Talking about diseases; developing a model of patient and public-prioritised disease phenotypes

Deep phenotyping uses standardised terminologies to create comprehensive phenotypic descriptions that facilitate secondary analysis, evidence synthesis, and clinical awareness. Most existing disease–phenotype knowledge bases reflect an academic perspective derived from literature and experimental databases, potentially misaligned with patient and public priorities. Using social media data, the authors develop a phenotype model representing a public perspective on disease and compare it to a model constructed from biomedical databases and literature. From social media, 52,108 possible disease–phenotype associations were identified across 311 diseases. A total of 24,618 novel phenotype associations not present in the biomedical/literature model were found across 304 diseases, of which 14,531 were significant. Social media-derived phenotypes over-represented manifestations affecting quality of life and phenotypes related to endocrine, digestive, and reproductive systems. Clinical expert review judged social media-derived associations similarly well-established as literature-derived associations and noted differences in their presence in clinical encounters. The social media phenotype model yields a perspective that differs significantly from existing resources and provides numerous novel associations. Integrating public perspectives into phenotype resources may enhance clinical awareness, support secondary analyses, and bridge understanding across healthcare stakeholders.

Karin Slater, Paul N. Schofield, James Wright, Paul Clifft, Anushka Irani, William Bradlow, Furqan Aziz, Georgios V. Gkoutos