Stanford Department of Neurology and Neurological Sciences, USA.
Yale University School of Medicine, and Consultant Houston, TX, USA.
Epilepsy Behav. 2020 Dec;113:107498. doi: 10.1016/j.yebeh.2020.107498. Epub 2020 Oct 20.
Online seizure diaries offer a wealth of information regarding real world experience of patients living with epilepsy. Free text notes (FTN) written by patients reflect concerns and priorities of patients and provide supplemental information to structured diary data.
This project evaluated feasibility using an automated lexical analysis to identify FTN relevant to seizure clusters (SCs).
Data were extracted from EpiDiary™, a free electronic epilepsy diary with 42,799 unique users, generating 1,096,168 entries and 247,232 FTN. Both structured data as well as FTN were analyzed for presence of SC. A pilot study was conducted to validate an automated lexical analysis algorithm to identify SC in FTN in a sample of 98 diaries. The lexical analysis was then applied to the entire dataset. Outcomes included cluster prevalence and frequency, as well as the types of triggers commonly reported.
At least one FTN was found among 13,987 (32.68%) individual diaries. An automated lexical analysis algorithm identified 5797 of FTN as SC. There were 2423 unique patients with SC that were not identified by structured data alone and were identified using lexical analysis of FTN only. Seizure clusters were identified in n = 10,331 (24.1%) of diary users through both structured data and FTN. The median number of SCs days per year was 13.7, (interquartile rank (IQR): 3.2-54.7). The median number of seizures in a cluster day was 3 (IQR 2-4). The most common missed medication linked to patients with SC was levetiracetam (n = 576, 29%) followed by lamotrigine (n = 495, 24%), topiramate (n = 208, 10.5%), carbamazepine (n = 190, 9.6%), and lacosamide (n = 170, 8.6%). These percentages generally reflected prevalence of medication use in this population. The use of rescue medications was documented in 3306 of structured entries and 4305 in FTN.
This exploratory study demonstrates a novel approach applying lexical analysis to previously untapped FTN in a large electronic seizure diary database. Free text notes captured information about SC not available from the structured diary data. Diary FTN contain information of high importance to people with epilepsy, written in their own words.
在线癫痫日记为了解癫痫患者的真实世界体验提供了丰富的信息。患者撰写的自由文本注释 (FTN) 反映了患者的关注点和优先事项，并为结构化日记数据提供了补充信息。
本项目使用自动词汇分析评估了识别与癫痫发作群 (SCs) 相关的 FTN 的可行性。
从 EpiDiary™ 中提取数据，这是一个免费的电子癫痫日记，拥有 42799 名独特用户，生成了 1096168 条条目和 247232 条 FTN。对结构化数据和 FTN 都进行了分析，以确定是否存在 SC。进行了一项试点研究，以验证一种自动词汇分析算法，以在 98 份日记的样本中识别 FTN 中的 SC。然后将词汇分析应用于整个数据集。结果包括簇的流行率和频率，以及常见的触发因素类型。
在 13987 名（32.68%）个体日记中至少发现了一条 FTN。自动词汇分析算法确定了 5797 条 FTN 是 SC。有 2423 名独特的患者，仅通过结构化数据无法识别，仅通过 FTN 的词汇分析才能识别。通过结构化数据和 FTN，在 n=10331 名（24.1%）日记用户中确定了癫痫发作群。每年 SC 天数的中位数为 13.7 天（四分位间距 (IQR)：3.2-54.7）。簇日发作的中位数为 3 次（IQR 2-4）。与有 SC 的患者相关的最常见漏服药物是左乙拉西坦（n=576，29%），其次是拉莫三嗪（n=495，24%）、托吡酯（n=208，10.5%）、卡马西平（n=190，9.6%）和拉科酰胺（n=170，8.6%）。这些百分比通常反映了该人群中药物使用的流行率。在结构化条目 3306 条和 FTN 4305 条中记录了抢救药物的使用情况。
本探索性研究展示了一种应用词汇分析的新方法，该方法应用于大型电子癫痫日记数据库中以前未开发的 FTN。自由文本注释记录了结构化日记数据中未包含的有关 SC 的信息。日记的 FTN 包含了患者用自己的语言描述的、对癫痫患者非常重要的信息。
