area was consistent with the relative socioeconomic status of the same area.

In Malindi, the unplanned, poorly drained; planned, poorly drained; and peri-urban strata have negative coefficients, while the planned, well-drained stratum has a positive coefficient value. One explanation for this unexpected directional effect is that many planned, well-drained areas have households with part-time residents and seasonal swimming pools, which when drained in the low (off) season, fill with rain water and provide excellent habitat for anopheline larvae. Moreover, most aquatic habitats in Malindi were characterized as human made (93%); thus, the ability of an area to drain excess water becomes less important as an independent variable. The abundance of larval sites in areas of high wealth, coupled with the prevalence of human-made aquatic habitats, helps to explain the unexpected results observed in Malindi, as well as the expected results observed in Kisumu, with respect to planning and drainage.

The informal economy of the two cities also plays a major role in the distribution and abundance of potential anopheline larval sites. In many of the planned, well-drained areas where water delivery systems exist, we found broken pipes and pools of residual water. Our sampling indicated that most contained anopheline larvae. These pipes are broken in an attempt to procure free water for sale or consumption, or were poorly installed and break as a result of passing vehicles. Moreover, community taps and urban garden wells in areas of high population density provide excellent aquatic habitat for anopheline mosquitoes. As people move into an area that lacks adequate infrastructure, sanitation services, and water delivery systems, there is more community use of the taps, increases in the number of artificial water storage containers, and greater levels of standing water in the area. Car-washing activities also contribute to the creation of potential larval sites, if engineered drainage systems are lacking or poorly maintained. Construction activities can also affect anopheline larval site abundance. Open containers, excavation projects, and container type equipment offer excellent habitats for larval development if exposed to the elements for sustained periods of time. In both Kisumu and Malindi, large and small-scale excavation and construction projects are common.

The modified environment normally reduces the abundance of mosquitoes by eliminating aquatic habitats. Vegetation is often replaced with asphalt, concrete, brick, stone, and housing and market centers. Drainage systems are installed and water delivery systems are introduced, thus further reducing anopheline habitat and the amount of standing water available for larval development. However, in both Kisumu and Malindi, populations are still actively involved in rural-type activities (e.g., urban farming/gardens, keeping livestock), and human settlements have developed in marginalized, unplanned, and peri-urban areas where patches of vegetation remain and deteriorating infrastructure is common. In these areas, domestic gray water is often dumped in the open environment, rainwater pools in the ruts and potholes of unpaved roads, and domestic water is often stored in the open environment to compensate for inadequate water delivery systems. Furthermore, trash and debris often accumulate in common areas, offering additional mosquito habitats, as shown by the number of larval habitats littered with human-made pollutants such as detergents, raw sewage, plastics, and an assortment of domestic debris. However, at high population densities, the amount of open space and vegetation for adult and larval mosquito development is extremely limited. Moreover, the aquatic habitats that are often found at this density are often too polluted, or too temporary for anopheline larval development.

In general, the sampling scheme was quite effective. Field teams were able to use the base maps, locate the randomly selected grid cells, and carry out entomologic and socioeconomic data collection with minimal error. Although the stratification process generally assigned grid cells to the correct strata, some grid cells had areas that were both planned and unplanned, and drained and poorly drained. In a few instances, grid cells that appeared to be well drained at the onset of the stratification process were in fact poorly drained. In general, drainage is a function of topography, location, and economic resources in both Kisumu and Malindi. Affluent areas receive the bulk of infrastructure funds and services, while the poorer areas and newly developed slums generally receive less municipal support. As a result, drains are installed and maintained in select areas only, which may explain the high abundance of aquatic habitats identified in the unplanned, poorly drained strata. Furthermore, areas with paved roads, evenly spaced houses, intact infrastructure, and other indicators of good planning are highly desirable. As such, squatter settlements, housing projects, and new low-market areas are developed in planned areas. These pockets of intense poverty tax the existing infrastructure and serve as a possible explanation for variations in the distribution and abundance of anopheline larval sites and house densities within and among planned strata. However, the unplanned, well-drained stratum in both cities had low numbers of aquatic habitats, suggesting that slope and soil type may also be important determinants of larval habitat abundance in urban environments.

