gms | German Medical Science

Introduction: Adverse drug events (ADEs) are a major patient safety issue. Approximately 5% of all hospital admissions are associated with ADEs [1]. Generally older patients taking several drugs have a manifold higher risk of ADEs than younger patients [2].

The aim of the project was to develop a tool to enable physicians to explore ADEs by means of diagnoses and prescribed drugs in Austrian claims data.

Methods: The anonymised data provided by the Main Association of Austrian Social Security Institutions, includes hospital discharge diagnoses of 72 million Austrian hospitalisations from the years 2001 to 2011. The ADE related diagnoses have been identified and classified using the 505 ADE-Diagnoses (segmented into 7 categories) defined by Stausberg for Germany [3]. They have been adapted, accordingly to the documentation habits in Austria resulting in 458 ICD-10 diagnoses. In combination with the ADE diagnoses, 16 identified comorbidities (e.g. diabetes, renal diseases) corresponding to 213 ICD-10 codes, can be analysed. In addition prescribed medications were available for the years 2006 and 2007. For this two-year period potential drug-drug interactions were estimated by integrating the Austrian interaction database Austria Codex.

Geospatial information of Austrian care regions has been inserted in the database schema. The care region of the hospital and of the residential district of the patient were assigned to each hospitalisation. The population of each care region grouped by age and gender was calculated using data from Statistics Austria. An application for geospatial analysis was developed with Postgres SQL, R, Java and “Shiny”, a web application framework for R.

Results: Using the developed tool “JADE regio” the rates of hospitalisations with ADEs for different time intervals can be visualised via a map of Austria. The developed tool provides an overview of hospital stays with ADE associated diagnoses and the preceding medication and potential interactions. Selected statistical methods like logistic regression model and the number needed to harm have been implemented. Analyses can be stratified by gender, age groups, geospatial parameters and comorbidities.

The amount of diagnoses indicating an ADE (over all ADE categories) relative to the total number of diagnoses documented during hospitalisations in the years 2001 to 2011 vary from 4.5 % to 5.5 %.

Discussion: The developed application “JADE regio” can support physicians to explore ADEs in Austrian claims data to generate hypotheses. Physicians from different medical areas have already evaluated the tool. Further development will include additional patient data, a standardised data model like OMOP [4] and enhanced visualization methods.

We have not distinguished between primary and secondary diagnosis. To compare detailed ADE rates with others it is necessary to consider the varying documentation and reimbursement rules. Notwithstanding the limitations (e.g. no over the counter drugs and no prescribed dose and time of intake are included in the database) the exploration of health claims data can still complement clinical studies with population-based analyses in the area of drug safety.

This work was supported by G Endel from the Main Association of Austrian Social Security Organisations.

Die Autoren geben an, dass kein Interessenkonflikt besteht.

Die Autoren geben an, dass ein positives Ethikvotum vorliegt.