Surveys were conducted according to study protocols approved by the Centers for Disease Control and Prevention (CDC) and University of Notre Dame Institutional Review Boards (IRBs) and the Ethics Committee of Ste. Croix Hospital. During previous research in the area, researchers observed low rates of literacy in the population; therefore, all information regarding the study was translated into Haitian Creole, and verbal consent and assent were requested from parents and participants in both the school and community surveys. The consent form was read to potential study participants and/or parents. The reader of the consent form and a witness were then asked to sign the form to indicate the subject's agreement, in accordance with the IRB-approved protocol. Before surveys began in the schools, approval was obtained from the Haitian Ministry of Health and Population (MSPP), the Ministry of Education, department directors, and schoolmasters. Subsequently, meetings were held with parents to provide an opportunity to explain the survey and the potential risks and benefits to their child. Prior to the community survey, community leaders were informed of the survey's procedures and provided approval for the study to commence. At the time of the survey, household members were informed of survey objectives and procedures, at which time oral consent or assent was obtained for all participants.
Low-Prevalence Study Sites
In 2001, national mapping for LF was performed, as previously described, using antigen testing of 100–250 schoolchildren aged 6–11 years of age per commune (a district-sized political entity in Haiti) across all Haitian communes 
. Since financial resources were limited, communes of highest antigen prevalence (≥10% prevalence) were prioritized for MDA. Our evaluation focused on five, predominantly rural, communes in which antigen prevalence in the survey was ≤1%: Grand Goâve (0.8%), Hinche (1.0%), Thomazeau (0.6%), Moron (0.8%), and St. Louis du Sud (0.4%).
Within each of the communes, five to seven schools were selected for antigen testing. These public and private schools were in urban and rural areas. Following the acquisition of informed consent and assent, as detailed above, blood samples were collected from students and tested as described below. ICTs were performed on all consenting children on site and additional blood was taken for ELISA testing upon return to laboratory facilities. Questionnaires were administered to a parent or guardian of ICT-positive children. The questionnaires were designed to identify autochthonous cases—defined as those children thought to have acquired the infection in the community where testing was conducted based on responses to questions about the absence of travel and residence in the same community for the past five years.
Index Case Selection
Five to eight ICT-positive children were chosen from each commune as index cases for that area. Index cases were children identified as antigen-positive by ICT in the school survey, with recoded GPS coordinates who responded to the questionnaire, and, when possible, were chosen based on those with confirmatory ELISA results. The index cases became central points in the subsequent community surveys.
For the community surveys, households of index cases defined the center for each testing radius. Circles of 50–75 meters were used in more densely populated urban or peri-urban areas, and circles of 100–250 meters were used in sparsely-populated rural settings. After index cases were identified, all consenting members of index households, and a systematic random sample of the neighboring households were selected for testing by ICT. All neighboring houses within the test radius were mapped using global positioning system (GPS) TerraSync (Sunnyvale, CA). In an effort to test 100 persons per community, approximately 20 households were chosen, based on an estimate of five persons per household (unpublished data). To select these 20 houses, the total number of houses in the zone was divided by 20 to determine the sampling interval. Houses were selected from a numbered list using a randomly selected starting point and this sampling interval. The questionnaire and methods used for blood collection and testing were the same as those used for the school survey. A total of 1,633 persons of all ages were evaluated in the community survey. For our study, subjects were included in the analysis if they had not been previously defined as an index case in the school study, received an ICT test result, and GPS coordinates were available for their households (n
Blood Collection and Serologic Testing
A blood specimen was collected from each person tested in the school or community survey. Filarial antigen-status was determined by ICT (Binax, Portland, ME) by trained laboratory personnel at the time of blood collection. Finger prick blood (100 µl) was collected and results were read at 10 minutes following application to the antigen test card. Technicians were supervised, and all efforts were made to uphold the quality of the test in regards to environment and timing. An additional 200 µl of blood was collected for confirmatory antigen testing. Collected blood was stored overnight at 4°C. The tubes were centrifuged the following day and the collected sera were stored in the field at 4°C for several days until return to Hôpital Ste. Croix where they were stored at −20°C. Sera were used for subsequent serologic assays, including confirmation of antigen status for persons with positive or questionable ICT results by Og4C3 antigen enzyme-linked immunosorbent-assay (ELISA, TropBio, Queensland, Australia). For the school survey and the subsequent community survey, ICT results were used as the indicator of antigen status for all persons; furthermore, autochthonous index cases were confirmed as antigen-positives by Og4C3 ELISA, the current gold standard for identifying circulating filarial antigens. Antigen-positive persons were treated with a single dose of diethylcarbamazine (DEC, 6 mg/kg). Due to the low percentage of ICT positive persons in each commune and the logistic difficulty of night blood surveys, microfilaria levels were not assessed.
Data were analyzed using SAS 9.3 (Cary, NC, USA), Epi Info 6 (CDC, Atlanta, USA) and ArcGIS (v. 9.3.1, Environmental Systems Research, Inc., Redlands, CA, USA). Univariate, Mantel Haenszel chi-square and logistic regression techniques were utilized.
The outcome of interest for this analysis was antigen positivity as denoted by the ICT results performed in the field. Two separate case definitions were employed as indicators of possible exposure in this analysis. The definition of index case only required a positive ICT among children tested in the school survey and the availability of GPS data. Index cases would therefore serve as potential, but unconfirmed, reservoirs of infection. A second, more stringent case definition was applied for children with positive Og4C3 results and who were defined as autochthonous cases based on their answers to the survey. These individuals were referred to as autochthonous index (AI) cases. The exposure of interest was the distance from each person tested to the nearest index or autochthonous index case. In order to determine the ordinal categories which best represent the distance to cases, a sensitivity analysis was performed for dichotomized distances of 10, 20, 40, 80 and 160 meters. Analysis of distance when using the AI case definition revealed no antigen-positive persons in the 59–99 m group, so the categories of 59–99 and 100+ meters were combined into a 60+ meter group, which was then used as the referent for the crude and multivariate regression analyses.
Potential confounders and effect modifiers, including age, gender and commune were also considered based on previous literature and anticipated heterogeneity among the communes. For the purpose of modeling, age was dichotomized into <15 years and ≥15 years, with the age of 15 or younger to denote school-aged children.
A spatial cluster analysis was performed on mapped households in the four communes recording antigen positivity. The analysis tested the spatial clustering of antigen-positive persons (excluding index cases) through the use of a Bernoulli model in SatScan (version 9.1.1, Boston, MA). A separate cluster analysis was performed for each of the four communes that included confirmed antigen-positives to better elucidate micro-clusters. Both general and isotonic simulations were performed on the commune-specific data, the latter of which accounts for the inverse relationship between risk and distance from the center of the cluster 
. This type of simulation holds biological plausibility in representing the transmission patterns of vector-borne diseases.