To identify clinical and treatment factors associated with refractory SE (RSE), in-hospital mortality, and resource use in an adult SE cohort.
Status epilepticus (SE) carries substantial morbidity and mortality in both classic population studies and contemporary regional cohorts.
Retrospective cohort of consecutive adults with chart-confirmed SE over one year. Variables included demographics, prior epilepsy, onset location, etiology, semiology, EEG findings, prehospital/ED benzodiazepines (BZDs), first antiseizure medication (ASM), GCS, LOS/ICU days, and discharge outcomes. Analyses: multivariable logistic regression for RSE; penalized logistic regression for mortality (to address separation with GCS/etiology); Gamma GLM for LOS.
We included 81 patients (52% female; median age 62 years). RSE occurred in 36 (44.4%), and in-hospital mortality was 22.2% (18). In adjusted analyses:
· A history of epilepsy was associated with reduced odds of RSE (aOR 0.15, p=0.022).
· Convulsive semiology (aOR 4.32, p=0.040) and EEG epileptiform abnormalities (aOR 3.91, p=0.046) were associated with increased odds of RSE.
· Compared to lorazepam, midazolam showed a trend toward reduced RSE risk (aOR 0.26, p=0.064).
· In the penalized mortality model, deaths were confined to acute etiologies (notably anoxic injury and cerebrovascular events) and low GCS (≤8).
· Myoclonic semiology was strongly associated with mortality (OR 25.0, 95% CI 5.7–110.5, p<0.001). In-hospital onset and acute etiologies were associated with mortality risk, while male sex, prior epilepsy, and convulsive semiology were linked to lower risk.
· Longer stays were associated with EEG epileptiform abnormalities (~65% increase, p=0.044) and older age (~1.7% per year, p=0.024).
In this cohort, independent determinants of RSE included prior epilepsy (protective), convulsive semiology, and EEG epileptiform abnormalities. Mortality was driven by acute etiologies (anoxia and cerebrovascular events), low GCS, and in-hospital onset. EEG abnormalities and age increased resource use. These findings complement historical and contemporary series and highlight targets for prognostication and protocol optimization.