Untangling operational failures of the Status Epilepticus Severity Score (STESS).

Abstract

To uncover clinical characteristics leading to false outcome prediction of the Status Epilepticus Severity Score (STESS), a validated and broadly used clinical scoring system for outcome prediction in status epilepticus (SE). From 2005 to 2016, adult patients with SE treated at the University Hospital Basel, Switzerland, were included. To assess independent associations of variables differing between patients with false and correct prediction of death (STESS ≥ 3), multivariable logistic regression models were computed using automated selection. Among 467 patients, 12% died. The median STESS was 3 (interquartile range 2-4). Regarding prediction of death, the STESS was false-positive in 51% and false-negative in 1%. Patients surviving despite having a STESS ≥3 had less fatal etiologies, less nonconvulsive SE with coma, and lower Charlson Comorbidity Index, Simplified Acute Physiology Score II, and Acute Physiology and Chronic Health Evaluation II scores. In multivariable analyses, odds for survival were high with SE types other than nonconvulsive status with coma and low with an increasing Charlson Comorbidity Index in patients with a STESS ≥ 3 (odds ratio [OR]for survival 4.23, 95% confidence interval [CI] 2.33-9.60; and ORfor survival 0.86, 95% CI 0.75-0.98). In patients with SE types other than nonconvulsive with coma, the STESS was mainly increased because they were frequently older than 65 years and had no seizure history. The STESS frequently and inadequately predicts death especially in patients with SE other than nonconvulsive with coma and few comorbidities. Clinicians are urged to interpret a STESS ≥3 with caution in such patients.