Dictionary construction and identification of possible adverse drug events in Danish clinical narrative text

Abstract

Objective: Drugs have tremendous potential to cure and relieve disease, but the risk of unintended effects is always present. Healthcare providers increasingly record data in electronic patient records (EPRs), in which we aim to identify possible adverse events (AEs) and, specifically, possible adverse drug events (ADEs).

Materials and methods: Based on the undesirable effects section from the summary of product characteristics (SPC) of 7446 drugs, we have built a Danish ADE dictionary. Starting from this dictionary we have developed a pipeline for identifying possible ADEs in unstructured clinical narrative text. We use a named entity recognition (NER) tagger to identify dictionary matches in the text and post-coordination rules to construct ADE compound terms. Finally, we apply post-processing rules and filters to handle, for example, negations and sentences about subjects other than the patient. Moreover, this method allows synonyms to be identified and anatomical location descriptions can be merged to allow appropriate grouping of effects in the same location.

Results: The method identified 1 970 731 (35 477 unique) possible ADEs in a large corpus of 6011 psychiatric hospital patient records. Validation was performed through manual inspection of possible ADEs, resulting in precision of 89% and recall of 75%.

Discussion: The presented dictionary-building method could be used to construct other ADE dictionaries. The complication of compound words in Germanic languages was addressed. Additionally, the synonym and anatomical location collapse improve the method.

Conclusions: The developed dictionary and method can be used to identify possible ADEs in Danish clinical narratives.

Keywords: Adverse Drug Event; Adverse Drug Reaction Reporting Systems; Data Mining; Dictionary; Electronic Health Records.

Figure 1

Method flowchart. (A) Adverse event descriptions were extracted from the summaries of product characteristics (SPCs), representing the baseline dictionary. (B) The lexemes in the baseline dictionary were assigned into seven dictionary groups, split in two groups according to whether they are involved in post-coordination or not. The first number below each group indicates the number of concepts and the second number the unique identifiers in the group. (C) Tagging using the baseline dictionary. (D) Tagging using the group dictionary and subsequent post-coordination of tagged lexemes. (E) The four filtering groups split according to whether they disqualify sentence subparts or the whole sentence in the filtering step. The number below each group indicates the number of concepts. (F) Filtering, where any disqualified possible ADE is removed. (G) The final output of the pipeline.

Figure 2

Synonymous coordinated terms. Lexemes from the groups localized event and laboratory event are combined with location and laboratory values, respectively, to produce coordinated terms. The method identifies the equality of two different ways of writing the same ADE and synonymous coordinated terms are merged to one common term. Order and possible prepositions used during post-coordination were excluded.

Figure 3

Location collapse. Locations are collapsed and merged into a single identifier.

Figure 4

Dictionary group synonyms and synonymous coordinated terms. Synonyms, inflections and spelling variants were merged into a common concept, where ordering and prepositions were omitted.