School of Interdisciplinary Informatics, University of Nebraska at Omaha, United States.
School of Interdisciplinary Informatics, University of Nebraska at Omaha, United States.
Int J Med Inform. 2018 Dec;120:92-100. doi: 10.1016/j.ijmedinf.2018.10.002. Epub 2018 Oct 9.
Twitter has become one of the most popular social media platforms that offers real-world insights to healthy behaviors. The purpose of this study was to assess and compare perceptions about chemotherapy of patients and health-care providers through analysis of chemo-related tweets.
Cancer-related Twitter accounts and their tweets were obtained through using Tweepy (Python library). Multiple text classification algorithms were tested to identify the models with best performance in classifying the accounts into individual and organization. Chemotherapy-specific tweets were extracted from historical tweetset, and the content of these tweets was analyzed using topic model, sentiment analysis and word co-occurrence network.
Using the description in Twitter users' profiles, the accounts related with cancer were collected and coded as individual or organization. We employed Long Short Term Memory (LSTM) network with GloVe word embeddings to identify the user into individuals and organizations with accuracy of 85.2%. 13, 273 and 14,051 publicly available chemotherapy-related tweets were retrieved from individuals and organizations, respectively. The content of the chemo-related tweets was analyzed by text mining approaches. The tweets from individual accounts pertained to personal chemotherapy experience and emotions. In contrast with the personal users, professional accounts had a higher proportion of neutral tweets about side effects. The information about the assessment of response to chemotherapy was deficient from organizations on Twitter.
Examining chemotherapy discussions on Twitter provide new lens into content and behavioral patterns associated with treatments for cancer patients. The methodology described herein allowed us to collect relatively large number of health-related tweets over a greater time period and exploit the potential power of social media, which provide comprehensive view on patients' perceptions of chemotherapy.
This study sheds light on using Twitter data as a valuable healthcare data source for helping oncologists (organizations) in understanding patients' experiences while undergoing chemotherapy, in developing personalize therapy plans, and a supplement to the clinical electronic medical records (EMRs).
Twitter 已成为最受欢迎的社交媒体平台之一，可提供有关健康行为的真实见解。本研究旨在通过分析与化疗相关的推文来评估和比较患者和医疗保健提供者对化疗的看法。
使用 Tweepy（Python 库）获取与癌症相关的 Twitter 账户及其推文。测试了多种文本分类算法，以确定将账户分类为个人和组织的最佳性能模型。从历史推文集中提取化疗专用推文，并使用主题模型、情感分析和词共现网络分析这些推文的内容。
使用 Twitter 用户个人资料中的描述，收集与癌症相关的账户并将其编码为个人或组织。我们使用带有 GloVe 词嵌入的长短期记忆（LSTM）网络将用户识别为个人和组织，准确率为 85.2%。从个人和组织分别检索到 13,273 条和 14,051 条公开的与化疗相关的推文。通过文本挖掘方法分析与化疗相关的推文的内容。个人账户的推文涉及个人的化疗经历和情绪。与个人用户相比，专业账户关于副作用的中性推文比例更高。关于对化疗反应的评估信息在 Twitter 上的组织中缺乏。
检查 Twitter 上的化疗讨论为了解癌症患者治疗相关的内容和行为模式提供了新的视角。本文描述的方法使我们能够在更长的时间内收集相对大量的与健康相关的推文，并利用社交媒体的潜力，为患者对化疗的看法提供全面的了解。
这项研究揭示了将 Twitter 数据用作有价值的医疗保健数据源的潜力，可帮助肿瘤学家（组织）了解患者在接受化疗时的体验，制定个性化的治疗计划，并补充临床电子病历（EMRs）。
