Online Evidence of Withdrawal Self-Medication

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Withdrawal symptoms from opioid use can be severe and are major contributing factors to relapse and continuing misuse. Many opioid users are actively experimenting with “remedies” that can alleviate withdrawal, and they are discussing their effectiveness in blogs and forums. In this pilot study we will use Natural Language Processing (NLP) and human expertise to examine over 50,000 recent posts in two Reddit forums OpiatesRecovery and Opiates to assess systematically which remedies are being used, how they are being used, and what are the reported consequences of such self-help experimentation. We will create a curated database of user-reported “remedies.” Information will be semiautomatically extracted from the online, self-reported use of alternative treatments (i.e., other prescription drugs, over the counter medications, food supplements, activities such as meditation and yoga). A team of a pharmacologist, physician, and ethnographer will evaluate database entries to uncover (1) potential harm associated with uncontrolled and unsupervised experimentation, (2) potentially effective available treatments (e.g., traditional medicine), (3) potentially promising compound leads, and (4) patients' needs and issues that are most important to them. Aim 1. To assemble an extensive database of opioid withdrawal and remedy-associated terminology from posts on OpiatesRecovery and Opiates Reddit communities. NLP will be used to build a language model that understands how words are used in context (word2vec). Aim 2. To develop a dataset of instances of self-reported remedy use from Reddit and conduct a bipartite network analysis of remedies and users. Using NLP tools and the word embedding model we will develop an exclusive dataset containing extracted information associated with remedies targeting withdrawal and craving. This aim will use elements of artificial intelligence and close human supervision to extract remedies, including variations of spelling, from the texts. The result of this aim will be a remedy database that includes spelling variations and slang references and a network analysis linking remedies and users. Aim 3. To organize, aggregate, and systematically assess information from mentions of remedy use. Potential compounds and other remedies will be classified to provide an initial assessment of their potential relevance to the opioid treatment process. This process will require the most human oversight and assessment. Network analysis tools will be used to assess and identify the relationships between the types of remedies and potential therapeutic effect and will create the benchmarks for similar future studies.