Identification of Generalized Convulsive Status Epilepticus from Emergency Medical Service Records: A Validation Study of Diagnostic Coding

Abstract

Background Generalized convulsive status epilepticus (GCSE) is a neurologic emergency demanding prehospital identification and treatment. Evaluating real-world practice requires accurately identifying the target population; however, it is unclear whether emergency medical services (EMS) documentation accurately identifies patients with GCSE. Objective To evaluate the validity of EMS diagnostic impressions for GCSE. Methods This was an analysis of electronic medical records of a California county EMS system from 2013 to 2018. We identified all cases with a primary diagnostic impression of “seizure-active,” “seizure-post,” or “seizure-not otherwise specified (NOS)” and within each diagnostic category, we randomly selected 75 adult and 25 pediatric records. Two authors reviewed the provider narrative of these 300 charts to determine a clinical seizure diagnosis according to prespecified definitions. We calculated a kappa for interrater reliability of the clinical diagnosis. We then calculated the positive predictive value (PPV), sensitivity, and specificity of an EMS diagnosis of “seizure-active” diagnosis for identifying GCSE. Sensitivity and specificity calculations were weighted according to the distribution of seizure cases in the overall population. We performed a descriptive analysis of records with an incorrect EMS diagnosis of GCSE or seizure. Results Of 38,995 total records for seizure, there were 3401 (8.7%) seizure-active cases, 12,478 (32.0%) seizure-NOS cases, and 23,116 (59.4%) seizure-post cases. An EMS diagnosis of “seizure-active” had a PPV of 65.0% (95% CI 54.8–74.3), sensitivity of 54.6% (95% confidence interval [CI] 39.3–69.0), and specificity of 96.6% (95% CI 95.1–97.6) for capturing GCSE. Limiting the case definition to patients who received an EMS diagnosis of “seizure-active” and were treated with a benzodiazepine increased the PPV (80.2%; 95% CI 69.9–88.2) and specificity (99.3%; 95% CI 98.7–99.6) while the sensitivity decreased (25.1%; 95% CI 17.0–35.3). Across the 300 records reviewed, there were 19 (6.3%) patients who had a non-seizure related diagnosis including non-epileptic spells (7 records), altered mental status (8 records), tremors (2 records), anxiety (1 record), and stroke (1 record). Conclusions EMS diagnostic impressions have reasonable PPV and specificity but low sensitivity for GCSE. Improved coding algorithms and training will allow for improved benchmarking, quality improvement, and research about this neurologic emergency.