Development of a nomogram model for early prediction of refractory convulsive status epilepticus

We aim to identify risk factors that predict refractory convulsive status epilepticus (RCSE) and to develop a model for early recognition of patients at high risk for RCSE.

This study involved 200 patients diagnosed with convulsive status epilepticus (CSE), of whom 73 were RCSE and 127 were non-RCSE. Variables included demographic information, lifestyle factors, medical history, comorbidities, clinical symptoms, neuroimaging characteristics, laboratory tests, and nutritional scores. A predictive model was developed through multivariable logistic regression analysis. The model's predictive performance and clinical utility were evaluated using various metrics, including the area under the receiver operating characteristic (AUROC) curve, GiViTI calibration belt, and decision curve analysis (DCA). Additionally, we performed internal five-fold cross-validation for this model.

We developed a nomogram model with six predictors: age ≤ 40 years, prior history of epilepsy, presence of epileptic foci, duration of CSE > 30 min, c-reactive protein > 6 mg/L, and nutritional risk screening ≥ 3 points. Our model has a high AUROC (0.838) and good consistency (P = 0.999). In DCA, the curve of our model exhibits a positive net benefit across the entire range of threshold probabilities. Moreover, our model achieved an accuracy of 0.778 and a Kappa value of 0.519 in the five-fold cross-validation.

We developed an objective, simple and accessible model to assess the risk of RCSE. This model shows promise as a valuable tool for evaluating the individual risk of RCSE.