Predicting the progression to super-refractory status epilepticus: A machine-learning study

Abstract

AimSuper-refractory status epilepticus (SRSE) is a status epilepticus (SE) that continues or recurs ≥24 h after the onset of anesthesia. We aimed to identify the predictors of progression to SRSE and the risk of 30-day mortality in patients with SRSE by using a machine learning technique. MethodsWe reviewed consecutive SE episodes in patients aged ≥14 years at Baggiovara Civil Hospital (Modena, Italy) from 2013 to 2021. A classification and regression tree analysis was performed to develop a predictive model of progression to SRSE in SE patients. In SRSE patients, a multivariate analysis was conducted to identify predictors of 30-day mortality. ResultsWe included 705 patients, 16% of whom (113/705) progressed to SRSE. Acute symptomatic hypoxic etiology and age ≤ 68.5 years predicted the highest risk (87.1%) of progression to SRSE. Etiology other than acute symptomatic hypoxic and absence of NCSE predicted the lowest risk (3.6%) of progression to SRSE. The predictive model was accurate in 96.1% of patients not evolving to SRSE and in 48.7% of those evolving to SRSE. Among patients with SRSE, 46.9% (53/113) died within 30 days compared to 25.2% (149/592) of patients without SRSE (p < 0.001). Among patients with SRSE, older age was associated with increased 30-day mortality (odds ratio 1.075; 95% confidence interval: 1.031–1.112; p = 0.001). ConclusionsAcute symptomatic hypoxic etiology and younger age are major predictors of progression to SRSE. In patients with SRSE, older age is associated with increased risk of short-term mortality.