Limited Accuracy of Prognostic Models for Predicting in-Hospital Mortality Among Patients with Refractory Status
Dannys Rivero Rodríguez1,2, Yanelis Pernas Sanchez3, Daniela DiCapua Sacoto4
1Department of Neurology, Hospital Eugenio Espejo, Quito, Ecuador. dannysr@sescam.jccm.es.
Background:
The prognostic value of scoring systems for status epilepticus (SE), including the Status Epilepticus Severity Score (STESS), modified STESS (mSTESS), and Epidemiology-Based Mortality Score in SE (EMSE), has been demonstrated primarily in studies from high-income countries (HICs). However, data on their performance in subgroups such as older adults, patients with refractory SE (RSE) in intensive care unit (ICU), and populations from low- and middle-income countries (LMICs) remain limited.
Methods:
Retrospectively analysis was performed on data from the Eugenio Espejo Hospital SE database (November 2015-January 2020). Among 112 evaluated patients, 109 SE episodes in 107 individuals were included. Receiver operating characteristic (ROC) curves for STESS, mSTESS, and EMSE variants were generated to assess predictive accuracy for in-hospital mortality in patients treated outside the ICU and those treated in the ICU. The DeLong test was used to compare areas under the ROC curve (AUROC).
Results:
A total of 50 of 109 SE episodes required ICU admission, and 42 were treated with anesthetic drugs, with induction of a therapeutic coma. Acute etiologies were more frequent among refractory episodes (40.7% vs. 78.0%, p = 0.03). Unfavorable outcomes at discharge were significantly higher among patients treated in the ICU (p = 0.04). Across the entire cohort, AUROC values for in-hospital mortality were 0.74 (95% CI 0.65-0.84) for STESS, 0.77 (95% CI 0.68-0.86) for mSTESS, 0.80 (95% CI 0.72-0.89) for EMSE-Etiology, Age, and Comorbidity (EAC), 0.82 (95% CI 0.74-0.90) for EMSE-ECLE, and 0.83 (95% CI 0.74-0.91) for EMSE-Etiology, Age, Comorbidity, and Electroencephalography Pattern (EACE). Among patients treated in the ICU, predictive performance decreased for STESS (0.77 vs. 0.71, p = 0.58), mSTESS (0.79 vs. 0.75, p = 0.65), EMSE-EAC (0.89 vs. 0.70, p = 0.03), EMSE-Etiology, Comorbidity, Level of Consciousness, and Electroencephalography Pattern (ECLE; 0.90 vs. 0.71, p = 0.02), and EMSE-EACE (0.92 vs. 0.71, p = 0.01).
Conclusions:
The prognostic performance of STESS, mSTESS, and EMSE scores for predicting in-hospital mortality declined among patients with RSE treated in the ICU. This reduction, particularly evident for EMSE, underscores the influence of disease severity and resource limitations in LMIC settings.