Hirut did the greater part of inserting the data elements. The number of data elements were past 1100, but a number of them were similar to the ones in Addis Ababa which Woinshet, the head of the HISP project in Ethiopia, had already typed into DHIS. We based our work on what Woinshet had done for Addis Ababa. The data elements and reports that the team from the bureau decided to be included in DHIS were different in a number of ways. The reports that were supposed to be sent to the federal authorities were the same, but most of the reports used internally in Tigray were different from those in Addis Ababa.
The months before my arrival Woinshet put a great effort into adapting DHIS to the requirements of Addis Ababa. Many of these requirements were the same in other regions we worked. The most important of these requirements was that they wanted the system to handle data based on The International Classification of Deceases (ICD). ICD is basically a number assigned to deceases according to a system of classification. In Ethiopia this codes were used for in-patients and out-patients to specify which diagnosis the patients got. In addition the gender and the age group the patient belonged to was recorded. Deaths were recorded in the same manner. If this kind of data were supposed to be included as normal data elements in DHIS each combination of ICD codes, gender, age group and whether it was in-patients, out-patients or death should have it own data element the number of data elements would pass 16,000 in Tigray.
DHIS is designed with a minimal data set in mind. Preferably below 100 data elements. In Tigray they wanted close to 1,200 data elements in addition to the ICD data. In Addis Ababa they wanted a similar number of data elements. Because of this Woinshet made a extension to DHIS to handle ICD based data. This extension was hard coded into DHIS version 18.104.22.168, therefore this was the version we used in Ethiopia. This extension being hard coded meant that it would have to be reimplemented for each new version of DHIS. It would probably be possible to make something similar to a patch that could easily be patched into new versions of DHIS. The fact that DHIS is made in MS Access and don’t have source code in the traditional sense makes this more difficult. MS Access is a RAD tool and stores its data and code in a binary format.
While Hirut inserted the organisational structure of Tigray and most of the data elements, I inserted the quarterly data elements and started on the reports. The reports given to us did not fit well to be implemented in a computer based system. To me it seamed like the bureau team had only gathered as many reports they could find and given it to me. They had not given thought to what data elements they wanted and how this data could be presented in reports. The list of data elements they had given us where also lacking data element necessary for a number of reports.
The reports they had given us where not made with computers in mind. Some reports required regular data to be given in a comment/note field. Some of the reports where weekly based, mostly from DPC, which DHIS 1.3 don’t support. What made things even more difficult where the fact that a number of reports where in Amharic or Tigrinian. The other people at my team understood Amharic, but none of us understood Tigrinian. Because of this challenges I was only able to complete the the quarterly reports, that is the reports on Tuberculosis and Leprosy.
Because I was unable to finish the reports with the information I had, I decided to postpone this task until I could talk to the departments at the bureau. To make use of the time I had left in Addis Ababa, before we were to go to Mekelle, I started on a small Java application. This application automated time consuming tasks like ordering the data elements and inserting test data. The test data were needed in order to test the reports. This application I have described in an other paper.
During our training at the bureau the following changes were proposed:
I implemented all this request except number 2 and 3. I didn’t find any easy way to change the month names in MS Access into Amharic names. The input clerks doing the typing could speak sufficient English to understand this. English is used as a administrative language alongside Amharic in Ethiopia. Requirement 3 i dropped because it would require more than changing the user interface. Months and years are coupled in the data model of DHIS. I could possibly had conceived a way to do this without changing anything, but the user interface. I didn’t deem it that important, however.