Development and external validation of a nomogram for predicting in-hospital mortality in status epilepticus patients in the intensive care unit: A MIMIC-IV and eICU-CRD study.

OBJECTIVE

To develop and externally validate a nomogram for predicting in-hospital mortality in ICU patients with SE, using readily available clinical data collected within the first 24 h of ICU admission.

METHODS

We identified adult SE patients from MIMIC-IV (training cohort, 2008-2019) using ICD-9/ICD-10 codes. Twenty-seven candidate predictors were screened by LASSO logistic regression with 10-fold cross-validation. The final model was constructed with multivariable logistic regression and presented as a nomogram. Internal validation used 1000 bootstrap resamples with optimism correction. External validation was performed in an independent SE cohort from the eICU-CRD database (2014-2015).

RESULTS

The training cohort comprised 299 patients (69 deaths, 23.1% mortality). LASSO selected nine predictors: age, first ICU care unit, Glasgow Coma Scale, temperature, SpO2, white blood cell count, glucose, bicarbonate, and the Elixhauser comorbidity index. The optimism-corrected AUC was .807 (95% CI, .754-.860) with a Brier score of .124. External validation in 597 eICU patients yielded an AUC of .750 (95% CI, .685-.815). In sensitivity analyses, a reduced model without the comorbidity score achieved an optimism-corrected AUC of .786 and an external AUC of .730, and a model without the ICU care-unit type achieved .817 and .764, respectively.

SIGNIFICANCE: This nine-variable nomogram provides a validated, bedside-applicable tool for mortality risk stratification in ICU SE patients, enabling manual risk calculation without electronic computation.