Enabling health information recommendation using crowdsourced refinement in online health information applications: EndoZone informatics study (Preprint)

Abstract

Nowadays, people predominantly rely on search engines and social media platforms as their initial source of health information. The recommendation algorithms and the underlying architecture behind these platforms are designed to rank the information by default with commercial rather than health intent, which result in large amounts of unrelated articles and posts to be recommended. Developing online health information applications by trusted sources is a promising way to bypass the information recommendation algorithms in search engines and social media. However, existing research lacks comprehensive guidance on integrating content ranking algorithms into these applications.

This paper aims to outline a generic methodology to guide the design and implementation of health information recommendation functionality within online health information applications.

The methodology concept and implementation-related considerations during software design and develop, operate, review, and optimize stages are firstly presented in detail. Using the methodology as a guide, the information recommendation functionalities in an endometriosis information platform named EndoZone was implemented. The applicability of the methodology is illustrated through the actual implementation process as well as the outcome of early operation data review.

The personalised health information recommendation component of the EndoZone platform (EndoZone informatics) has been implemented strictly following the methodology. The implementation process and volunteer group feedback show that the methodology is practical for enabling information recommendation functionalities for online health information applications that have targeted health content sharing requirements. Early data shows that the solution built using this methodology is effective in reflecting users’ feedback, providing highly personalized information recommendation and highly flexible in adjusting information recommendation logic. It has been observed that the methodology is not able to identify project specific design defects during early design and development stage, but issues can be identified and addressed in later review and optimize stage.

We propose a generic methodology to guide the design and implementation of health information recommendation functionality within online health information applications. This methodology recommends trusted health information based on people’s medical needs, and utilizes crowdsourced data, i.e., user feedback post content access to rank the recommendations. The successful application of the methodology for EndoZone informatics is a solid step towards enabling personalized health information recommendation at scale based on people’s medical needs.