Introduction

As you quickly learn in ABT 180, there are countless applications of GIS to many "real world" problems , which scientists are attempting to solve. We initially decided to focus our project on some aspect of disease outbreaks in Africa, where we knew that GIS could have a measurable impact in determining either where diseases were occurring, or what parts of the population were being most dramatically affected, or perhaps who was most in need of help. Historically there has been a problem in Africa where medical facilities, which are often located in the large metropolitan cities, were not actually best serving the populations who most desperately needed them. This situation can undoubtedly be attributed to Benito Mussolini's invasion of Ethiopia, Eritrea, and Somalia. Italy was the colonial power which took over Ethiopia, and when their efforts to build medical facilities began, they were focused primarily on serving the Italians living in Ethiopia as colonial civil servants. Naturally , these Italians lived in the major metropolitan cities, and so this is where the medical facilities were placed. The policies of the Italians clearly left many Ethiopians without access to medical facilities or health care. Since independence in 1941-42 , efforts to reform these colonial practices and policies were addressed in the Declaration of Socialism of 1974 and the National Democratic Revolution Programme Declaration of 1976. ( It should be noted here that our study includes modern Eritrea, which gained independence Ethiopia from 1993-95. ) "The approach to national health planning has changed ... stressing expansion and development of rural health services reflecting the primary health care approach." (Kloos/Zein, p.11)

Objective / Hypothesis

The objective of our project is to identify the locations of rural cities in Ethiopia where medical facilities are most needed. The results will be determined by the geographic location of endemic areas of major disease outbreaks of Malaria, Yellow Fever, Schistosomiasis, and Onchocerciasis in Ethiopia , in combination with the known locations of current medical facilities and the population densities of the endemic areas.

The proposed facilities are located in major cities (c. 1994) based on specific factors such as :

• geographic areas where all four diseases are prevalent (c. 1988)
• location of current existing medical facilities (c. 1988)
• population density and total population in these geographic areas which are served by the proposed medical facilities (c. 1984)
  
  
  

Data Sources

As we searched for data sources in this area through MELVYL and the Internet WWW, we discovered a very detailed study entitled The Ecology of Health and Disease in Ethiopia, published in 1988 by Prof. Helmut Kloos and Zein Ahmed Zein, which was the most up-to-date information we found for any particular country or region. The study contained maps which depicted endemic regions (i.e. polygons) of malaria, onchocerciasis, and yellow fever, as well as point locations of schistosomiasis incidence. It also contained a map of point locations of existing medical facilities , depicting both Hospitals and Health Centers. In addition the study contained a map of population density regions (i.e. polygons). We used the more recent map of existing cities from Harold Marcus' History of Ethiopia.

As we learned from ABT 180, a GIS project is largely dependent on the accuracy of the spatial data used to conduct the study. This is one area where our project is limited in that the maps from this study have no "metadata" (i.e. there is no indication of the projection or process used to compose the maps), thus we could not determine the relative accuracy of the maps.

Faced with this problem, we then needed to come up with some "geo-referenced" base maps which we could then register the maps from the study to. We decided to use the ARC World Digital Database coverages of the Ethiopian Political Boundary and Administrative Boundaries, available in the Visualization Lab, as our base maps.

Importing Spatial Data

All of the maps of disease outbreaks, the map of existing medical facilities, and the map of major cities first needed to be "imported" into a digital, vector format which the GIS application ARC/Info could use. There are a number of different ways to accomplish this task such as using the digitizing table, scanning and then using GRID to convert the scanned images to vector coverages, or what turned out to be our chosen method, "on-screen" digitizing.

Because the maps from the study were all relatively small in size (i.e. less than 8 ½ x 11), we decided that it would be easiest to scan them on a HP 4c flatbed scanner and "on-screen" digitize rather than using the digitizing table in order to achieve greater accuracy. We saved the scanned images as ".tif " images. Then we used the ".tif " images as background coverages in ARC Edit to create vector coverages from the raster ".tif " images.

To accomplish this we first used the IMAGE command to bring up the ".tif " in the background environment in ARC Edit. Next we used CREATE to create a new coverage. Then we digitized the attributes and features of our scanned raster ".tif " image into a vector coverage format.

For creating the polygon coverages we first used the INTERSECTARCS ALL command to ensure that nodes were placed at all intersecting arcs. We then set the EDITFEATURE to ARC and ADDed the arcs using the SPLINE function to smooth the lines we digitized from the ".tif " images. Finally, when all arcs were added , we used the CLEAN command to create arc and polygon topology.

For creating the point coverages we simply set the EDITFEATURE to POINT and used the ADD command to digitize all necessary point features from the ".tif " images. Then we used the BUILD command with the POINT option to create topology.

Editing and Processing Spatial Data

The next step was to "register" or assign real-world coordinates to our digitized coverages. This was accomplished by placing tics in the digitized coverages in locations which nodes existed in the ARC World coverages, which we already determined to have a Geographic projection in decimal degrees using the DESCRIBE command. We then created new tic coverages using the tics and boundary file from the ARC World coverage. We deleted the tics from our new tic coverages and placed new tics in the same locations as our digitized coverages. We brought up the ARC World coverages as BACKCOVERAGES and used them as the SNAPCOVERAGE with the SNAPFEATURE set to tic node. In this manner, we were able to place each tic in the most accurate position possible relative to its corresponding tic in the digitized coverage. We used nodes from both the political and administrative boundary coverages from ARC World to "geo-reference" our new tic coverages in decimal degrees. We are now left with tic coverages with "geo-referenced" tics at the same location as the tics in our digitized coverages and are ready to begin the TRANSFORM process. Using the TRANSFORM command, we assigned "real-world" coordinates, in decimal degrees, to our digitized coverages.

