Recommending Patient Expertise Within Online Health Communities

Internet-enabled technologies have created unprecedented opportunities for sharing experiences with virtually anyone worldwide. Within this context, online health communities serve as platforms for the exchange of information and advice, thereby facilitating the sharing of patient expertise. This expertise, acquired from managing personal health issues in daily life, is often reflected in valuable insights such as disease management strategies. However, locating these strategies and other forms of patient expertise in online health communities poses a significant challenge for their members, primarily due to two key reasons. First, the continuous influx of user-generated content can bury valuable knowledge previously constructed through social interaction. Second, the information search tools available on these platforms could unintentionally exacerbate the very problem they are designed to address. This doctoral dissertation investigates user-centered approaches for recommending patient expertise within online health communities, aiming to alleviate the burden of explicit query formulation and directly address the challenge of finding relevant information amidst vast volumes of online discussions. The dissertation's core objective was addressed through an incremental and iterative research process. Specifically, findings from earlier studies informed the direction of subsequent research. Furthermore, we adopted a research-through-design approach, where design activities were central to knowledge generation. Chapter 2 presents the first contribution of this dissertation. Specifically, we investigate how individuals with cardiovascular disease conceptualize sharing personal health information with their peers. Our findings underscore that individuals primarily value learning from each other's experiences, rather than their physiological data. Chapter 3 reports on a usability evaluation of a low-fidelity mobile application prototype, investigating how we may design to facilitate locating peers and learning self-care ideas from them in online health communities. Our findings reveal the central role of the lifestyle profile in locating and choosing peers. In light of this, we argue that because individuals with this condition seek to lead normal lives, the most helpful suggestions in this quest may come from peers with similar habits and activity levels—attributes that operationalize lifestyle. Building upon these findings, Chapter 4 further examines the influence of lifestyle, alongside demographic and clinical profiles, through user-controllable peer recommendations. We conducted a user-centered evaluation of a web-deployed application for a simulated online health community, which recommended potential connections with varying levels of user control. A second direction emerging from the research in Chapter 3 focused on the collection of everyday self-care strategies and the provision of personalized recommendations for these strategies. Chapter 5 presents this research direction, beginning with a crowdsourcing study on self-care strategies that individuals with cardiovascular disease employ in their daily lives. Using the data collected from this study, we then simulated an online health community and implemented a collaborative filtering recommender system. This chapter, therefore, also reports on a user-centered evaluation of recommendations for everyday self-care, specifically comparing the user experience with personalized versus non-personalized recommendations. By systematically investigating the user experience through an iterative and incremental research process, this dissertation proposes practical considerations for designing user-centered recommender systems that facilitate locating individual experts and the community's aggregate expertise.