Clinico-psychosocial prediction for seizure emergencies: A multicenter derivation-validation study.

Abstract

Objectives: To develop a multidimensional predictive model for emergency hospitalization due to recurrent epileptic seizures, aiming to reduce the burden on healthcare systems and improve patient outcomes through timely interventions and individualized support.

Methods: We conducted retrospective and prospective multicenter derivation-validation cohort studies (n = 230 and 505, respectively). The derivation cohort, comprising patients retrospectively assessed between 2019 and 2020 at a single epilepsy center, was used to analyze clinical, social, and psychological factors associated with seizure worsening. Three predictive models were constructed: clinical, clinico-social, and clinico-psychosocial. These models were prospectively validated at three tertiary care hospitals between 2023 and 2024 to predict seizure recurrence requiring emergency hospitalization. Predictive performance was evaluated using receiver operating characteristic (ROC) analysis, and comparative performance was assessed.

Results: In the derivation cohort, focal epilepsy (p = 0.015), high seizure frequency (p = 0.002), psychogenic non-epileptic seizures (p < 0.001), mood disorders (p < 0.001), and living alone (p < 0.001) were identified as significant predictors of seizure worsening. Three predictive scores were developed: a clinical score (including focal seizures and high seizure frequency); a clinico-social score (incorporating social factors); and a clinico-psychosocial score (incorporating psychosocial variables). The clinical score yielded an area under the curve (AUC) of 0.87 (95% CI, 0.76-0.93), whereas the clinico-psychosocial score demonstrated superior predictive performance, with an AUC of 0.96 (95% CI, 0.92-0.98; p=0.016).

Significance: The clinico-psychosocial model, reflecting drug-resistant focal epilepsy, social isolation, and psychological distress, exhibited excellent predictive accuracy for emergency hospitalization. This model can support clinicians in early identification of high-risk patients, thereby potentially reducing the incidence of recurrent seizures and associated social consequences.