[Validation of Methods to Identify Known Diabetes on the Basis of Health Registers]

Introduction: In 2003 the government introduced a national diabetes plan. One of the recommendations was to establish a national diabetes database targeted at monitoring the prevalence of diabetes and quality of diabetes care. The aim of this study is to validate a national algorithm for identification of known diabetes and compare the results with the results from the use of a regional algorithm.

Materials And Methods: Patients with diabetes residing in Aarhus County on 31 December 2003 were identified by data from The National Patient Registry, The National Health Insurance Service Registry, the prescription database and the laboratory database in the county.

Results: This study identified a total of 8,802 patients with a diagnosis of diabetes which was confirmed by the patients' general practitioners (GP). This corresponds to a prevalence of 2.32% (95% CI: 2.27%2.37%). The national algorithm found 86% of this diabetes population while the regional algorithm found 96%. The sensitivity was increased to 91% by supplementing with information of dispensed prescriptions for anti-diabetics in the national algorithm. The positive predictive value was 89% for the national algorithm as well as for the regional algorithm.

Conclusion: The national algorithm may be used as a tool for establishing a national diabetes database. Despite a higher sensitivity, the regional algorithm cannot currently be recommended at a national level as it depends on the collection of person-related data which are not available nationally at the present time.