Drug-Event Pairs as Indicators for the Detection of Adverse Drug Reactions during Hospitalization in Routinely Collected Electronic Data Sources

Adverse drug reactions (ADRs) are a common cause of morbidity and mortality in hospitalized patients. Identification of ADRs in clinical practice, surveillance and research is essential to prevent further harm. The aim of this study was to assess the likelihood of drugs contributing to clinically important inpatient adverse events, in order to provide a list of drug-event pairs indicating ADRs in electronic health record (EHR) data, referred to as "indicators of ADRs". We conducted a consensus process based on the RAND/UCLA Appropriateness Method for 14 ADRs. Experts were asked to rate the strength of the causal link between adverse events and potentially causative drugs on a 4-point Likert scale. Based on the median rating, drug-event pairs were categorized according to the likelihood of an ADR being present. Drug-event pairs with a median rating of >= 3 without disagreement were defined as indicators of certain and probable ADRs. Of the 255 drug-event pairs evaluated, 2 (1%) and 42 (16%) achieved consensus validation that they certainly and probably indicate an ADR. In addition, 137 drug-event pairs were considered as indicators of possible (54%) and 74 drug-event pairs were considered as indicators of unlikely (29%) ADRs. The provided set of content-validated indicators of clinically important inpatient ADRs can be used in clinical practice (e.g., decision support), surveillance (e.g., quality indicators) and research (e.g., outcome measures). They will be implemented in EHR data from German university hospitals to determine the prevalence of ADRs, support efficient use of pharmacist resources, and develop models predicting ADRs.