Organizational aspects and implementation of data systems in large-scale epidemiological studies in less developed countries

Abstract

Background: In the conduct of epidemiological studies in less developed countries, while great emphasis is placed on study design, data collection, and analysis, often little attention is paid to data management. As a consequence, investigators working in these countries frequently face challenges in cleaning, analyzing and interpreting data. In most research settings, the data management team is formed with temporary and unskilled persons. A proper working environment and training or guidance in constructing a reliable database is rarely available. There is little information available that describes data management problems and solutions to those problems. Usually a line or two can be obtained in the methods section of research papers stating that the data are doubly-entered and that outliers and inconsistencies were removed from the data. Such information provides little assurance that the data are reliable. There are several issues in data management that if not properly practiced may create an unreliable database, and outcomes of this database will be spurious.

Results: We have outlined the data management practices for epidemiological studies that we have modeled for our research sites in seven Asian countries and one African country.

Conclusion: Information from this model data management structure may help others construct reliable databases for large-scale epidemiological studies in less developed countries.

Figure 1

The model data management office. It shows all the data staff, necessary equipment, tables, file storage, should be connected.

Figure 2

The data management team of the Disease of the Most Impoverished (DOMI) projects. The team included systems and operation units. The responsibilities of the team members have been described in the text.

Figure 3

The schema of database for the DOMI typhoid vaccine trial programs. The data table names are shown inside brackets next to the form name. The linkage keys are shown next to the box of the data table. The logical relationships between the entities are shown in parenthesis. In a relationship (1,N), "1" indicates each entity of the table will link to at least one entity of the other table, and "N" indicates multiple entities of the table may link to at least one entity of the other table. In a relationship of (0,N), "0" indicates not all entities of the table will link to another table. And, in a relationship of (1,1), the later "1" indicates a single entity will link to at least one entity of the other table.

Figure 4

The features of the batch processing system. It describes operation procedures for each functional subdivision of the batch data processing system.

Figure 5

The flow of data validation process. The data validation process is started with dual data entry by two different persons, resolving keypunching errors, and then identifying the data errors that contain duplicate entries, outliers, inconsistencies, and data linkage problems. The data errors usually solved through field verifications unless the mistake is done at the level of data entry.