Annual Meeting Abstracts: View

  • (Abst. 1.424), 2019
  • Understanding Seizure Clusters: Epidiary Text Notes Compared with Diary Events
  • Authors: Robert S. Fisher, Stanford Comprehensive Epilepsy Center; Joyce A. Cramer, Yale University School of Medicine; Eyal Bartfeld, Irody, Inc.
  • Content:

    Rationale: Patients describe seizure clusters in many ways. Free-text comments in digital epilepsy diaries represent an under-utilized information resource and give insight (in their own words) into priorities of people with epilepsy. We evaluated patient notes using lexical analysis and diary events to characterize features of seizure clustering, furthering prior observations from diary forms (Fisher RAS, Bartfeld E, Cramer JA. Use of an online epilepsy diary to characterize repetitive seizures. Epilepsy Behav, 2015;47:66-71).  Methods: EpiDiary (www.EpiDiary.com) data includes structured fields for seizure frequency per calendar day, a checkbox for seizure clusters on a given seizure event and a free-text option. For purposes of this analysis, clusters were defined as at least two seizures reported by the patient (including auras or myoclonic jerks) in a calendar day. Lexical analysis was employed to extract comments containing the words: 'lots / many / repeat* / frequent* / cluster*' (*=wildcard string, such as 'repeatedly'). The authors reviewed every extracted comment to limit results only to text correctly pertaining to clusters. Entries referring to prior clusters or plans for clusters were not included in the analysis. Data were used to characterize repetitive seizures. Results: 1,839 users, generating a total of 5,955 comments, were flagged by the lexical analysis. Clinician review validated that the comments actually pertained to seizure clusters, as opposed to clusters of migraines, etc., in 2,645 (44.4%) of comments. Comments highlighting clusters did not show a corresponding checkmark in the diary cluster box in 1,560 instances (58.9%). Seizure count was not recorded for 1,192 comments, and 512 comments indicating clusters presented neither a cluster checkmark, nor a seizure count. Extra medicine was taken for clusters by 553 users on 3,818 days, representing 30.1% of all users and 56.5% of those commenting on clusters. Rescue medicines specifically named in a comment were: lorazepam in 99, clonazepam in 84, midazolam in 54, clobazam in 31, rectal diazepam in 29, other diazepam in 18, and clorazepate in 9. Some used more than one rescue medicine. The VNS magnet also was mentioned as a treatment for clusters. Comments often provided potentially useful clinical information, such as “massive ongoing cluster with about 20% apneic events.” Interpretations are limited, as not all comments might have been captured. Some subjects used multiple accounts. Words used to identify seizures, e.g., “puffs,” might or might not have been about seizures. Occasional comments were copied verbatim from one date to another. Subjects used their own definitions of clusters.  Conclusions: Free-text comments in the digital seizure diary provided additional information about the seizure clusters. More sophisticated lexical and grammatical models will likely improve accuracy of relevant comment identifications in the future. Funding: RSF holds stock or options in Irody, Avails Medical, Smart-Monitor, Cerebral Therapeutics, Eysz and consults for Medtronic.JC is Consultant for Eisai, Irody, Sunovion, UCB.EB is Founder and President of Irody, Inc, which supported this study.