Integrating health-related data from various sources: combining surveys, records and routine data.

The paper discusses the necessity of combining data from various sources in order to enhance their usefulness for a variety of applications. As future data use can hardly be forseen in advance, major data sets in health services should fulfill several formal requirements in order to make them suitable for future linkage. These formal requirements are that there be references to defined populations, to specific persons, to defined time periods, to specific places or regions. It would be necessary for terms, definition and classification schemes to agree between data sets which are to be linked and be in wide use. Three facets of data linkage are discussed specifically namely linking data at one level of aggregation, linking different data components, and combining data sets from different sources at several levels of aggregation. Three examples are provided, describing linkages of data from various sources for epidemiological studies and a study in health services research. They show that at this point in descriptive epidemiological studies linkage on the basis of regions is of great importance. This implies that it would be desirable for large scale data collection activities in health services to provide for a uniform representation of the geographic areas. Such uniformity would greatly enhance the linkage potential of data sets and thus their usefulness for small area and regional analyses.