Adverse Drug Events (ADEs) tend to be registered as free text mentions in clinical narratives, such as clinical progress notes and discharge letters. Free text is considered as unstructured data and extensive pre-processing and formatting is needed for a computer to accurately analyse it. Natural Language Processing may help to address these challenges.

To gain insight in the extent and type of ADEs in the ICU setting, by leveraging unstructured data sources within the electronic health records of the ICU patients. To analyse the unstructured data sources, Natural Language Processing tools for Dutch clinical language will be developed and evaluated.

 We developed and validated a set of key words to identify ADEs in clinical notes of ICU patients. Using these key words we were able to identify much more ADEs in comparison to structured data sources in the electronic health records. Next step is to optimise our screening of clinical notes using Natural Language Processing.