1998-2000

By Dr. Scott Anderson, Col. Kalyanee Torugsa, Col. Sorachai Nitayapan, Lt. Col. Arthur Brown, and Maj. Gen. Suebpong Sangkharomya - 2002

Source of Article - http://www.esri.com/news/arcuser/0702/thaihiv2of2.html

Although Thailand has experienced one of the world's most threatening HIV/AIDS epidemics, it has also conducted a most successful public health AIDS prevention campaign. Recently, GIS has become an important tool in the country's fight against the disease. The adoption of GIS as part of these prevention efforts provides an interesting case study in technology transfer across the globe made possible through personal relationships.

The AIDS Epidemic in Thailand

The first reported case of AIDS in Thailand occurred in August 1984. The HIV epidemic in the country started in the late 1980s and initially involved high risk groups--intravenous drug users and commercial sex workers and their customers. The virus then spread to the friends and families of members of these groups.

By 1992, the epidemic had spread so rapidly that 12.5 percent of the young men being inducted into the Royal Thai Army (RTA) from the most heavily impacted part of the country, the Upper North, tested positive for HIV-1. The following year, 3.7 percent of all young men screened before induction and 5.1 percent of men above the age of 21 tested positive. This population of young Thai men, chosen by national lottery each May and November, has become an important indicator not only of the overall prevalence of HIV-1 in the country, but also of the impact of the national AIDS prevention campaign.

Beginning in November 1991, the RTA began testing all army personnel for virus antibodies before induction. These young men were also given counseling on AIDS risk factors. About the same time, government officials initiated a number of strategies to combat the disease including widespread dissemination of information on AIDS prevention and the use of condoms. The government distributed free condoms, closed many commercial sex establishments, provided counseling to commercial sex workers, and required that sex workers use condoms.

The results were dramatic. By 1999, just six years after the growth peak of the epidemic, overall HIV-1 prevalence rates among young men tested prior to induction declined by two-thirds to just 1.3 percent. Among 21-year-olds, prevalence rates declined to just 1.1 percent. In the heavily impacted Upper North region, rates declined from 12.5 percent to just 1.6 percent, representing a reduction of 81.6 percent. In only one region, the South, did the epidemic continue to grow. The prevalence rates in this area increased from 2.0 percent to 2.5 percent over the same period.

The RTA and United States Army medical personnel, working jointly under the auspices of the Armed Forces Research Institute of Medical Sciences (AFRIMS), noted that there were broad regional variations in the occurrence of disease and recognized the need to use GIS methodologies to analyze the epidemic. As a result, in January 2001, officials from the RTA and the Research Foundation of the State University of New York (SUNY) signed a memorandum of agreement that initiated a fruitful collaboration between SUNY faculty and AFRIMS staff.

Personal Connections Lead to Technology Transfer

The principal players in this collaboration are Kalyanee Torugsa, an RTA colonel and veterinarian; Dr. Scott W. Anderson, a geography professor at the SUNY College at Cortland; and Lt. Col. Arthur E. Brown, a physician with the U.S. Army Medical Corps.

1996-1998

It was Brown who brought Torugsa and Anderson together from opposite sides of the globe. Brown has spent much of his professional life in Thailand, first as a civilian conducting medical research on leprosy and malaria, and later as a medical officer attached to AFRIMS. His current assignment is coordinating a preventive AIDS vaccine trial among more than 15,000 residents in two Thai provinces. He and Anderson have been friends since their undergraduate days at Dartmouth College. A casual conversation between the two about their current professional interests during Brown's stateside leave in the summer of 2000 became the impetus for the collaboration.

Torugsa is responsible for the AFRIMS unit that collects and analyzes data on all RTA inductees. She and her superior officers had been looking for a way to use GIS to improve the analysis and presentation of the immense amount of information collected on the epidemic. Implementing GIS proved to be a challenge because GIS professionals are still rare in Thailand. Torugsa needed advice on choosing hardware and software, collecting and managing data for future GIS purposes, and--especially--acquiring GIS skills. Working with Anderson provided a straightforward way to introduce an unfamiliar technology to her program, and her collaboration with Anderson was supported by the RTA chain of command.

Exchanging Data Across the Globe

Torugsa's staff was quickly able to provide Anderson with annual data on HIV-1 prevalence rates by province for young men prior to induction. The data, covering 1991 through 1999, was e-mailed to Anderson in Microsoft Excel files. This data included province-level information on the number of men tested, the number of those testing positive for HIV-1, and the percentage of men testing positive.

1994-1996

The GIS Lab at SUNY, Cortland, is primarily a teaching facility. Initially, Anderson was interested in the project primarily for its teaching applications. He soon had several advanced students performing the manipulations necessary to bring this data into an ArcView 3.2 project. His students converted the Excel tables into dBASE files, joined these files to shapefiles of the Thai provinces, and created and saved a common legend (.avl) file for each of nine annual maps. These maps were assembled in a single layout for printing.