After the digitized coverages were converted to decimal degrees, we used ARC Edit to clean the digitized coverages and to ensure that they would overlay properly with the ARC World coverages. Because we knew that the ARC World coverage was more accurate than the digitized coverages, we used it's border in place of the digitized coverage border instead of RUBBERSHEETING.

Our next step was to project the images into the same projection of the base maps. In our study, our base maps were not in a specific projection so we chose a projection that would minimize distortions of both area and direction in proximity to the equator. For our project, we chose the Mercator projection.

Adding Attribute Information

Finally, once our coverages were "geo-referenced", transformed into decimal degrees, projected into the Mercator projection system, and then edited to clean up mismatching border arcs, we were ready to code the coverages with the appropriate attribute data necessary to perform the spatial analysis.

For the malaria, onchocerciasis, and yellow fever coverages, we simply had to code each polygon with a value designating it as either an "endemic" or "non-endemic" area. We used the ADDITEM command to add an integer item called "endemic" to the polygon attribute table, using 1 to indicate endemic areas while all the rest with zero would be the non-endemic areas. We used the SELECT MANY command to choose all the polygons which were endemic and the CALCULATE command to update the value to 1. We used the same process to add density ranges for the population density coverage.

For the Hospitals coverage, we needed to code each point to it's corresponding type of Medical Facility . Again we used the ADDITEM command to add an integer item to the point attribute table called "rank", using 1 to designate a Hospital and 2 to designate a Health Center. We used the same process to add "rankings" for the cities coverage and the schistosomiasis coverage.

When all of our "on-screen" digitizing, editing, and processing was completed, and the appropriate attribute data had been added, we had the following coverages:

Spatial Analysis

We decided that the placement of the medical facilities should be dependent on two major factors; population density and geographic areas of endemic disease or incidence of major disease outbreak. The strength of our spatial analysis lies in two areas. First was our ability to overlay polygons on polygons to determine areas where all diseases were endemic and points on polygons to determine where cities and hospitals lie relative to those intersecting endemic areas. Secondly we again overlaid polygons of 10 Km buffers around prospective cities on population density polygons in order to calculate the total potential population of people served by the proposed medical facilities

In order to isolate the areas in which all of these diseases were present, we used the INTERSECT command to retain only those areas of our disease coverages that were common to all the disease data sets being spatially joined. This was done for our three polygon disease coverages as well as for our point disease coverage. The end result was a clearly defined area where incidence of Schistosomiasis, endemic Malaria, Yellow Fever, and Onchocerciasis all intersected. By overlaying point coverages of major cities and existing medical facilities on this defined area, we were able to identify four major cities located in this area which were in the most need of new medical facilities according to our criteria.

Once these prospective cities were identified, population density data was used to calculate the approximate number of people that the proposed facility could theoretically service. Assuming that each facility could be utilized by individuals located less than 10 km away (Kloos/Zein, p. 52), buffers of this size were created around the prospective cities to simulate the service area of a medical facility placed in that city. Again the INTERSECT command was used to retain the population density of the area covered by the 10 km buffers. This was done for each of the four service buffers and each was saved as an individual theme. This was done so we could save unique attribute information for each proposed site to approximate the number of people who would be theoretically served by each proposed medical facility (See Table 1). Those sites with service areas encompassing denser populated areas were ranked higher than the sites that encompassed less populated areas. Using this ranking, we were able to identify which of these four cities was in most need of a new medical facility.

Results

The results of our analysis indicate that a Medical Facility needs to be placed immediately in Jima , secondly in Nekemte, then in Asosa, and finally in Metu. The facility in Jima will serve 51,416 +/- 10,158 people. The facility in Nekemte will serve 28,731 +/- 4,141. The facility in Asosa will serve 1,562 +/- 1,562. The facility in Metu will serve 9,557 +/- 2,643. Initially it would seem that Metu would be in line first for a facility before Asosa, however, there already is a health center in Metu, while Asosa, which lies in the heart of the endemic regions has no facility at all.

Conclusions

We have investigated an area in which GIS is being applied to achieve stunning results in the improvements to quality of life and access to health care. We wish that we had more time to add coverages of topography, rivers, and roads to our study to more accurately determine accessibility to our proposed facilities. We also would have liked to come up with an index of who is not being served by the facility, in addition to the total population who is served by each facility. We believe our methodology could have been strengthened by these additions.

Bibliography / Map Data Sources

ESRI Inc. ARC World (1:3m Digital Cartographic Data Set) . Redlands, CA : ESRI, Inc., 1992 .

Kloos, Helmut and Zein, Ahmed Zein, editors. The Ecology of Health and Disease in Ethiopia. Addis Ababa : Ministry of Health, 1988.

Marcus, Harold G. A History of Ethiopia. Berkeley / Los Angeles : UC Press, Ltd. , 1994.

Submitted By :

Jim Mullins &

Gregory McHugh

Final Laboratory Project for :

Applied Biological Systems & Technology 180 ,

Tues. / Thurs. , 2:10 - 5:00 Lab Section ,

WinterQtr.97, UC Davis.

Thanks to :

Prof. Wes Wallender, Lab Instructor Paul Grant,

& T.A.'s Alicia Palacios & Michael O'Neill