gms | German Medical Science

Background: Adverse drug events (ADEs) occur frequently during hospitalisation. According to a systematic review and meta-analysis of in-hospital ADEs, 19% of inpatients suffer from an ADE and approximately one third of these ADEs are judged as preventable [1]. Given that ADEs vary in their relevance, depending on their frequency, severity, and preventability [2], [3], [4], a medication safety screening tool with focus on the most relevant ADEs could lead to a more targeted and efficient use of limited resources in both clinical practice and research. Additionally, combining ADEs that are identified as important with potentially causative drugs could restrict the detection of adverse drug reactions (ADRs) to those where a drug-related cause is (at least) possible.

The aim of this study was to develop drug-event pairs as valid indicators for the detection of clinically relevant ADRs during hospitalisation in routine data.

Materials and Methods: We conducted a two-stage consensus process based on the RAND/UCLA Appropriateness Method (RAM) [5], in order to prioritise in-hospital ADEs (stage one) and to identify drug-event pairs as valid indicators for the most important ADRs (stage two). Similar to the Delphi method, we developed an assessment form for each stage with candidate ADEs for stage one and with drug-event pairs as indicators for stage two based on systematic literature searches. Panelists independently rated each ADE/indicator on two occasions, with the first-round ratings fed back to them before the second-round ratings were placed. In contrast to the Delphi method but consistent with the RAM, two moderated face-to-face meetings (one for each stage) were held in between rounds. In the first stage, experts were asked to rate the “overall importance” of each ADE on a 4-point Likert scale (1= Not important to 4= Very important). ADEs with a median rating of ≥ 3 without disagreement were defined as „prioritised“. ADEs with a median importance rating of 4 were considered for the second stage, where the experts were asked to rate the strength of the causal link between ADEs and potentially causative drugs for an average patient on a 4-point Likert scale (1= Unlikely to 4= Certain). Drug-event pairs with a median rating of ≥ 3 without disagreement were defined as „valid indicators” of ADRs. In both stages, disagreement was pre-defined to be present if at least 30 % of expert ratings were ≤ 2 (for items with a median of ≥ 3), or ≥ 3 (for items with a median of < 3).

Results: In the first stage, the RAM panel comprised a total of 6 physicians and 6 pharmacists with clinical and/or scientific background from 9 university sites in Germany which prioritised 34 out of 63 ADEs according to their overall importance rating. The highest rated events were rhabdomyolysis, acute kidney injury, hypoglycaemia, liver damage and anaphylactic shock [6]. In the second stage, 5 physicians and 5 pharmacists from 9 university sites in Germany confirmed 44 out of 255 drug-event pairs as valid indicators of ADRs in routine data.

Conclusion: The first stage of our survey led to a set of prioritised ADEs and provided a basis for our second stage, resulting in systematically developed indicators of ADRs for application in clinical practice (e.g., decision support), clinical surveillance and research (e.g., as outcome measures). As this modified RAND consensus process is embedded in the Medical Informatics Initiative (MI-I) overarching use case POLAR (POLypharmacy, drug interActions, Risks) the developed indicators will be operationalised and implemented in routine data of university hospitals throughout Germany [7]. Ultimately, they will be used to determine the prevalence of potential ADRs during hospital stay and to develop automated risk models for the prediction of ADRs.