Can Natural Language Processing Estimate Patient Preferences?

Project Status: Active

Project Summary

“Social media listening” (SML) is an emerging method of preference exploration that overcomes some of the limitations of interviews and focus groups in which qualitative analysis techniques are applied to patient-generated discussions on social media (e.g., Cook et al. 2019a, 2019b, 2019c). This approach averts some of the time and resources that would otherwise be devoted to patient recruitment in traditional qualitative studies, and provides an easy way of engaging with broader patient populations (from both demographic and geographical standpoints), as well as of exploring historical changes in patient experience and preferences in a way that is not feasible using existing methods. However, achieving the promise of SML is hampered by the significant human resources that are still required to conduct qualitative analysis of social media data, given the large volume of text obtainable from this source. In order to facilitate the use of SML by health system stakeholders engaged in patient-oriented research, we developed EXPECT-NLP (EXploration of Patient Experiences in Collected Texts using Natural Language Processing), a software tool that can automatically extract key themes from patient-generated discussions, groups them by sentiment, relatedness, and topic, and allows users to explore an underlying corpus of text using a graphical user interface. Our hope is that this will allow potential users to quickly leverage the vast amount of social media data available to generate insights and hypotheses on patient experiences and preferences that will inform the development of new medical products, health services, and policies.

Cook NS, Cave J, Holtorf AP. Patient Preference Studies During Early Drug Development: Aligning Stakeholders to Ensure Development Plans Meet Patient Needs. Front Med (Lausanne). 2019 Apr 24;6:82. doi: 10.3389/fmed.2019.00082. PMID: 31069227; PMCID: PMC6491461.

Project Leads: Nick Dragojlovic and Larry Lynd

Methods Matters Webinar Recording & Slides


Download the slides

Methods Matters Webinar Highlights

Unexpected challenges adapting non-health tools for health research

Beyond research: how patients can use "social media listening"

Further uses of "social media listening" - COVID vaccines & student experience

Project Themes


dark purple bar


 Patient and Public Values in Research



Health Economics and Simulation Modelling
This project is a part of the Health Economics and Simulation Modelling cluster to learn more about this cluster, please click here.