Social Media Analytics in Health

Introduction

Nowadays, with the increasing popularity of social media, Internet users are becoming more accustomed to sharing daily life, commenting on topics as well as expressing opinions on social media ^[1]. In addition, there are more and more Internet users using social media as a channel for information search ^[2]. At the same time, with the rapid development of big data, machine learning, natural language processing, sentiment analysis, opinion mining algorithms, social network analysis, recommendation and other related technologies, it becomes possible to perform reliable Social Media Analytics in various fields, so does in health.

Imagine that in the near future, based on people’s activities on social media such as Facebook, Twitter, Instagram and Weibo, to assess personal health and predict illness, then conduct timely treatment or prevention, it will actively and effectively ensure healthy lives and promote well-being.

Main Steps

Below are 10 shared steps to social media analytics success. And in the following, I will make detail explanation about some main steps in health area in the workflow of Social Media Analytics mentioned in my first post.

Figure 1 10 simple steps to Social Media Analytics success ^[3]

1. Motivation Setting

First and foremost is to clarify the purpose of Social Media Analytics. In contributing to promote human health, motivations can be set like to investigate the leading causes of premature death and disability, such as heart disease, diabetes, hypertension, obesity, chronic lung disease, depression and substance abuse; predict public health events; assess personal health; equip users with emotional and health support; provide early warning of adverse drug reactions; help build a healthy lifestyle, etc.

2. Data Collection

When conducting data collection, in addition to activities such as posts, reposts, or searches on social media that are directly related to health or illness, we should also pay attention to some other life details that seem unrelated to the user’s health, such as posts on opinions about events, time and types of daily meal, family relationships, etc. Because it turns out that every post can somehow expose the user’s health and even underlying illnesses.^[4]

At the same time, we must pay attention to privacy issues, obtain user consent, and collect data in a transparent way. A study involving social media and healthcare at Penn’s Center for Health Care Innovation shows that the vast majority of 1,432 Facebook and Twitter users in the University’s medical system agree to share their social media activities and compare them with their electronic medical records, accounting for about 71%. Centre director David Asch said that, in his personal experience, in experiments conducted in laboratories so far, people seem to be comforted by the idea that their health is being “surveilled” by the local hospital or health system. Although people really care about their privacy, they also realize the value of sharing this information to themselves and society. As a result, we can reasonably assume that most people will agree if they were told that their social media data was collected to improve their personal and social health. ^[5]

3. Preprocessing

The user scale of social media is huge, and the topics covered are wide. There is no doubt that social media is a valuable resource library containing a large number of user opinions and behaviors. However, social media is an unstructured data source. There are many ways for users to express opinions, which makes the opinion text in social media appear fragmented, noisy, and unstructured. Therefore, algorithms must be run for preprocessing, separating noise and useful information, turning text strings into some kind of meaningful statistics. It is still an important challenge to automatically analyze and summarize the opinions of all users of social media by means of computer.

4. Analysis

There have been many methods of Social Media Analytics existed, such as automatic recognition of posts related to a certain disease through neural networks and deep learning methods ^[6]. Semantic analysis is also worth attention. In machine learning, semantic analysis for a corpus is to establish a structure that can approximately express the concepts contained in a large number of documents. Semantic analysis is therefore a useful analysis tool to evaluate the structure and meaning of a large number of information flows in social networks, human behavior, and network communication. There have been several studies on social media semantic analysis in the field of computer science, such as combining semantic analysis and sentiment analysis to measure the correlation between certain concepts in Twitter text and negative/positive emotions. However, there is not much research on the semantic nature and structure of health-related information in social media. ^[7]

5. Result Presentation

Just as the genome reflects human genes, combined with various useful data and relevant analysis results, we can use “social health genes” to describe the health of the public on social media. At the same time, a health footprint database composed of people’s shared information can be used to build a social health map.

6. Applications

Figure 2 6 Do’s And 5 Don’ts with Social Media for Healthcare ^[8]

• Personal Health

The use of social media enhances the interaction between users and provides health support for online social media users by searching, utilizing, sharing and commenting on health information as well as building communities and discussion groups. Interactions between individuals in social media can also help users improve their health or become aware of health issues.

More and more people with mental disorders are starting to look for help through social media, share their medical history or seek advice from others on similar health conditions. This emerging form of spontaneous communication between online communities and individuals with different health issues is called peer-to-peer support. ^[9]

• Public Health

In addition to focusing on individuals, Social Media Analytics can also be applied to a wide range of public health trends research ^[4].

Such as health monitoring and prediction of large-scale diseases. A widely reported example was in the late 2000s, when Google struggled to analyze search queries to predict flu outbreaks earlier than the Centers for Disease Control (CDC). According to Smithsonian Journal, the project was not very successful. It has always greatly overestimated the incidence of influenza, but its basic concept of predict public disease based on social media is still promising.

Another application is using social media to analyze people’s understanding of diseases such as heart disease, hypertension, and diabetes. If find people generally have wrong views, we can use the information diffusion capabilities of social media platforms to spread health information, build relevant knowledge and quickly popularize the public.

Social Media Analytics on patients can help hospitals get feedback that is useful for their services. Andy Schwartz’s team published a study on Yelp in Health Affairs that analyzed how people think about hospitalization. American hospitals usually use standard patient satisfaction surveys called HCAHPS (Hospital Consumer Assessment of Healthcare Providers and Systems Survey) for evaluation. But Yelp’s research shows that HCAHPS failed to ask questions that are important to patients, such as parking and responding to tolls, which are widely discussed in patients’ social media. Hospitals can use the findings to improve their services.

Significance

The empowerment of social media enables the networked public to assume a more important role in sharing health knowledge, experience and emotional feelings. Public participation will therefore play a greater role in subsequently promoting awareness of health information, inspiring healthy behaviors, strengthening decisions about healthy behaviors, and maintaining loyalty in healthy behaviors. Hence, Social Media Analytics in health has certain theoretical and practical significance for achieving SDG3, i.e. ensuring healthy lives and promoting well-being for all at all ages.

Reference

[1] Xie Songxian, Open Mining and Application on Social Media[D]

[2] Kevin H G, Dan Z , Chengrui C , et al. Talking about Cancer on Twitter: Health Semantics and Social Media[J]. Chinese Journal of Journalism & Communication, 2017.

[3] Is Your Social Media Strategy Working? 10 Steps to Analyse Performance [Infographic], https://blog.red-website-design.co.uk/2017/06/01/social-media-strategy-working-10-steps-analyse-performance-infographic/

[4] Social Media Posts Expose Your Health, http://www.knowledgeatwharton.com.cn/article/8785

[5] Abbasi A, Adjeroh D, Dredze M, et al. Social Media Analytics for Smart Health[J]. Intelligent Systems, IEEE, 2014, 29(2):60-80.

[6] Gkotsis G, Oellrich A, Velupillai S, et al. Characterisation of mental health conditions in social media using Informed Deep Learning[J]. Scientific Reports, 2017(7): 45141.

[7] Kotov A, Social Media Analytics for Healthcare[C]. IEEE/WIC/ACM International Joint Conferences on Web Intelligence. IEEE, 2015.

[8] What Are 6 Do’s And 5 Don’ts with Social Media for Healthcare? #HCSM #infographic, https://www.dr4ward.com/dr4ward/2016/02/what-are-6-dos-and-5-donts-with-social-media-for-healthcare-hcsm-infographic.html

[9] Luo Xiaolan. Health Information Analysis and Health Promotion in Social Media [J]. Chinese Journal of Medical Library and Information Science, 2017 (10): 26-33.