Regional administrative health registries as a resource in clinical epidemiologyA study of options, strengths, limitations and data quality provided with examples of use.

The present thesis, which is based on a review and 12 published articles, concerns clinical epidemiological methods [176-187].The Nordic countries have for many years established numerous registries. The establishment in Denmark of the National Population Registry in 1924 and the personal registration number (the CPR number) in 1968 allowed person-identification of remarkable quality, and made it possible to gather information on the same person in several registries. This situation is unique to the Nordic countries.The administrative registries were not primarily established for research purposes but have often proved a valuable tool in research. Despite the extensive use of registries in research, the methodological literature on this subject is limited. The purpose of the present thesis was: 1) to analyse strengths and limitations in using regional administrative registries in research, 2) to develop a framework for evaluation of existing registries for use in clinical epidemiological research, 3) to develop methods for evaluation of the data quality in regional registries, and 4) to evaluate four regional Danish administrative health registries for use in clinical epidemiological research.The analyses of strengths, limitations and data quality were based on studies of data from the regional hospital information systems, health service registries, and public health officers' surveillance system for strong analgesics and notifiable diseases.Against the background of the studies, the many advantages of using registries in research are discussed. The most important advantage is that data already exist and time consumption is thus considerably reduced, compared with studies based on collection of primary data. Costs are also considerably reduced. Other advantages included the generally large sample sizes, which provide great precision in estimates and which allow the study of rare exposures, diseases and other effects. Typically, the registries are complete as far as the persons in the target population are concerned; this ensures representativeness. Collection of registry data has been done independently of the present study, and this often reduces various types of bias such as recall, non-response, and influence on the diagnostic process determined by the study. A number of health problems manifest themselves many years after the exposure, and existing registries are thus especially valuable when studying diseases with a long period between exposure and disease manifestation.The limitations in the use of registries are considerable and often ignored. The most significant problems are related to data selection and data quality because the data collection method is determined by the registry. The studies have varied in quality, which is often evaluated at regional level, since the same information is usually stored in several registries. This allows for comparison with respect to the completeness and validity of the data. The use of discharge diagnoses from hospital registries involves considerable data quality problems, while data from the health service registries, and from the public health officers' surveillance system for strong analgesics and notifiable diseases are of satisfactory quality. Misclassification of data exists, however, in all data sources; this can only be revealed by comprehensive studies of the validity of the data quality.