Latent topic discovery of clinical concepts from hospital discharge summaries of a heterogeneous patient cohort
Name
nihms773968.pdf
Size
105.65 KB
Format
Adobe PDF
Checksum (MD5)
6311036e98645cda946b8463e2052e61
Author(s)
Saeed, Mohammed
Lehman, Li-Wei
Long, William F.
Mark, Roger G
Date Issued
November 2014
Journal
2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Citation
Lehman, Li-wei et al. “Latent Topic Discovery of Clinical Concepts from Hospital Discharge Summaries of a Heterogeneous Patient Cohort.” 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, August 26-30 2014, Chicago, Illinois, USA, Institute of Electrical and Electronics Engineers (IEEE), November 2014 © 2014 Institute of Electrical and Electronics Engineers (IEEE)
Version
Author's final manuscript
Abstract
Patients in critical care often exhibit complex disease patterns. A fundamental challenge in clinical research is to identify clinical features that may be characteristic of adverse patient outcomes. In this work, we propose a data-driven approach for phenotype discovery of patients in critical care. We used Hierarchical Dirichlet Process (HDP) as a non-parametric topic modeling technique to automatically discover the latent "topic" structure of diseases, symptoms, and findings documented in hospital discharge summaries. We show that the latent topic structure can be used to reveal phenotypic patterns of diseases and symptoms shared across subgroups of a patient cohort, and may contain prognostic value in stratifying patients' post hospital discharge mortality risks. Using discharge summaries of a large patient cohort from the MIMIC II database, we evaluate the clinical utility of the discovered topic structure in identifying patients who are at high risk of mortality within one year post hospital discharge. We demonstrate that the learned topic structure has statistically significant associations with mortality post hospital discharge, and may provide valuable insights in defining new feature sets for predicting patient outcomes.
MIT Department
Institute for Medical Engineering and Science
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1109/EMBC.2014.6943952
