Smartphone-based digital phenotyping for dry eye toward P4 medicine: a crowdsourced cross-sectional study

• DOI: 10.1038/s41746-021-00540-2
• Publication Date: 2021-12-20T11:13:41Z
• Authors: Takenori Inomata; Masahiro Nakamura; Jaemyoung Sung; Akie Midorikawa-I...; Masao Iwagami; Kenta Fujio; Yasutsugu Akasaki; Yuichi Okumura; Keiichi Fujimoto; Atsuko Eguchi; Maria Miura; Ken Nagino; Hurramhon Shokirova; Jun Zhu; Mizu Kuwahara; Kunihiko Hirosawa; Reza Dana; Akira Murakami
• Keywords: Interpretability; Cross-validation

Multidimensional integrative data analysis of digital phenotyping is crucial for elucidating the pathologies multifactorial and heterogeneous diseases, such as dry eye (DE). This crowdsourced cross-sectional study explored a novel smartphone-based strategy to stratify visualize heterogenous DE symptoms into distinct subgroups. were collected from 3,593 participants between November 2016 September 2019. Dimension reduction via Uniform Manifold Approximation Projection stratified seven clusters symptomatic DE. Symptom profiles risk factors in each cluster identified by hierarchical heatmaps multivariate logistic regressions. Stratified subgroups visualized chord diagrams, co-occurrence networks, Circos plot analyses improve interpretability. Maximum blink interval was reduced 1, 2, 5 compared non-symptomatic Clusters 1 had severe symptoms. A data-driven multidimensional with may establish predictive, preventive, personalized, participatory medicine.