Adverse drug reactions (ADRs) represent a significant proportion of medical complications, burdening both the patients and the entire healthcare system. The consequences of ADRs are diverse – ranging from mild, transient problems that may affect treatment adherence and quality of life, to more serious acute conditions requiring urgent medical intervention. Electronic health records play an increasingly important role in the identification of ADRs, as they enable rapid review of large volumes of patient data.
The aim of this master’s thesis was to define and evaluate a set of 18 specific keywords related to specific ADRs, which could be applied in daily clinical practice to identify ADRs from electronic patient records and to enhance the hospital pharmacovigilance system.
The study included 1064 patients aged 18 years or older, admitted through the Internal Medicine Emergency Department to the departments of the Internal Clinic, University Medical Centre Ljubljana, during January and February 2023. A retrospective manual review of their electronic records (medical histories and discharge letters) identified at least one ADRs in 416/1064 patients (39,1%). In total, 737 ADRs were detected by manual review, most frequently at admission (209; 28,4%). During hospitalization, 203 ADRs (27,5%) were recorded, 197 (26,7%) were documented as known allergies, and 128 (17,4%) were listed in the medical history (past ADRs). In the second part of the study, using a search engine with predefined specific keywords in the hospital information system, we identified 171 ADRs, representing 23,2% of all ADRs found by manual review, or 64,5% (171/265) of those estimated to be detectable with the selected set of specific keywords. Seven combinations of specific keywords were identified as most suitable for clinical use, achieving the highest positive predictive values: »iztirjen« + »marevan« (75,0%), »hipokal« + »edemid« (62,7%), »hipoglikemij« + »inzulin« (55,6%), »melen« + »marevan« (53,8%), »bradikardij« + »concor« (44,4%), »hipokal« + »diuver« (42,9%), and »krvavit« + »marevan« (40,0%). The majority of specific keywords (12/18) demonstrated a sensitivity above 40%. The highest sensitivities were observed for »hematemez« (90,0%), »hipokal« + »edemid« (82,1%), »hipotenzi« (73,0%), »izpuščaj« (70,0%), »iztirjen« + »marevan« (65,2%), and »hiperkal« + »aldacton« (54,5%). Specificity was high, with 16/18 keywords exceeding 90%.
|