Moreover, because vector distribution is a function of topography, and site-specific topographic conditions can lead to a non-random distribution of vectors or vector habitat, a cross-sectional study design and random sampling strategy, which results in a random selection of samples, may fail to capture important spatial and temporal entomologic information in specific instances. Although the topographic effect was minimal in both Kisumu and Malindi, it may be an important consideration in urban areas with significant topographic variation. It is also important to note that the standard error of the prediction for both locations in this analysis is quite large. As such, the relationship between larval site development and human activity in Kisumu and Malindi needs further investigation. Larger sample sizes will be needed to reduce the standard error of any estimate made, and to increase the power of the model to detect correlations and relationships that may exist between planning and drainage, socioeconomic status, and entomologic factors in urban areas.

In conclusion, we plan to develop a spatial surface that will allow us to predict variation in population dynamics by species based on differentials in environmental variables and human ecology. Using that surface, we will attempt to specify and validate a model linking human and anopheline population dynamics, while controlling for potential spatial autocorrelation in the residuals. This will enable us to systematically delineate and map significant sources of variation in human activity and environmental attributes that generally affect the risk of encountering potentially infectious mosquitoes in urban areas. In this way, we can provide key entomologic and human ecologic data for specific urban areas, and estimate the risk of mosquito-human contact within those communities. This will allow local mosquito control groups an opportunity to concentrate resources in the areas at the greatest risk for larval site development and subsequent mosquito-human contact.

Table 1

Distribution of grid cells across strata, the percent of the total grid cells per strata, and the number of grid cells selected for survey and sampling work in Kisumu and Malindi, Kenya

StrataTotal number of grid cellsPercent of totalGrid cells selected
Kisumu
    Planned, well-drained10733.756
    Planned, poorly drained134.11
    Unplanned, well-drained278.522
    Unplanned, poorly drained17053.6311
Total31710020
Malindi
    Planned, well-drained5723.364
    Planned, poorly drained6526.645
    Unplanned, well-drained31.331
    Unplanned, poorly drained11948.7710
    Urban total24410020
    Peri-urban1641008
    Total65210028
Table 2

Summary socioeconomic statistics from the household survey administered in Kisumu and Malindi, Kenya

Strata*
12345
*1 = planned, well-drained; 2 = planned, poorly-drained; 3 = unplanned, well-drained; 4 = unplanned, poorly-drained; 5 = peri-urban.
†NA = not available.
‡High education = if member of household attended form 4, highest secondary classes or higher.
Kisumu
    Number of households10310011395
    Mean number of people per night per household5.025.255.314.03
    % of households with electricity67.3076.0064.0011.50NA†
    % of households with acces to piped water60.0085.0033.3338.71
    % of households with access to main sewer49.00761.026.709.70
    % of households harboring livestock30.9025.5029.6028.30
    % of households with high education‡61.8069.0069.9041.10
    % of total aquatic habitats identified18.005.0010.0067.00
Malindi
    Number of households10010080100100
    Mean number of people per night per household3.973.985.805.756.01
    % of households with electricity83.0088.0044.0027.0011.00
    % of households with access to piped water59.0066.0055.0033.006.00
    % of households with access to main sewer15.008.002.503.001.00
    % of households harboring livestock32.0024.0016.3029.0061.00
    % of households with high education‡52.0043.0030.0037.0026.00
    % of total aquatic habitats identified18.0018.002.0041.0021.00
Table 3

Summary statistics for multivariate regression model estimating the relationship between house density, high income, planning and drainage, and the abundance of potential larval sites in Kisumu and Malindi, Kenya

Independent variablesCoefficientSEP
*Constant= Bo.
Kisumu
    Constant*−0.8061.2860.542
    House density0.006360.014<0.001
    House density2−0.0001390.0000.003
    High income−1.8045.0690.782
    Planned, well-drainedBaseline
    Planned, poorly drained2.1861.6700.213
    Unplanned, well-drained−1.2181.3080.369
    Unplanned, poorly drained2.2970.9830.036
R20.822
Global F test9.977
P value of Global F test<0.001
n = 20
Malindi
    Constant*2.4121.5670.139
    House density0.06470.015<0.001
    House density2−0.0001740.0000.003
    High income−0.9813.6300.790
    Planned, well-drainedBaseline
    Planned, poorly drained−1.4921.2960.263
    Unplanned, well-drained−3.8312.2090.098
    Unplanned, poorly drained−1.8931.3510.176
    Peri-urban−2.0721.4820.177
R20.492
Global F test4.165
P value of Global F test0.006
n = 28
Figure 1.
Figure 1.

