HealthTrust: trust-based retrieval of health social media videos

Abstract

There is a global trend towards the use of the Internet to search for information about health issues. We have access to a wide range of online health information; especially the so-called social media (e.g. blogs, videos). However, finding good quality resources is not easy in the current context of information overload. Today, very relevant and valuable health social media has to compete in visibility with misleading information such as anti-vaccination and pro-anorexia content. General web information retrieval approaches, such as Google, tend to retrieve popular content that can be misleading or even repulsive. For example, people searching for videos about diabetes foot care will discover that the top videos retrieved by a YouTube search include macabre amputations. Traditional health information retrieval approaches based on quality labels face many scalability challenges. The PhD project described here focuses on the unmet need for better technical solutions for the retrieval of high quality and relevant health social media. This thesis summarizes nearly six years’ work in the field of health social media summarized in ten research papers. I have applied a wide range of research methods such as qualitative research with patients, web-data analysis and literature reviews. My first research challenge was to grasp some understanding of the emerging health social media avalanche where research literature was virtually nonexistent. Secondly, I explored a wide range of technical solutions for the retrieval of relevant and trustworthy health information such as web search-engines, recommender systems and personalized health education systems. Building on the knowledge acquired during the dissertation, I proposed a new trust-based metric called HealthTrust for the retrieval of health social media. HealthTrust is a metric measuring the trustworthiness of the content within a health community and it can be used to rank search results of health social media. The rationale for choosing an approach based on social network analysis within a health community relies on the assumption that health communities have a common shared knowledge about the relevance and trustworthiness of the content and their providers. The HealthTrust algorithm was successfully tested for the retrieval of diabetes social videos.