Spatial Analysis Optimizes Malaria Prevention Measures
After each day's mapping, the crews returned to Kisian and uploaded the location and attribute data from TDC1s into Pathfinder Office. Although the GPS equipment used was capable of real-time differential correction, this was not done because it would have required radio broadcasting of the correction signals from the base station. In Kenya, broadcast licenses are difficult for foreigners to obtain, and the field site was located 40 kilometers away over rocky terrain.
Instead, Pathfinder Office used the base station GPS data to perform differential postprocessing to correct the accuracy of the feature location measurements collected in the field. The attribute data was edited and cleaned up using Pathfinder Office software and added, along with feature points, to the GIS for Asembo and Gem.
Over the course of the project, CDC upgraded its equipment. In late 2003, it began using integrated Trimble GeoXT receivers. These units combine a rugged GPS receiver and data collector in one unit that eliminates complicated cable hookups and greatly simplifies fieldwork.
With the initial basemap and data collection completed, CDC began distributing bed nets to half the villages in the study area. Houses in villages with the nets were noted in the GIS. From that point, information gathering shifted primarily to the collection of health and related statistics. Crews kept careful records of those who became infected with malaria, those who became ill, and those who died. Field teams counted mosquito larvae found in stagnant water as well as live mosquitoes captured in traps placed throughout the project area. In all cases, the results were mapped in the GIS.
In addition to tracking illness, the crews continually updated existing data and gathered new data on the study area. As new houses were built or families moved to other villages, these occurrences were recorded. In later phases of the project, feature map updating became a much simpler process. Trimble TerraSync software allowed the researchers to download GIS map and attribute data from the office PC in Kisian to the GeoXT units. Crews took GIS data into the field, updated it there, and uploaded the updated files to the enterprise GIS at the end of the day.
After several years, clear trends began emerging in the bed net study. Results were quantified by multivariate analysis of the relationships among villages with bed nets; villages without nets; mosquito density surveys; and instances of malarial infection, sickness, and death. In the early phases of the program, GIS data was downloaded to a statistics software package to help unravel these complex spatial relationships. However, after CDC upgraded to ArcGIS ArcView 8.3, this analysis could be performed and visualized using the ArcGIS Geostatistical Analyst extension.
"The results were surprising," said CDC's Hawley. "Bed nets not only reduced malarial infection within the villages where they were used, but they also benefited people in villages without nets." The study confirmed that bed nets generated significant results in all statistical categories in the villages that received them. Incidents of bites, sickness, and death all fell. In children less than two years old, mortality was reduced by 17 percent, clinical malaria attack rates were reduced 75 percent, and mosquito biting rates were reduced by more than 90 percent.
It may seem redundant to track all three of these seemingly related statistics (bites, sickness, and death), but this data was the key to gauging the nets' overall effectiveness. Some scientists feared that mosquito bites were occurring in such large numbers in the study area that simply reducing the total might have no impact on the overall spread of malaria. However, the study proved otherwise.
As several researchers observed, the real surprise was that infection, sickness, and mortality also dropped in areas where no nets were used. Spatial analysis revealed that houses within 300 meters of a bed net village enjoyed protective benefits. This halo effect diminished at a distance of 300 to 600 meters and beyond. "We would not have been able to detect this community effect without information on the exact location of each household," said ter Kuile.
Approximately 22 percent of households without bed nets benefited from this community effect. Traditional estimates of bed net effectiveness compare the health of people without bed nets to people with bed nets. Because people without nets were benefiting from their neighbors' nets, bed nets were more effective than researchers thought.
Clues to why the community effect occurs are emerging from analysis of mosquito densities recorded within the study area. "The nets profoundly reduced the population of mosquitoes in and around the villages that had them," said Hawley. "We believe that people sleeping under nets attract mosquitoes to the permethrin-coated net surface, which then kills, or at least sickens, the mosquitoes so they never have the chance to travel elsewhere to find a meal."
This project continues, but the focus has shifted to determine the ideal pattern of distribution for insecticide-treated bed nets so the limited supply can be used to benefit the greatest number of people. The most important result of this study was proof that the nets were effective. Millions of children in Africa and around the world are now sleeping more safely beneath bed nets donated by international aid organizations.
About the Author
Allen Hightower is chief of data management for the Division of Parasitic Diseases, which is part of the National Center for Infectious Diseases at CDC in Atlanta, Georgia. He may be reached at firstname.lastname@example.org.
Note: Use of trade names and commercial sources is for identification only and does not imply endorsement by the U.S. Department of Health and Human Services.