Map of Kenya showing relative positions of study areas.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 68, 3; 10.4269/ajtmh.2003.68.357

Figure 2.
Figure 2.
Figure 2.

Sampling frame for urban Kisumu (A) and urban/peri-urban Malindi (B).

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 68, 3; 10.4269/ajtmh.2003.68.357

Figure 3.
Figure 3.
Figure 3.

Selected grid cells by strata in Kisumu (A) and Malindi (B).

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 68, 3; 10.4269/ajtmh.2003.68.357

Figure 4.
Figure 4.

Number of potential larval sites plotted against the number of households per selected grid cell in Kisumu (A) and Malindi (B).

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 68, 3; 10.4269/ajtmh.2003.68.357

Authors’ addresses: Joseph Keating and Kate Macintyre, Department of International Health and Development, TB46, School of Public Health and Tropical Medicine, Tulane University, 1440 Canal Street, New Orleans, LA 70112-2699, Telephone: 504-988-8258, E-mail: jkeating@tulane.edu. Charles M. Mbogo and Lydiah Kibe, Centre for Geographic Medicine Research, Coast, Kenya Medical Research Institute, PO Box 428, Kilifi, Kenya. Andrew Githeko and Bryson Ndenga, Centre for Vector Biology and Control Research, Kenya Medical Research Institute, PO Box 1578, Kisumu, Kenya. James L. Regens, Institute for Science and Public Policy, Sarkeys Energy Center, University of Oklahoma, 100 E. Boyd Room 510, Norman, OK 73019-1006. Chris Swalm, Tulane University Health Sciences Center, SL18, 1430 Tulane Avenue, New Orleans, LA 70112-2699. Laura J. Steinberg, Department of Civil and Environmental Engineering, 206 Civil Engineering Building, Tulane University, New Orleans, LA 70118. John I. Githure, Human Health Division, International Centre of Insect Physiology and Ecology, PO Box 30772, Nairobi, Kenya. John C. Beier, Department of Tropical Medicine, School of Public Health and Tropical Medicine SL17, Tulane University, 1430 Tulane Avenue, New Orleans, LA 70112-2699.

Acknowledgments: We are grateful for the assistance of all scientific and technical staff at the Center for Geographic Medicine Research-Coast and the Kenya Medical Research Institute–Center for Vector Biology and Control Research (Kisian, Kenya), particularly Maurice Ombok, Richard Amimo, Francies Atieli, Nellie Njoki, Samuel Kahindi, Salim Omar, and Mtawali Chai. We also thank Allen High-tower (Centers for Disease Control and Prevention, Atlanta, GA) for his continued support of the field-based GPS/GIS teams in Kenya. This paper is published with the permission of the Director of the Kenya Medical Research Institute.

Financial support: This research was supported by National Science Foundation grant DEB-0083602 and National Institutes of Health grant U19 AI-45511.

REFERENCES

  • 1

    Minakawa N, Mutero CM, Githure JI, Beier JC, Yan G, 1999. Spatial distribution and habitat characterization of anopheline mosquito larvae in western Kenya. Am J Trop Med Hyg 61 :1010–1016.

    • Search Google Scholar
    • Export Citation
  • 2

    Gimnig JE, Ombok M, Kamau L, Hawley WA, 2001. Characteristics of larval anopheline (Diptera: Culicidae) habitats in western Kenya. J Med Entomol 38 :282–288.

