Human demographics and socio-economic conditions
Household surveys will also record ecological disturbance created by development activities such as deforestation, construction and aquaculture production. Detailed information will be gathered about aquaculture production: presence or absence of a pond and other water sources near the home; surface area and distance of water from home; and condition and age of pond, degree of use, and malaria control methods. A large sample of households is required to ensure statistical power for analysis, model development, and accurately mapping epidemiological data. Local assistants will be used to make this endeavor possible. Sampling efforts will take place within a subset of all communities. Half of the participating households will be selected at random from lists maintained by municipal agents and aquaculture promoters. The remaining will be filled in to ensure a stratified sample. This compromise approach will allow a description of the “mean” population but will also ensure enough differentiation occurs among variables of interest for effective analysis. One adult from each selected household will be asked a series of questions in Spanish in a face-to-face interview. The information gathered from the questionnaire will allow a multivariate analysis of continuous (e.g. size of pond and distance to household) and categorical (e.g. degree of use and condition of pond) variables. Each home will receive an identification number and an exact location will be recorded using a global positioning system along with the presence and location of any constructed ponds, thus creating the spatial variables that will allow further insight into areas of greater malaria risk. Each household will also appropriately be categorized by level of assistance and contact with NGO or other aquaculture consultant. Consistent involvement with a supervising organization or greater community participation increases the probability of maintaining production.
The degree and extent of mobility at the household level will also be established through survey data. Respondents to the survey will be asked short, open-ended questions regarding their reason for migrating, time of settlement, and the rationale behind their economic and land use choices. This section will include occupational information, to establish patterns of movement and exposure, such as: daily, periodic or seasonal commuting, wood cutting, feeding fish and maintaining pond, and other cultivation activities. Additionally, respondents will be asked whether there are economic activities that would be preferred to the current ones, and major limitations preventing changes will be discussed. Land use decisions will allow us to cluster households and compare potential occupational risk factors of individual activities. Socioeconomic information will play an important role in deciphering this condition. For example, in relation to aquaculture activities and malaria, a household with substantial assets can afford preventive measures, implement mosquito control strategies, and maintain an active pond.
Landscape data
Historical remote sensing and land cover data will be assembled from multiple sources, including aerial photographs, maps, and satellite data from sensors such as Landsat MSS, TM, ETM, and ASTER. The interpretation of these data will depend on finer-scale ground-truthing derived from the household surveys. Using the data from medical clinics, epidemiological maps will also be created using the GIS package ArcView and MapInfo (Briet et al., 2003).
The selection of remote sensing data follows a research design based on relevant events occurring during the past 15 years in all study units. Events include changes in demographics, such as the agrarian colonization project (e.g. the late 1970s relocation of unemployed oil workers); changes in economics, such as the 1990s boom in NGOs development projects (e.g., fishpond building), landscape disturbances (e.g. the major floods of the 1980s and 1990s) and changes in epidemiology, such as the arrival of the invasive vector species Anopheles darlingi. The area to be covered includes the Iquitos-Nauta peri-urban communities, the urban centers of Iquitos and Nauta, and the neighboring rural districts of Indiana, Mazan and Fernando Lores (figure 1).
Epidemiological data
Primary epidemiological data will be collected from two main sources. First, the annual incidence records for malaria will be obtained from regional health centers. Health center data will represent passive case detection and not reflect asymptomatic individuals. This will not present a complete account of malaria incidence, but this is the current approach to reporting and diagnosis in all health centers in the Iquitos region. Second, in addition to health center data, we will conduct household surveys in which we will record symptomatic infection events based on self-reported malaria episodes. For areas like the Amazon basin where prevalence is high and communities are highly mobile, self-reported malaria is the most appropriate prevalence measure (Singer and Castro, 2001).
These two sources of data will allow us to construct a GIS model of malaria risk covering our study area. From this model we will establish areas of high, medium, and low malaria risk along a spatial gradient that includes urban, peri-urban, and rural areas. Our objective is to establish a systematic sampling protocol to determine the composition and relative abundance of mosquito species in all combinations of land use type and malaria risk in all of the sites where we will collect demographic and socioeconomic information. Characterizing the mosquito communities is essential to our study because vector capacity and Plamodium infection rate will allow us to estimate the intensity of malaria transmission using ecological parameters and evaluate the accuracy of our initial model (Sithiprasasna et al. 2004).
Vector distribution and abundance information has previously been collected by UNAP researchers, and will be utilized in this study. With our assistance, local researchers will be able to bolster the existing data using outdoor Center for Disease Control (CDC) light traps at multiple locations throughout the study area. Locations will include; households, fish ponds, cleared land, road margins, forested parcels, and traditional farm plots. CDC light traps will be used because long-term sampling with human-bait collection methods poses higher risk for participants and is labor-intensive (Sithiprasasna et al. 2004). Furthermore, CDC light traps may prove important for detecting Plasmodium infection rates (IR) in mosquito samples. Hii et al. (2000) found that P. falciparum and P. vivax infection rates in mosquitoes were higher in light trap collections than human landing collections, indicating that light traps may select for older mosquitoes. Using both methods will ensure an accurate depiction of mosquito abundance and distribution in each of our sites.
Mosquitoes will be tested for P. falciparum and P. vivax infection by local technicians at the Peruvian Amazon Research Institute (Instituto de Investigaciones de la Amazonía Peruana, IIAP) using ELISA (de Arruda et al. 2004). ELISA allows rapid and accurate determination of malarial sporozoite loads within mosquito samples by using monoclonal antibodies to identify proteins on the sporozoite surface (de Arruda et al. 2004). Determination of infection rate in sampled mosquitoes is important in understanding transmission dynamics within our spatial gradients. Roshanravan et al. (2003) reported that the proportion of Plasmodium-infected An. darlingi is less than 0.1-0.5%; therefore, our local partners’ collection will indicate whether the infection rate has increased or remained stable over time. This information will also be crucial in determining if An. darlingi acts as the primary vector for malaria transmission within our study areas. Recent studies suggest that An. benarrochi may serve as a principle vector for malaria transmission, especially in those areas where An. darlingi is absent (Flores-Mendoza et al. 2004).
By controlling for the composition of vector communities and the IR in rural and urban sites, we will establish the relative importance of increased mobility from peri-urban areas in determining the malaria risk level (hypotheses 1 and 2).
Using remotely sensed data, we will look for correlations between malaria risk and hydrological features (fish ponds, lakes and flood plains), vegetation type and density, and structural features of urbanization. We expect that peri-urban and rural areas of high malaria risk will be located in areas with a high density of fish ponds, newly excavated roads, deforestation and recently settled communities (hypothesis 3).
