Social Media Mining for Pharmacovigilance

项目摘要 在我们的母公司拨款中，总体目标是开发新的NLP方法，以利用社交媒体(SM)数据 特异性药物警戒(PV)努力受到传统药物警戒方法已知缺陷的阻碍。 我们专注于促进使用纵向SM数据来探索(A)影响用药的因素的方法 在普通民众中坚持和坚持(目标1)，以及(B) 怀孕期间服用的药物和妊娠结局(目标2)。我们还解决了以下方面的主要障碍 基于这些数据的病例对照研究：寻找对照(目标3)。总的来说，方法的重点是 Parent Grant正在推进NLP方法，使SM能够大规模集成为内容丰富的 流行病学研究中用户生成数据的来源，包括随时间推移发布的相关数据。我们是 寻求扩展为父母资助的目标1、2和3开发的方法，以专注于阿尔茨海默氏症 和阿尔茨海默病相关痴呆(AD/ADRD)。本附录的目标是回应 Not-AG-20-034，并包括以下内容：目标1是开发和评估NLP方法以识别 并描述了一群受AD/ADRD诊断影响的Twitter用户，特别是 声明其直系亲属(父母、祖父母、兄弟姐妹)、重要他人，甚至其本人， 都被诊断出患有AD/ADRD。收集的队列中用户的时间线将被进一步挖掘以 人口特征(年龄、性别、种族和居住的地理位置)和具体情况 将开发方法来提取AD/ADRD患者的家庭关系，以及与 两个特定的重点研究：1)关于早发(也称为年轻发病)和潜在的终生因素的研究 反映在SM和2)对整个队列中使用者服用药物的研究，并特别关注 神经精神症状也与AD/ADRD有关。目标2是开发和评估 一种语言模型，可以在SM中自动对患者的书面成果进行分类，以确定潜在的 通过分析被诊断为EoAD的人随着时间的推移在SM上发布的帖子来发现认知障碍的迹象 认知障碍明显表现在他们的叙事文本的内容、语法和形式上。除了方法之外 和已经在母公司Grant下开发的系统，这些任务利用基线方法以及 我们自己的推特收藏。这些资源使我们处于完成研究方法的独特地位 这通常需要几年的时间，完全在行政部门涵盖的一年内 补充，使我们能够迅速加强我们的合作和研究，以便进一步提交提案 NIA或NIGMH。

Project Summary In our parent grant, the overall goal is to develop novel NLP methods to leverage Social Media (SM) data for specific pharmacovigilance (PV) efforts that are hindered by known drawbacks of traditional PV approaches. We focus on methods to facilitate the use of longitudinal SM data for exploring (a) factors affecting medication adherence and persistence among the general population (Aim 1), and (b) possible associations between medications taken during pregnancy and pregnancy outcomes (Aim 2). We also address a major roadblock for case-control studies based on this data: finding controls (Aim 3). In general, the methodological focus of the parent grant is on advancing NLP methods that enable the large-scale integration of SM as a content-rich source of user-generated data in epidemiological studies, including relevant data posted over time. We are seeking to extend methods developed for Aims 1, 2, and 3 of the parent grant to focus on Alzheimer's Disease and Alzheimer's Disease Related Dementia (AD/ADRD). The Objectives for this supplement are responsive to NOT-AG-20-034, and include the following: Objective 1 is to develop and evaluate NLP methods to identify and characterize a cohort of Twitter users affected by an AD/ADRD diagnosis, specifically, Twitter users who declare that their direct relatives (parents, grandparents, or siblings), significant others, or even themselves, have been diagnosed with AD/ADRD. The collected timelines of users in the cohort will be further mined for demographic characterization (age, gender, ethnicity, and geographic location of residence) and specific methods will be developed to extract familial relationship to the AD/ADRD patient, and information relevant to two specific focus studies: 1) a study on early onset (also known as young onset) and potential lifetime factors reflected in SM and 2) study of medications taken by users in the whole cohort, with specific attention to neuropsychiatric symptoms also associated with AD/ADRD. Objective 2 is the development and evaluation of a linguistic model that can automatically classify patients' written productions in SM to determine potential cognitive impairment by analyzing postings over time on SM by people diagnosed with EOAD for signs of cognitive impairment evident in the content, grammar, and form of their narrative text. In addition to methods and systems already developed under the Parent Grant, the tasks leverage baseline approaches as well as our own collection of Tweets. These resources put us in the unique position to complete research approaches that would normally take several years well within the span of the year covered by the administrative supplement, allowing us to quickly ramp up our collaborations and research for further proposal submissions to the NIA or NIGMH.