Abstract
BACKGROUND AND PURPOSE: Epilepsy is associated with increased risk of sudden death (SD). The risk factors of developing SD in patients with epilepsy are not fully addressed. This study aimed to develop and validate a short-term prediction model for SD in patients with epilepsy using nationwide data from Taiwan. METHODS: A retrospective cohort study used data from the National Health Insurance Research Database from 2011 to 2020. It focused on epilepsy patients over 18 years old who were prescribed anti-seizure medication (ASM). The study evaluated patient comorbidities, medication adherence, and recent hospital admissions. It aimed to assess the association between these factors and the occurrence of SD within 30 days. The analysis used a multiple logistic regression model and decision-tree classifier and assessed predictive accuracy using the area under the receiver operating characteristic curve. RESULTS: Out of 161,773 treatment events, 3,454 SD events were identified (2.1%). Factors associated with increased risk of SD included older age, intensive care unit admission, chronic kidney disease, psychotic disorder, poor ASM adherence (medication possession rate <0.5), and recent intravenous ASM use. The logistic model's area under curve was 0.752 in the 2020 testing dataset, and the testing dataset's sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were 0.569, 0.832, 3.391, 0.518, and 6.543, respectively. A calibration curve showed reasonable alignment between predicted and observed probabilities of SD. CONCLUSIONS: Patients with epilepsy recently admitted to the hospital, showing poor ASM compliance and using intravenous ASMs, face an increased risk of SD within 30 days. Improving ASM adherence and treatment optimization may reduce this risk.