New publication on text-based approach to detect adverse drug events

SafePolyMed colleagues from University of Tartu published their results on using a text-based approach for the detection of adverse drug events (ADEs). Colleagues Dage Särg et al. used electronic health records of the participants of the Estonian Biobank and combined rule-based and machine learning approaches to make it easier to detect ADEs in health records. ADEs are usually recorded in free text fields in those records, such as the summary section, making it difficult to systematically retrieve information on ADEs. Focusing specifically on antidepressants and antipsychotics, the researchers found that their text-based filtering approach significantly reduced the effort in creating a data set of individuals who developed an ADE from a certain drug. Creating such data sets allows for the closer study of associations between pharmacogenetic phenotypes and ADEs. This paves the way for researchers to further improve ADE detection, for example by tailoring models to specific languages and clinical contexts. Read the full publication here.