The result was a very dramatic set of maps that demonstrated the rise and decline of prevalence rates over the study period. However, because the number of Thai provinces was limited (73 provinces in 1991 and 76 in 1999), the maps provided only a coarse representation of the geography of this epidemic. Furthermore, province-level maps using the same data had already been published in several medical journals.

Seeking a Finer-Grained Analysis

Anderson recognized that GIS could provide little new insight into the geography of the HIV/AIDS epidemic in Thailand unless the data could be made available for analysis at a finer level. Sharing this insight with Brown and Torugsa set in motion a chain of events that intensified and improved their collaboration. Torugsa set her staff to work revising data tables on each induction site for each of the 16 induction periods so district-level data could be mapped showing numbers of inductees tested, numbers testing positive for HIV-1, and percentage testing positive for the period from November 1991 through May 2000.

Fixing Flaws in the Data

Because there are more than 900 districts in Thailand, this was not an easy task. It took Torugsa's staff of six several months to complete the revisions. During the process, several flaws in the original project became apparent. The first problem related to the time periods being mapped. The May and November cohorts [i.e., groups of inductees] came from different regions of the country. Any map of annual data that did not include both cohorts was geographically skewed. Because the RTA began its preinduction testing in November 1991 and did not conduct tests in November of 1993 and May of 1994, there were only six annual periods with complete and reliable data.

Fortunately, the solution to this problem was fairly straightforward. Anderson collapsed the data across years so the November cohort of one year was combined with the May cohort of the next. This resulted in eight annual periods with reliable data. By joining these new data tables to a province shapefile obtained from the Thai Environmental Institute, Anderson was able to create a set of dramatic maps providing a more fine-grained portrayal of the geography of the AIDS epidemic in Thailand. Especially informative was a layout that contained eight choropleth maps showing changes in the percentage of inductees testing positive for HIV-1 by district over time. A dot density map that represented each positively testing inductee with a color-coded single dot was even more powerful.

Dealing with Insufficient ns

In August of 2001, Anderson flew to Bangkok to present these maps to AFRIMS personnel and conduct further consultations. At this time, Anderson, Brown, and Torugsa, along with other AFRIMS staff, agreed that another flaw in the project data needed to be addressed.

1991-1993

In many districts, so few young men were tested during a 12-month period that the percentage positive yielded outliers based on insufficient numbers (or ns). For example, one district might be assigned a zero incidence rate because only five men were tested and none tested positive, while another district might indicate a disturbingly high incidence rate because only five were tested, but one tested positive.

The collaborators agreed on a two-stage strategy to address this problem. First, they agreed to combine the data into four two-year periods for each district. Each period used data from four cohorts from May and November so the district ns would be higher. After consultation with an AFRIMS medical statistician, they also agreed to join contiguous districts together, when necessary, so that no district mapped during any of the four periods had fewer than 20 ns.

Back in Cortland, Anderson collapsed the data into time periods that encompassed four cohorts each. This involved little more than calculating data into new fields. However, joining contiguous districts required a more complex set of procedures. Anderson's first challenge was to develop a protocol for determining which districts to combine. The second challenge was to figure out how to do it.

A Protocol for Joining Districts

After consulting with AFRIMS staff via e-mail, Anderson developed a set of priorities for combining districts. The protocol he developed used the following logic in sequence to determine which adjacent districts to combine with low n districts.

1. Districts in the same province were given preference over all others.
2. Districts with historic connections were favored. For example, if two districts had previously been part of the same district, they were combined.
3. Districts with similarly low ns were joined rather than joining low n districts with very high n districts. In effect, this meant rural districts were combined rather than combining rural districts with urban ones.
4. Districts with similar demographics, especially districts with similar numbers of males aged 18-50, were preferred although ns could serve as a proxy for this variable.
5. Districts with similar shapes, topography, or other geographic features were combined. For example, contiguous districts along the Myanmar, Cambodia, or Laos borders were joined together.

Coding and Dissolving Districts Together

Once the districts that would be joined were identified, the actual process was simple. Anderson used the dissolve function supplied by the Geoprocessing extension for ArcView 3.x. After adding a new field containing code numbers for all Thai districts, he changed the code number for each low n district to the code number of the district to which it would be joined. He dissolved these districts together, and the resulting shapefile contained no low n districts. After this process was completed for each of the four time periods, Anderson produced a set of maps that showed the most accurate and fine-grained detail available to date on the geography of the HIV/AIDS epidemic in Thailand.

An Ongoing Process

The SUNY Cortland/AFRIMS collaboration continues. Torugsa has assigned an AFRIMS staff member, Major Panpaka Supakalin, the task of bringing GIS methodologies in-house. Her unit has acquired ArcView 8.1. Brown, Torugsa, and Anderson continue discussing ways to improve their work, and Anderson is now redrawing the Thai district-level shapefiles to correct a number of inaccuracies. This collaboration, based on a personal relationship, has resulted in a permanent transfer of technology and has ensured that GIS methodologies will play an increasingly important role in the fight against one of humankind's most feared epidemics.

For more information on this project, contact

Scott Anderson, Geography Department/GIS Lab
SUNY College at Cortland
Cortland, New York
E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it