Risk factors and a predictive model for the development of epilepsy after Japanese encephalitis.

Background: We aimed to study seizure characteristics during the acute phase of Japanese encephalitis (JE) in children, determine the risk factors of postencephalitic epilepsy (PEE), establish a risk prediction model for the disease, and construct a nomogram to visualize the model.

Methods: We retrospectively analyzed the clinical data and follow-up results of 328 children with JE who were hospitalized between January 2011 and December 2020. Risk factors were screened using univariable analysis, a predictive model was built using binary logistic analysis, lasso regression was used for variable screening, and a nomogram was developed.

Results: Of the 328 children with JE enrolled in the study, 216 (65.9%, 216/328) had seizures in the acute phase. The incidence of PEE was 14.7% (39/264), The cumulative percentages of PEE after discharge was 10.6% (28/264)at 6 months, which increased to 13.6%(36/264)at 3 years. 38.5% of patients with PEE had generalized onset seizures, and 17.9% had focal motor seizures. Univariable analysis revealed that 22 clinical indicators were related to the PPE; Multivariable analysis identified seizure number >5 (OR (95%CI) = 3.013 (1.046-8.676), P = 0.041), status epilepticus (OR (95%CI) = 3.918 (1.212-12.669), P = 0.023), and Coma (OR (95%CI) = 22.495 (8.686-58.285), P<0.001) as independent risk factors for PEE. The risk prediction model: ln(p/1p)= -3.533 + 1.103 × (seizures number > 5) +1.366 × (status epilepticus) + 3.113 × (Coma) was developed, and a nomogram was constructed. The area under the ROC curve (AUC), calibration plot, and Hosmer-Lemeshow test showed that the model had good discrimination and calibration. Ordinary bootstrapping was used for internal validation, and the predictive results of the original and test sets were consistent.

Conclusions: Seizure is a common manifestation during acute encephalitis and sequelae in children with JE. The nomogram constructed in this study could be used for early prediction, and could facilitate early intervention.