Determining the reasons for medication prescriptions in the EHR using knowledge and natural language processing.

Area of Science:

• Medical Informatics
• Clinical Natural Language Processing

Background:

• Accurate medication indication knowledge is crucial for patient safety systems like computerized physician order entry.
• Electronic Health Records (EHRs) contain prescribing information but often lack explicit reasons for medication use.
• Automating the extraction of medication indications from clinical notes is essential for enhancing EHR utility.

Purpose of the Study:

• To describe a novel method for determining medication use reasons from outpatient clinical notes.
• To leverage acquired drug-indication knowledge and natural language processing (NLP) for this task.
• To evaluate the performance of the developed method.

Main Methods:

• Utilized a combination of external drug-indication knowledge and NLP techniques.
• Applied the method to information extracted from outpatient clinical notes within the EHR.
• Performed an evaluation to quantify the method's accuracy and effectiveness.

Main Results:

• The method achieved a sensitivity of 62.8%, specificity of 93.9%, and precision of 90%.
• An F-measure of 73.9% was obtained, indicating a strong performance in identifying medication indications.
• This pilot study demonstrated the feasibility of linking external knowledge to EHR data.

Conclusions:

• Linking external drug indication knowledge to EHR data for determining medication reasons is a promising approach.
• The study identified challenges and areas for future improvement, including enhancing knowledge base accuracy and coverage.
• Future work will focus on expanding the drug knowledge base and validating performance on a larger outpatient drug dataset.