A Predictive Model for Short-Term Mortality After Status Epilepticus: The ACARD Score

Background and objectives: Reliable prediction of short-term mortality in status epilepticus (SE) can contribute to guide clinical decisions. Current prognostic systems achieve only acceptable predictive power and show lack of generalizability or poor calibration. We aimed to identify clinical predictors of short-term mortality in patients with nonhypoxic SE and develop a predictive score. Methods: This was a multicenter, multinational cohort study based on registry data. Participants were consecutive episodes of SE in participants aged 14 years or older from Modena (Italy) (derivation cohort) and in participants aged 18 years or older from Salzburg (Austria) (validation cohort). The predefined outcome was 30-day mortality after the onset of SE. Age, sex, level of consciousness before treatment, semiology of SE, level of disability at baseline before SE, etiology, and treatment refractoriness were assessed. Adjusted regression coefficients of each independent predictor were transformed to produce a points-based risk scoring system. Results: The Italian cohort included 689 episodes, and the Austrian cohort comprised 569 episodes of SE. In the derivation cohort, the 30-day mortality rate was 27.3%. The independent risk factors were aged 75 years or older (odds ratio [OR] 5.52, 95% CI 3.45-8.83; p < 0.001), consciousness impairment (stuporous or comatose) before SE treatment (OR 1.76, 95% CI 1.09-2.82; p = 0.020), acute etiology due to primary CNS pathology (OR 2.60, 95% CI 1.60-4.23; p < 0.001), refractoriness to treatment (OR 6.40, 95% CI 3.91-10.46; p < 0.001), and disability before SE onset (OR 3.53, 95% CI 2.22-5.61; p < 0.001); remote etiology was independently associated with a lower likelihood of 30-day mortality (OR 0.28, 95% 0.13-0.61; p = 0.001). An integer-based scoring system termed Age, Consciousness, Aetiology, Refractoriness, Disability (ACARD) was developed by combining these independent predictors. In the validation cohort, the 30-day mortality rate was 11.6%. The area under the curve of the ACARD score was 0.864 (95% CI 0.836-0.891) in the derivation cohort and 0.845 (95% CI 0.801-0.888) in the validation cohort. Calibration plot indicated good fit of predicted and observed data in both cohorts. Discussion: The ACARD score is a user-friendly tool developed to predict 30-day mortality after nonhypoxic SE. It has the potential to identify participants at high risk of short-term mortality and outperform the performance of other available scoring systems.