Extraction Of Adverse Events From Clinical Documents To Support Decision Making Using Semantic Preprocessing

Area of Science:

• Health Informatics
• Clinical Natural Language Processing
• Medical Data Mining

Background:

• Clinical documentation in electronic health records (EHR) is predominantly unstructured.
• Manual processing of this data is inefficient and time-consuming.
• Automated methods are needed to enhance information extraction from clinical notes.

Purpose of the Study:

• To develop and evaluate a rule-based method for identifying adverse events in clinical documentation.
• To transform unstructured clinical text into a semantic structure for adverse event extraction.
• To assess the performance of the proposed method compared to existing techniques.

Main Methods:

• A rule-based system was developed to parse clinical documents.
• Clinical text was converted into a semantic structure.
• Adverse events were extracted from the structured data.
• The method was evaluated by neurosurgeons, comparing results to a support vector machine (SVM) approach.

Main Results:

• The rule-based method achieved a recall of 65% and a precision of 78% in identifying adverse events.
• Performance was comparable to a bag-of-words classification using SVM.
• The structured output facilitates data reuse for various applications.

Conclusions:

• The rule-based method shows promise for supporting physician decision-making by extracting adverse events.
• Structuring unstructured clinical data enhances its utility and reusability.
• This approach offers an efficient way to process and analyze clinical documentation.