[Administrative data as source for epidemiological research: clinical pathways of diabetic patients]

Objectives 1. To assess if record linkage of two different databases could improve the quality and understand the epidemiology of diabetes and its complications. 2. To analyse how hospitalization relates to the natural history of the disease and to its pharmaceutical management. 3. To document how pharmacoepidemiology could be a challenging tool for clinicians. 4. To identify critical areas where improvement of care would be specifically important.

Data came from two large databases, the drugs prescription and the hospital discharge forms of the Local Health Unit of Rovigo, Italy, collected through 2000. A casecontrol design was adopted to compare two cohorts identified by prescription of antidiabetic vs any other drugs and the linkage to their hospitalizations over the index period. The study was focused on people > or = 50 year in order to concentrate the attention on NIDDM. A population of 5.603 patients were identified as diabetic and 63.155 were the controls. The prevalence of diabetes was 3.6% in the general population and 8.1% in 50 year and older. The hospitalizations analysis revealed differences between cases and controls in term of longer duration of stay (10.1 vs 8.4 days), higher in-hospital mortality (5.5% vs 5%) and higher presence of cardiovascular complications. Of the 2.922 hospitalizations registered for diabetics, 43% did not report the specific ICD-9-CM code for diabetic disease (250.x). Record linkage of these administrative databases offers new opportunities to improve the comprehension of the natural history of diabetic disease. The identification of diabetic patients from prescription data allows a more reliable picture of the hospitalization than the simple analyses of hospital discharge forms. Up to 43% of hospitalizations in the diabetic cohort did not report the specific diabetic disease diagnosis, compared to the 25% reported in a previous observation where linkage was not applied. The lack of this code registration during hospitalization can reflect a scant perception of the importance of diabetes as the major determinant of complications and could open a discussion table with other clinicians, general practitioners and health care professionals.