Using Natural Language Processing and Machine Learning to Identify Opioids in Electronic Health Record Data.

Area of Science:

• Medical Informatics
• Computational Linguistics
• Pharmacology

Background:

• Electronic health records (EHRs) contain vast amounts of unstructured data.
• Manual review of medication data in EHRs is time-consuming and prone to errors.
• Accurate classification of medications, such as opioids, is crucial for clinical research and patient care.

Purpose of the Study:

• To evaluate the effectiveness of machine learning (ML) and natural language processing (NLP) for classifying medication names in EHRs.
• To develop and assess an automated method for distinguishing between opioid and non-opioid medications.
• To determine the utility of ML/NLP in processing EHR data for retrospective analyses.

Main Methods:

• A dataset of 4216 distinct medication entries from EHRs was manually labeled as opioid or non-opioid.
• A bag-of-words NLP approach combined with supervised ML classification was implemented.
• The model was trained on 60% of the data and validated on the remaining 40%.

Main Results:

• The automated classification achieved high performance: 99.6% accuracy, 97.8% sensitivity, and 0.998 AUC.
• Out of 4216 entries, 225 were classified as opioids (5.3%) and 3991 as non-opioids (94.7%).
• Effective classification ( >90-95% accuracy, sensitivity, AUC) was achieved with approximately 15-20 opioid and 80-100 non-opioid training examples.

Conclusions:

• The ML/NLP approach demonstrates excellent performance in classifying opioid and non-opioid medications from EHR data.
• This automated method significantly reduces the need for manual chart review.
• The approach can be adapted for broader EHR data analysis and predictive modeling in pain research and other big data studies.