Machine learning validation through decision tree analysis of the Epidemiology-Based Mortality Score in Status...

This study was undertaken to validate the accuracy of the Epidemiology-Based Mortality Score in Status Epilepticus (EMSE) in predicting the risk of death at 30 days in a large cohort of patients with status epilepticus (SE) using a machine learning system.

We included consecutive patients with SE admitted from 2013 to 2021 at Modena Academic Hospital. A decision tree analysis was performed using the 30-day mortality as a dependent variable and the EMSE predictors as input variables. We evaluated the accuracy of EMSE in predicting 30-day mortality using the area under the receiver operating characteristic curve (AUC ROC), with 95% confidence interval (CI). We performed a subgroup analysis on nonhypoxic SE.

A total of 698 patients with SE were included, with a 30-day mortality of 28.9% (202/698). The mean EMSE value in the entire population was 57.1 (SD = 36.3); it was lower in surviving compared to deceased patients (47.1, SD = 31.7 vs. 81.9, SD = 34.8; p < .001). The EMSE was accurate in predicting 30-day mortality, with an AUC ROC of .782 (95% CI = .747-.816). Etiology was the most relevant predictor, followed by age, electroencephalogram (EEG), and EMSE comorbidity group B. The decision tree analysis using EMSE variables correctly predicted the risk of mortality in 77.9% of cases; the prediction was accurate in 85.7% of surviving and in 58.9% of deceased patients within 30 days after SE. In nonhypoxic SE, the most relevant predictor was age, followed by EEG, and EMSE comorbidity group B; the prediction was correct in 78.9% of all cases (89.6% in survivors and 46.1% in nonsurvivors).

This validation study using a machine learning analysis shows that the EMSE is a valuable prognostic tool, and appears particularly accurate and effective in identifying patients with 30-day survival, whereas its performance in predicting 30-day mortality is lower and needs to be further improved.