A web‐based algorithm to rapidly classify seizures for the purpose of drug selection. (22nd August 2021)

• Record Type: Journal Article
• Authors: Beniczky, Sándor; Asadi‐Pooya, Ali A.; Perucca, Emilio; Rubboli, Guido; Tartara, Elena; Meritam Larsen, Pirgit; Ebrahimi, Saqar; Farzinmehr, Somayeh; Rampp, Stefan; Sperling, Michael R.
• Is Part Of: Epilepsia. Volume 62:issue 10(2021)
• Page Start: 2474
• Page End: 2484
• Publication Date: 2021-08-22
• Subjects: algorithm -- classification -- epilepsy -- seizure -- web‐based application; Epilepsy -- Periodicals; 616.853
• DOI: 10.1111/epi.17039
• Languages: English
• ISSNs: 0013-9580

Abstract

Abstract: Objective: To develop and validate a pragmatic algorithm that classifies seizure types, to facilitate therapeutic decision‐making. Methods: Using a modified Delphi method, five experts developed a pragmatic classification of nine types of epileptic seizures or combinations of seizures that influence choice of medication, and constructed a simple algorithm, freely available on the internet. The algorithm consists of seven questions applicable to patients with seizure onset at the age of 10 years or older. Questions to screen for nonepileptic attacks were added. Junior physicians, nurses, and physician assistants applied the algorithm to consecutive patients in a multicenter prospective validation study (ClinicalTrials.gov identifier: NCT03796520). The reference standard was the seizure classification by expert epileptologists, based on all available data, including electroencephalogram (EEG), video‐EEG monitoring, and neuroimaging. In addition, physicians working in underserved areas assessed the feasibility of using the web‐based algorithm in their clinical setting. Results: A total of 262 patients were assessed, of whom 157 had focal, 51 had generalized, and 10 had unknown onset epileptic seizures, and 44 had nonepileptic paroxysmal events. Agreement between the algorithm and the expert classification was 83.2% (95% confidence interval = 78.6%–87.8%), with an agreement coefficient (AC1) of .82 (95% confidence interval = .77–.87), indicating almost perfect agreement. Thirty‐two health care professionals from 14 countries evaluated the feasibility of the web‐based algorithm in their clinical setting, and found it applicable and useful for their practice (median = 6.5 on 7‐point Likert scale). Significance: The web‐based algorithm provides an accurate classification of seizure types, which can be used for selecting antiseizure medications in adolescents and adults.