Health care use and costs of adverse drug events emerging from outpatient treatment in Germany: a modelling approach

Abstract

Background: This study's aim was to develop a first quantification of the frequency and costs of adverse drug events (ADEs) originating in ambulatory medical practice in Germany.

Methods: The frequencies and costs of ADEs were quantified for a base case, building on an existing cost-of-illness model for ADEs. The model originates from the U.S. health care system, its structure of treatment probabilities linked to ADEs was transferred to Germany. Sensitivity analyses based on values determined from a literature review were used to test the postulated results.

Results: For Germany, the base case postulated that about 2 million adults ingesting medications have will have an ADE in 2007. Health care costs related to ADEs in this base case totalled 816 million Euros, mean costs per case were 381 Euros. About 58% of costs resulted from hospitalisations, 11% from emergency department visits and 21% from long-term care. Base case estimates of frequency and costs of ADEs were lower than all estimates of the sensitivity analyses.

Discussion: The postulated frequency and costs of ADEs illustrate the possible size of the health problems and economic burden related to ADEs in Germany. The validity of the U.S. treatment structure used remains to be determined for Germany. The sensitivity analysis used assumptions from different studies and thus further quantified the information gap in Germany regarding ADEs.

Conclusions: This study found costs of ADEs in the ambulatory setting in Germany to be significant. Due to data scarcity, results are only a rough indication.

Figure 1

Probabilities of morbidity and mortality associated with medication intake used in the modelling approach. Enclosed in the oval is the part of the model used for the scenario. Probability of drug prescription in the original model was 0.55, and no drug was 0.45. Overall outcome probabilities obtained by multiplying probability of type of drug-related problem with probability of outcome of drug-related problem (probabilities originate from the model published by Ernst et al., 2001 [17]).