Abstracts

Rationale: As electronic medical record (EMR) systems become more available, they will serve as an important resource for collecting epidemiologic data in epilepsy research. However, since clinicians do not have a systematic method for coding non-epileptic seizures (NES), patients with NES are often misclassified as epilepsy, leading to sampling error. This study validates a Natural Language Processing (NLP) tool that excludes linguistic information to help identify patients with NES. Methods: Using the VA national clinical database, 2200 notes of Iraq and Afghanistan Veterans who completed video electroencephalograph (VEEG) monitoring were reviewed manually and identified as having documented NES or not. Reviewers identified NES-related vocabulary to inform a NLP tool called Yale cTakes Extension (YTEX). Using NLP techniques, YTEX annotates syntactic constructs, named entities, and their negation context in the EMR. These annotations are passed to a classifier to detect NES patients. The classifier was evaluated by calculating positive predictive values (PPV), sensitivity and F-score. Results: Of the 742 Iraq and Afghanistan Veterans who received a diagnosis of epilepsy or seizure disorder and had VEEG documented events; 22 Veterans had definite NES and 33 had probable NES, and 2 had both NES and epilepsy documented. Our classifier achieved a PPV of 94%, a sensitivity of 99%, and a F-score of 96%. Conclusions: Our study validates a NLP tool used to identify NES patients, who have completed VEEG monitoring, in an EMR. The classifier tool can be critical in reducing false positive diagnoses of epilepsy and seizure disorders based on ICD-9 diagnosis of seizure disorder and/or epilepsy but have been diagnosed with NES. Therefore, the classifier may be valuable in conducting future EMR based epidemiologic research in epilepsy and seizure disorders.