Automatisierte Verfahren in der Arzneimittelsicherheitsforschung

Abstract

The evaluation of routinely collected data, such as claims data of statutory health insurances, has been pursued in drug safety research in the recent years as these data generally contain more information than spontaneous reporting registers. In this thesis it is examined how the analysis of claims data using automated signal detection methods can lead to reliable conclusions about the risk profile of drugs. The idea of disproportionality analysis is described. The theoretical background of the Gamma- Poisson Shrinkers and its extension for the analysis of longitudinal data is given as well as results of a practical application. To address the problem of the often missing consideration of confounders in signal detection, the propensity score and the High-dimensional propensity score (HDPS) are discussed and compared in an application study. Finally, the ideas of disproportionality analysis and automated covariate selection via HDPS are theoretically combined and practically applied to claims data. In conclusion, all investigated automated methods can be applied successfully to claims data, even if the latter approach is still in need of further adjustments.