Signal detection in drug safety research traditionally relies on the use of spontaneous reporting (SR) data. Recently, the interest in electronic health care data for signal detection purposes grows, and data mining techniques developed for SR data are now being modified and used to highlight possible safety risks in such longitudinal data. We present simple frequentistic methods such as the reporting odds ratio (ROR) and the proportional reporting ratio (PRR), the widely-used Bayesian Gamma-Poisson shrinkage algorithm (GPS) and extensions for the use on longitudinal healthcare data. In an exemplary application to German claims data, we aim to recapture known signals for intracerebral hemorrhage under exposure to phenprocoumon and to compare the results obtained by the longitudinal Bayesian shrinkage algorithm with risk estimates from classical pharmacoepidemiological studies. We also highlight the still unsolved problem of adequate controlling for confounding in automated database studies.
CITATION STYLE
Suling, M., Weber, R., & Pigeot, I. (2013). Data mining in pharmacoepidemiological databases. In Robustness and Complex Data Structures: Festschrift in Honour of Ursula Gather (pp. 351–364). Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-35494-6_21
Mendeley helps you to discover research relevant for your work.