    • Search Google Scholar
    • Export Citation
  • 3

    Robert V, Awono-Ambene HP, Thioulouse J, 1998. Ecology of larval mosquito, with special reference to Anopheles arabiensis (Diptera: Culicidae) in market-garden wells in the urban area of Dakar, Senegal. J Med Entomol 35 :948–955.

    • Search Google Scholar
    • Export Citation
  • 4

    Trape JF, Zoulani A, 1987. Malaria and urbanization in central Africa: the example of Brazzaville. Part II: Results of entomological surveys and epidemiological analysis. Trans R Soc Trop Med Hyg 81 :10–18.

    • Search Google Scholar
    • Export Citation
  • 5

    Khaemba BM, Mutani A, Bett MK, 1994. Studies on the anopheline mosquitoes transmitting malaria in a newly developed highland urban area: a case study of Moi University and its environs. East Afr Med J 71 :159–164.

    • Search Google Scholar
    • Export Citation
  • 6

    Chinery WA, 1984. Effects of ecological changes on the malaria vectors Anopheles funestus and the Anopheles gambiae complex of mosquitoes in Accra, Ghana. J Trop Med Hyg 87 :75–81.

    • Search Google Scholar
    • Export Citation
  • 7

    Barbazan P, Baldet T, Darriet F, Escaffre H, Djoda DH, Hougard JM, 1998. Impact of treatments with Bacillus sphaericus on Anopheles populations and the transmission of malaria in Maroua, a large city in a savannah region of Cameroon. J Am Mosq Control Assoc 14 :33–39.

    • Search Google Scholar
    • Export Citation
  • 8

    Coene J, 1993. Malaria in urban and rural Kinshasa: the entomological input. Med Vet Entomol 7 :127–137.

  • 9

    Gillies MT, Coetzee M, 1987. A Supplement to the Anopheline of Africa South of the Sahara (Afrotropical Region). Johannesburg, South Africa: South African Institute for Medical Research. Publication no. 55.

  • 10

    Gillies MT, DeMeillon B, 1968. The Anopheline of Africa South of the Sahara (Ethiopian Zoogeographical Region). Johannesburg, South Africa: (South African Institute for Medical Research.

  • 11

    Mendis C, Jacobsen JL, Gamage-Mendis A, Bule E, Dgedge M, Thompson R, Cuamba N, Barreto J, Begtrup K, Sinden RE, Hogh B, 2000. Anopheles arabiensis and An. funestus are equally important vectors of malaria in Matola coastal suburb of Maputo, southern Mozambique. Med Vet Entomol 14 :171–180.

    • Search Google Scholar
    • Export Citation
  • 12

    Kenya National Bureau of Statistics, 1999. Executive Report: Summary Statistics. Nairobi, Kenya.

  • 13

    Githeko AK, Service MW, Mbogo CM, Atieli F, Juma FO, 1993. Plasmodium falciparum sporozoite and entomological inoculation rates at the Ahero rice irrigation scheme and the Miwani sugar-belt in western Kenya. Ann Trop Med Parasitol 87 :379–391.

    • Search Google Scholar
    • Export Citation
  • 14

    Petrarca V, Beier JC, 1992. Intraspecific chromosomal polymorphism in the Anopheles gambiae complex as a factor affecting malaria transmission in the Kisumu area of Kenya. Am J Trop Med Hyg 46 :229–237.

    • Search Google Scholar
    • Export Citation
  • 15

    Mbogo CM, Snow RW, Khamala CP, Kabiru EW, Ouma JH, Githure JI, Marsh K, Beier JC, 1995. Relationships between of Plasmodium falciparum transmission by vector populations and the incidence of severe disease at nine sites on the Kenyan coast. Am J Trop Med Hyg 52 :201–206.

    • Search Google Scholar
    • Export Citation
  • 16

    Schneider DC, 1994. Quantitative Ecology. New York: Academic Press.

  • 17

    Mendenhall W, Sincich T, 1996. A Second Course in Statistics: Regression Analysis. Englewood Cliffs, NJ: Prentice-Hall, Inc.

  • 18

    Sudman S, 1978. Applied Sampling. New York: Academic Press.

  • 19

    Service MW, 1976. Mosquito Ecology. New York: John Wiley and Sons Publishing.

Save