engleski

Development of disease registration systems at population level in the EU

Issue: The increasing capacity of the hardwares has allowed during the last 25 years the storage of huge amounts of health related information. Large computerised data bases are now available in most European countries in hospitals and other public health institutions. The most common archives concern hospital admission and discharges, outpatients, pathology records, death certificates, drug consumptions, and, more recently, laboratory and instrumental exams. The most common use of this information is for administration purpose, while little effort has been so far carried out aimed at an epidemiological use of these data. Aims: This proposal concerns the potential use of routinely computerised health data for improving the knowledge on the geographical and temporal trends of disease at population levels, by implementing an automated process of disease registration. Description of the project: The only experience in this field is the so for the Automated Cancer Registration (ACR). This methodology, first applied to the Venetian Cancer Registry for the period 1987-1989, consists of an algorithm which summarises the existing information from three computerised data sources, mainly hospital discharges, pathology records, and death certificates, and, through a simple decisional process, assigns an ICD code to the various combinations of diagnoses. The proportion of cases registered through the automated process was about 2/3 of the total number of incident cases. This methodology has subsequently been applied to other few registries in Europe, and the results tend to confirm the replicability of the methods. Lesson learned: It is hypothesised that intelligent procedures leading to automated registration of diseases are characterised by higher sensitivity and lower specificity as compared to more traditional systems of registration. It therefore crucial to parallel the automated process with quality control procedures in charge of detecting and sizing the proportion of errors produced by the system, which should mainly concern the risk of false positives. This check should regularly be carried out through appropriate sampling techniques with the aim of improving the algorithm and measuring the error affecting the automated system. Results so far available indicate that automated processes are not affected by large systematic errors.

nije evidentirano

nije evidentirano