Automated Identification of Aspirin-Exacerbated Respiratory Disease Using Natural Language Processing and Machine...

Area of Science:

• Computational medicine and bioinformatics.
• Clinical informatics and artificial intelligence in healthcare.

Background:

• Aspirin-Exacerbated Respiratory Disease (AERD) is an inflammatory condition often diagnosed late, delaying effective treatment.
• Current diagnostic delays for AERD average over 10 years, impacting patient access to therapies like aspirin desensitization.

Purpose of the Study:

• To develop an integrated algorithm using natural language processing (NLP) and machine learning (ML).
• To identify patients with AERD from electronic health records (EHR).

Main Methods:

• Developed a rule-based decision tree algorithm incorporating NLP-extracted features from clinical notes.
• Extracted 7 key features: AERD, asthma, NSAID allergy, nasal polyps, chronic sinusitis, elevated urine leukotriene E4, and no-NSAID allergy.
• Utilized MedTagger for feature extraction and optimized the decision tree classifier on a training set.

Main Results:

• The combined NLP and ML algorithm achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.86.
• The algorithm demonstrated 80.00% sensitivity and 88.00% specificity in identifying AERD patients on the test set.

Conclusions:

• A promising algorithm for AERD diagnosis has been developed, requiring further refinement.
• Advancements in NLP and ML technologies hold potential to significantly reduce diagnostic delays for AERD.